Measurement of exhaled nitric oxide concentration in asthma: a systematic review and economic evaluation of NIOX MINO, NIOX VERO and Nobreath

Introduction

Asthma is a chronic disorder of the airways, caused primarily by inflammatory processes and constriction of the smooth muscle in airway walls (bronchoconstriction). It is characterised by airflow obstruction and increased responsiveness of the airways to various stimuli. Symptoms include recurrent episodes of wheezing, breathlessness, chest tightness and coughing. Typical asthma symptoms tend to be variable, intermittent and worse at night. Asthma is commonly triggered by viral respiratory infections, exercise or external factors such as smoke, a change in weather conditions and allergens, for instance pollen, mould and house dust mites.

Asthma usually develops in childhood but may start at any age. It runs in some families but many people with asthma have no other family members affected. In adults, asthma is more common in women than in men. 1 There is no cure for asthma, although people may experience long periods of remission. Poorly controlled asthma can have a significant impact on the quality of life of the affected individual and his or her family. However, there may be variation in an individual’s perception of the symptoms and how he or she adapts to the condition over time. Clinical measures such as lung function may not correlate with an individual’s quality of life scores, but if asthma is well controlled near-maximal scores on quality of life instruments can be achieved.

Classification of asthma

There are several ways of categorising different types of asthma, including:

• Intrinsic and extrinsic asthma. Asthma can be divided into extrinsic (external cause) and intrinsic (when no causative agent can be found) asthma. Extrinsic asthma is triggered by allergens and hence it is also termed ‘allergic asthma’. In extrinsic asthma the immune system reacts to substances such as pollen and produces antibodies. Individuals with a predisposition to developing such allergies are said to be atopic and may develop any combination of the triad of hay fever, eczema and asthma. In the case of asthma, the allergic reaction is observed in bronchi and bronchioles, which results in the production of excess mucus that obstructs the air passages. Extrinsic asthma is commonly seen in children. About 90% of childhood asthma cases are caused by specific allergens. Individuals with a family history of atopy are at a higher risk of developing extrinsic asthma. In contrast, intrinsic asthma is a non-seasonal, non-allergic form of asthma, which usually first occurs at a later point in life than allergic asthma. Intrinsic asthma tends to be chronic and persistent rather than episodic. It is not related to specific allergens and may be provoked by the inhalation of chemicals such as cigarette smoke or cleaning agents, non-steroidal anti-inflammatory drugs, chest infections, emotion, exercise, cold air, food preservatives or various other non-specific irritants.

• Eosinophilic and non-eosinophilic asthma (neutrophilic asthma). Asthma can also be categorised as eosinophilic or non-eosinophilic. There is some evidence that eosinophils may play an important proinflammatory role in the pathogenesis of asthma,2,3 although there remains some uncertainty around this and other pathogenic mechanisms associated with asthma. Eosinophils are found in the airways of asthmatics but not healthy subjects and are believed to be related to exacerbations. It has also been noted that suppression of eosinophil infiltration is often associated with amelioration of symptoms2 but that the relationship is not close. Poor inflammation control is most closely related to the risk of future exacerbations. The presence of eosinophils may be used to direct treatment as patients without eosinophilic inflammation are thought to be less responsive to inhaled corticosteroid (ICS) treatment. 4 High levels of eosinophils are correlated with high levels of fractional exhaled nitric oxide (FeNO) and it is thought that FeNO could be used as a biomarker of eosinophilic inflammation and therefore of ICS responsiveness. 5,6 However, the presence of eosinophils is not always a marker of severity of disease; fatal asthma may be associated with neutrophilia rather than eosinophilia. 7 Targeting the type of inflammation may be a better guide to treatment than measures of disease severity alone. For instance, glucocorticosteroids are typically very effective in eosinophilic inflammation but less so if the inflammation is neutrophilic.

• Eosinophilic and non-eosinophilic airway disease. Eosinophilic inflammation occurs in both asthma and chronic obstructive pulmonary disease (COPD) and in both cases the appropriate treatment is ICSs. 6 There is a view held by some clinicians that, rather than a diagnosis of asthma, a diagnosis of responsiveness to ICSs [irrespective of diagnostic label (asthma or COPD)] may be a more helpful approach in terms of directing treatment, reducing costs and reducing exacerbations. 6 However, this form of classification has not yet been officially adopted in the British Thoracic Society (BTS) and Scottish Intercollegiate Guidelines Network (SIGN) guidelines8 and this report will focus on the diagnosis of asthma as described in these guidelines.

• Molecular approaches to classifying asthma phenotypes. There is an increasing trend to characterise asthma by molecular and cellular factors to enable more targeted and personalised therapy. Such efforts are ongoing and specific phenotypes and the implications of these are not yet fully elucidated. 9

• Exercise-induced bronchoconstriction (EIB). Most patients with asthma will experience EIB but approximately 11% of the population without other forms of asthma also experience this. It is characterised by a reduction in the forced expiratory volume in the first second (FEV₁) of > 10% after exercise and can be treated pharmacologically with short-acting beta2-agonists (SABAs) or leukotriene receptor antagonists (LTRAs) and non-pharmacologically with a light warm-up before vigorous exercise for example. The exact mechanisms behind EIB are not fully understood but may include neural and biochemical mediators. 10

Impact of asthma on patients

The principal symptoms of asthma are wheezing attacks and episodic shortness of breath. An acute onset of symptoms is known as an exacerbation. Coughing, which worsens at night, may also be a symptom. Asthma exacerbations tend to vary considerably in terms of frequency and duration. Some people experience one or two per year lasting for a few hours, whereas others have exacerbations lasting for weeks or experience them more frequently. Exacerbations may be precipitated by a wide range of triggers, as described in Classification of asthma. Asthma is a major cause of impaired quality of life and may impact on a patient’s work, recreational activities, physical activities and emotions. However, although patients’ health-related quality of life (HRQoL) may be impacted on by poor asthma control and the incidence of exacerbations, it has been noted elsewhere that meeting clinical treatment goals may not result in noticeable changes in a patient’s quality of life. 13

In the long term, asthma may lead to permanent airflow obstruction and associated loss of quality of life, especially when it is persistent or poorly controlled. 19 Asthma also has a substantial impact on a patient’s ability to work and study and has been estimated to result in at least 12.7 million lost working days per year. 12 Many patients will undergo regular monitoring and will be required to take medication for the rest of their life. There have been concerns that long-term ICS use may reduce growth rates in children, although evidence is conflicting and it appears that any reduction in growth may be transient, with patients eventually achieving a normal adult height. 20,21

Burden on the NHS

Given the high prevalence of people with asthma, asthma treatment represents a significant cost to the NHS. The Health Survey for England 2010 estimated that direct health-care costs associated with asthma are £1B per year. 22 In addition, estimates from 2002 indicate that general practitioner (GP) prescriptions alone are worth approximately £600M per year. 22

As asthma is an incurable condition, treatment, or at the least monitoring, is usually required for the remainder of the patient’s lifetime. However, as the diagnosis of asthma is not definitive there is the potential for misdiagnoses to go undetected for many years or even an entire lifetime. Misdiagnosis can occur when a patient appears to respond to treatment but in fact has experienced a natural resolution of the symptoms of another underlying condition such as a cold, a respiratory infection or allergy. In these cases, patients will appear well controlled and a treating physician may simply assume that the treatment is working. The BTS/SIGN guidelines8 recommend that patients who are well controlled should ‘step down’ their therapy dose. This could result in a patient being taken off treatment altogether and their diagnosis being reconsidered. However, clinical input to this review suggests that step down of doses does not always occur as treatment is relatively cheap per patient and physicians are cautious not to risk exacerbations. As such, there may be long-term unnecessary NHS expenditure associated with these misdiagnoses. Similarly, both overtreatment and undertreatment of patients who have been correctly diagnosed with asthma may be sources of substantial NHS expenditure. Undertreatment may increase costs to the NHS as poor control may lead to an increased rate of severe exacerbations, which require additional primary care management and acute hospital admissions. Overtreatment may increase costs to the NHS because a patient may be able to receive the same level of symptom control with less medication and so the condition could have been treated as effectively at a lower cost.

Diagnosis of asthma

The diagnosis of asthma is a clinical one and there is no standardised definition of the condition. Central to all definitions in adults is the presence of symptoms (wheezing, breathlessness, chest tightness and cough) and of variable airflow obstruction measured through objective tests of lung function [such as peak expiratory flow rate (PEFR) and FEV₁ divided by forced vital capacity (FVC), known as the Tiffeneau–Pinelli index (FEV₁/FVC)] and percentage of predicted FEV₁ (FEV₁%; calculated as the percentage of the predicted FEV₁ for a person of the same height, sex and age without diagnosed asthma). Variability in PEFR and FEV₁, either spontaneously or in response to therapy, is a characteristic feature of asthma. The BTS/SIGN guidelines8 indicate that the severity of asthma should be judged according to symptoms and the amount of medication required to control symptoms.

More recently, descriptions of asthma have included airway hyper-responsiveness and airway inflammation. It is unclear how these features relate to each other, how they are best measured and how they contribute to the clinical manifestations of asthma.

Figures 3 and 4 present the diagnostic pathways for children and adults, respectively, as they currently stand. 8

FIGURE 3.

Diagnosis of asthma in children according to BTS/SIGN guidelines. 8 –ve, negative; +ve, positive. a, Lung function tests include spirometry before and after bronchodilator (test of airway reversibility) and possible exercise or methacholine challenge (tests of airway responsiveness). Most children over the age of 5 years can perform lung function tests. Reproduced with permission from BTS/SIGN. British Guideline on the Management of Asthma: a National Clinical Guideline. Edinburgh and London: BTS/SIGN; 2012. 8

FIGURE 4.

Diagnosis of asthma in adults according to BTS/SIGN guidelines. 8 PEF, peak expiratory flow. a, See section 2.5.1 of the BTS/SIGN guidelines; b, see table 6 of the BTS/SIGN guidelines. Reproduced with permission from BTS/SIGN. British Guideline on the Management of Asthma: a National Clinical Guideline. Edinburgh and London: BTS/SIGN; 2012. 8

Diagnosis in children is clinically based on recognising a characteristic pattern of episodic symptoms in the absence of an alternative explanation. Lung function tests are less useful because of variability and the inability of very young children to perform these tests reliably. According to the BTS/SIGN guidelines,8 clinical features that increase the probability of asthma include:

1. More than one of the following symptoms: wheeze, cough, difficulty breathing, chest tightness, particularly if these symptoms:

   • are frequent and recurrent

   • are worse at night and in the early morning

   • occur in response to, or are worse after, exercise or other triggers, such as exposure to pets, cold or damp air, or with emotions or laughter

   • occur apart from colds

2. Personal history of atopic disorder

3. Family history of atopic disorder and/or asthma

4. Widespread wheeze heard on auscultation

5. History of improvement in symptoms or lung function in response to adequate therapy.

Reproduced with permission from BTS/SIGN guidelines8

If asthma is suspected, an initial clinical assessment should be carried out to estimate the probability of asthma. According to the BTS/SIGN guidelines,8 based on initial clinical assessment a child can be classified according to their risk of having asthma as:

• high probability, where an asthma diagnosis is likely

• low probability, where a diagnosis other than asthma is likely

• intermediate probability, where the likely diagnosis is uncertain.

For children identified as having a low probability of asthma, a more detailed investigation and specialist referral should be considered. For children with a high probability of asthma, a trial of treatment should be started immediately, with review at 6–8 weeks. When the response is good, the ICS dose should be reassessed every 6 months. Those with a poor response to treatment should undergo more detailed investigations.

There is insufficient evidence at first consultation to make a firm diagnosis of asthma in some children, particularly those aged < 4–5 years. 8 For those children who can perform spirometry and for whom airway obstruction is evident, change in forced expiratory flow volume or peak expiratory flow monitoring should be assessed in response to an inhaled bronchodilator and/or in response to a trial of treatment for a specified period.

In children with an intermediate probability of asthma who can perform spirometry and who have no evidence of airway obstruction, tests for atopic status, assessment of bronchodilator reversibility and, if possible, assessment of bronchial hyper-responsiveness using methacholine, exercise or mannitol should be considered, although these last three would be performed in secondary care. In such cases specialist referral should always be considered.

Other investigations to support a diagnosis of, or alternatively rule out, asthma in children include tests of eosinophilic airway inflammation using induced sputum or exhaled nitric oxide concentrations, tests of atopy by skin prick test or blood eosinophilia and chest radiography or other imaging techniques to investigate other causes.

Diagnosis in adults is also based on clinical history and includes the recognition of a characteristic pattern of symptoms and signs and the absence of an alternative explanation for them. However, in contrast to the diagnostic pathway for children, in adults spirometry is performed at the first consultation to assess the presence and severity of airflow obstruction.

As in the diagnosis of children, adults are also classified as having a high, low or intermediate probability of asthma. Chest radiography and specialist referral may be considered in any patient presenting atypically or with additional symptoms or signs.

Monitoring and management of diagnosed asthma

Asthma management aims to control symptoms (including nocturnal symptoms and exercise-induced asthma), prevent exacerbations and achieve the best possible lung function, with minimal side effects of treatment. For both children and adults, asthma is monitored and managed in primary care by routine clinical review on at least an annual basis. These reviews include (but are not limited to) assessment of a patient’s symptom score (using a validated questionnaire), exacerbations, oral corticosteroid (OCS) use, time off school or work, growth and inhaler technique; in adults, lung function is also assessed by spirometry of peak expiratory flow. Patients are managed in a stepwise manner, with escalation of medication until control is reached. This approach to pharmacological management for children and adults is represented in Figures 5 and 6 respectively. 8 Treatment is started at the step most appropriate to the initial severity of the asthma, with the aim of achieving early control of symptoms and optimising respiratory function. Control is maintained by stepping up treatment as necessary and stepping down when control is good.

FIGURE 5.

Management of asthma in children. BDP, beclomethasone dipropionate. Reproduced with permission from BTS/SIGN. British Guideline on the Management of Asthma: a National Clinical Guideline. Edinburgh and London: BTS/SIGN; 2012. 8

FIGURE 6.

Management of asthma in adults. BDP, beclomethasone dipropionate. Reproduced with permission from BTS/SIGN. British Guideline on the Management of Asthma: a National Clinical Guideline. Edinburgh and London: BTS/SIGN; 2012. 8

The potential role of FeNO devices in the diagnosis and management of asthma

Nitric oxide monitors measure FeNO. High FeNO levels in a patient with symptoms suggestive of asthma, such as coughing and wheezing, may suggest that the patient has eosinophilic asthma that could be treated with ICSs (see Classification of asthma). In individuals already diagnosed with asthma, changes in FeNO levels may indicate how well a patient is responding to ICS-based medication, whether medication is being adhered to and whether the dosage of medication should be increased or decreased (titrated or step-up/step-down adjustment). Consequently, FeNO monitors may have a role in the diagnosis, monitoring and management of patients with asthma.

However, current opinion is divided as to the utility of this measurement, in large part because of the potential for various factors to confound FeNO levels. Amongst these are age, sex, smoking status, exposure to environmental tobacco, pregnancy, height, measurement technique and atopic status and medication. 23,24 A further consideration is the observation that the dose–response plateaus within the therapeutic range of ICSs,25,26 although doses up to 800 µg of beclomethasone dipropionate have been reported to be distinguishable from placebo. 27

Current service provision

A number of FeNO devices have been developed. Some of these are hand-held portable devices (such as the devices that are the focus of this assessment) and others are stationary devices that measure FeNO through chemiluminescent techniques. Both types of FeNO monitor have been available for use in the NHS for a number of years. However, they are not available in all secondary care settings and their use in primary care is extremely rare. There are a number of possible reasons why FeNO devices have not had a more widespread diffusion into care, including the lack of clear guidance in the BTS/SIGN guidelines8 on how they should be used, which itself is a consequence of contradictory research, and the previously prohibitive cost and operational requirements of large chemiluminescent devices.

A number of other diagnostic interventions are commonly used in the diagnosis of asthma in England and Wales, as described in Diagnosis of asthma. Some of these are performed in primary care, such as spirometry, reversibility testing and trials of treatment, whereas others are performed in secondary care, such as airway hyper-responsiveness [methacholine challenge test (MCT)] and sputum induction. As noted earlier, monitoring and management of asthma in diagnosed patients is guided by BTS/SIGN guidelines. 8

Technologies under assessment

The three hand-held FeNO devices included in this assessment are NIOX MINO^® (Aerocrine, Solna, Sweden), NIOX VERO^® (Aerocrine) and NObreath^® (Bedfont Scientific, Maidstone, UK).

Anticipated costs associated with the intervention

The marginal per-test costs of each of the three technologies considered within this assessment depend on both fixed costs, such as the initial cost of the devices, and variable costs, such as the costs of consumables.

The NIOX MINO device has a unit cost of £2100 and has an effective unit lifetime of 3 years or 3000 tests (whichever comes first). The NIOX VERO device has a unit cost of £2310 and has an effective unit lifetime of 5 years or 5000 tests (whichever comes first). The NObreath device costs £1995 and has an unlimited unit lifetime. Maintenance for the NObreath device is provided free of charge by Bedfont Scientific.

Test kits for NIOX MINO are available in packs of 300 at a price of £1350, packs of 500 at a price of £2100 or packs of 1000 at a price of £3950. Test kits for NIOX VERO are available in packs of 300 at a price of £1500, packs of 500 at a price of £2200 or packs of 1000 at a price of £4200. Mouthpieces for NObreath are available in packs of 50, 100, 300 or 1000 at prices of £195, £365, £995 and £2995 respectively.

The NObreath device requires replacement of the sensor unit every 2 years at a cost of £295. Besides test kits, there are no other replacement costs for the NIOX MINO and NIOX VERO devices.

This information is summarised in Table 2.

Definition of the interventions

Two monitors were identified at the scoping stage for this appraisal: NIOX MINO, which is manufactured by Aerocrine, and NObreath, which is manufactured by Bedfont Scientific. During the latter stages of the assessment, Aerocrine alerted the EAG to a follow-up device to NIOX MINO, the NIOX VERO device. This device is also considered within this assessment although the evidence base is limited. All three interventions are evaluated in the context of the diagnosis and management of asthma.

Populations and relevant subgroups

Comparators

The relevant comparators are diagnosis or management according to the current UK guidelines, as described in Chapter 3. In the diagnostic setting, the relevant comparator consists of the current diagnostic pathway without the use of FeNO measurements; this is different for children and adults (see Guidelines for the diagnosis and management of asthma). In the management setting, the relevant comparator is management according to current guidelines without the use of FeNO.

Relevant outcomes for the assessment

The assessment includes consideration of the available evidence across a wide range of clinical and economic outcomes.

Place of the intervention in the diagnostic/treatment pathways

During the scoping phase of this appraisal, workshop attendees considered that the interventions should be assessed when added to current practice. There are a number of potential places within the diagnostic/treatment pathways where FeNO may be of clinical use and each is likely to have different consequences for clinical effectiveness and cost-effectiveness.

Position of FeNO in the diagnostic pathway: children

During the scoping workshop it was agreed that FeNO is likely to be of most use in positions 1, 2 and 3 in Figure 7. This figure is based on the BTS/SIGN clinical guidelines,8 with input from a clinician about how the tests are used in practice (Dr John White, York Teaching Hospital NHS Foundation Trust, 17 July 2013, personal communication). This equates to patients who are difficult to diagnose. Depending on whether FeNO is used as a direct replacement for a test or as a rule-in or rule-out test at these positions in the pathway, it may have the ability to prevent expensive secondary care visits if used in primary care. In secondary care it may have additional value alone or in conjunction with existing secondary care tests. FeNO could also be considered to replace the whole pathway or be inserted at other points along the pathway. Tables 3–5 detail the actions and consequences associated with some different replacement and rule-in/rule-out scenarios. In rule-in scenarios, patients testing positive are assumed to have asthma and those testing negative go on to have further tests for asthma. In rule-out scenarios, those who test negative are assumed not to have asthma and those who test positive go on to have further tests for asthma.

FIGURE 7.

Potential positions for FeNO in the diagnostic pathway: children. Source: BTS/SIGN guidelines8 with clinical input from Dr John White (17 July 2013, personal communication).

TABLE 3 - Consequences of using FeNO as a direct replacement for the whole pathway or for airway hyper-responsiveness in patients indicated for this test within the pathway: adults and children

Replacement scenario	FeNO measurement	Action taken	Consequence 1	Consequence 2
TP	High FeNO measurement	Treat as asthma	Correct diagnosis of asthma reached	None
FP			Patient’s misdiagnosis goes undetected until worsening of symptoms or routine review or continues lifelong	None
TN	Low FeNO measurement	Treat as not asthma	Further tests for other conditions	Correct diagnosis reached
FN				Further tests negative, re-enter asthma pathway or remain misdiagnosed until exacerbation or return to GP with ongoing symptoms

TABLE 4 - Consequences of using FeNO as a rule-out test before airway hyper-responsiveness: adults and children

Rule-out scenario	FeNO measurement	Action taken	Consequence 1	Consequence 2
TP	High FeNO measurement	Treat as possibly asthma and undertake further confirmatory tests	Further tests confirm asthma diagnosis	Treat as asthma
FP			Further tests reject asthma diagnosis	Further tests for other conditions or diagnose as non-specific symptoms
TN	Low FeNO measurement	Treat as not asthma	Further tests for other conditions	Correct diagnosis reached
FN				Further tests negative, re-enter asthma pathway or remain misdiagnosed until exacerbation or return to GP with ongoing symptoms

TABLE 5 - Consequences of using FeNO as a rule-in test before airway hyper-responsiveness: adults and children

Rule-in scenario	FeNO measurement	Action taken	Consequence 1	Consequence 2
TP	High FeNO measurement	Treat as asthma	Correct diagnosis of asthma reached	None
FP			Patient’s misdiagnosis goes undetected until worsening of symptoms or routine review or continues lifelong	None
TN	Low FeNO measurement	Further tests for asthma	Tests for asthma negative	Further tests for other conditions or diagnose as non-specific symptoms
FN			Correct diagnosis of asthma reached	None

Position of FeNO in the diagnostic pathway: adults

For the diagnostic pathway in adults, FeNO is thought to be of most use in positions 1 and 2 in Figure 8. This equates to patients who are difficult to diagnose. This figure is based on the BTS/SIGN clinical guidelines,8 with input from a clinician about how the tests are used in practice (Dr John White, 17 July 2013, personal communication). This led to the understanding that, in nearly all or at least most cases, patients would undergo a trial of treatment or airway reversibility testing before being referred to secondary care, regardless of their FEV₁/FVC ratio. This is slightly different from our initial reading of the BTS/SIGN guidelines, in which only patients with a FEV₁/FVC ratio of < 0.7 would undergo these tests, with those with a FEV₁/FVC ratio of > 0.7 going on to secondary care for airway hyper-responsiveness testing. Our initial diagrammatic representation of the adult pathway can be viewed on the NICE website (www.nice.org.uk/guidance/dg12/documents/measurement-of-exhaled-nitric-oxide-concentration-in-asthma-niox-mino-and-nobreath-final-protocol2).

FIGURE 8.

Potential positions for FeNO in the diagnostic pathway: adults. PEF, peak expiratory flow. Source: BTS/SIGN guidelines8 with clinical input from Dr John White (17 July 2013, personal communication).

Depending on whether FeNO is used as a direct replacement for an existing test or as a rule-in or rule-out test at these positions in the pathway, it may have the ability to prevent expensive secondary care visits if used in primary care. In secondary care it may have additional value alone or in conjunction with existing secondary care tests. FeNO could also be considered to replace the whole pathway or be inserted at other points along the pathway.

Clinical evidence review

Two systematic reviews and one rapid review were conducted concurrently to identify clinical evidence relevant to the decision problem:

• Rapid review of the equivalence of FeNO devices. It was not clear at the outset if there would be sufficient primary research evidence relating to the three devices to inform the appraisal. As such, a review of the equivalence of these devices to other FeNO measurement devices was anticipated and appropriate searches were conducted. The review of equivalence was conducted in full when it became apparent that sufficient evidence was not available from the diagnostic accuracy review and management efficacy review. The equivalence review aimed to establish whether measurements from different FeNO measurement devices could be considered to be equivalent to one another and therefore whether studies that used other devices could helpfully inform this appraisal. This review was thought to be the least critical in terms of informing key model inputs and a rapid review using systematic methods was therefore conducted because of time and resource constraints. This represents a change to the published assessment protocol. 30 (www.nice.org.uk/guidance/dg12/documents/measurement-of-exhaled-nitric-oxide-concentration-in-asthma-niox-mino-and-nobreath-final-protocol2.)

• Systematic review of the diagnostic accuracy of FeNO measurements for asthma. The ideal study would recruit patients with symptoms of asthma, have a cohort design or randomise patients to diagnosis using FeNO or diagnosis using other methods and follow them to clinical outcomes. Such studies are known as end-to-end studies and demonstrate the ability of the test to improve patient outcomes. In the absence of such studies, diagnostic cohort studies represent the next best level of evidence, with modelling of clinical outcomes based on the numbers of patients classed as TP, TN, FP and FN. Below this are correlation studies. All levels of evidence were searched for in this review; lower levels of evidence were consulted when the higher levels of evidence were not identified. When available, three pairs of sensitivity and specificity values were selected: those that produced the highest sum of sensitivity and specificity; those that had the highest sensitivity for rule-in scenarios; and those that had the highest specificity for rule-out scenarios. In rule-in scenarios, patients testing positive are assumed to have asthma and those testing negative go on to have further tests for asthma. In rule-out scenarios, those who test negative are assumed not to have asthma and those who test positive go on to have further tests for asthma.

• Systematic review of the efficacy of FeNO-guided management of asthma. Existing systematic reviews of randomised controlled trial (RCT) evidence in adults31 and children32,33 meant that only RCT evidence was searched for in this review, with additional interrogation of the database for data on subgroups when RCT evidence was not found.

Cost-effectiveness assessment

The cost-effectiveness assessment of FeNO includes two components: a systematic review of existing economic analyses and the development of two de novo health economic models:

• Systematic review of the cost-effectiveness of FeNO for the diagnosis and/or management of asthma. A systematic review was undertaken to identify all existing economic analyses of FeNO testing for asthma; this includes published studies as well as evidence submitted by the manufacturers of NIOX MINO, NIOX VERO and NObreath. This included a critical appraisal of the available evidence and a summary of methodological problems and concerns relating to these analyses.

• Development of two de novo models. Independent health economic models were developed to assess the incremental cost-effectiveness of FeNO compared with standard care in the diagnosis and management of asthma.

Search methodology for the clinical reviews

Systematic searches were carried out between March 2013 and April 2013. Update searches were conducted in September 2013 for the diagnostic and management reviews. For the review of device equivalence and for both diagnostic and management reviews, the following databases were searched:

• MEDLINE and MEDLINE In-Process & Other Non-Indexed Citations (Ovid): 1948–present

• EMBASE (Ovid): 1974–present

• The Cochrane Library (Wiley Interscience):

  • Cochrane Database of Systematic Reviews (CDSR): 1996–present

  • Database of Abstracts of Reviews of Effects (DARE): 1995–present

  • Cochrane Central Register of Controlled Trials (CCRCT): 1898–present

  • Health Technology Assessment (HTA) database: 1995–present

  • NHS Economic Evaluation Database (NHS EED): 1995–present

• Science Citation Index Expanded (SCIE) (Web of Science): 1899–present

• Conference Proceedings Citation Index – Science (CPCI-S) (Web of Science): 1990–present.

The search strategies used in MEDLINE are provided in Appendix 1.

The following trial registers and websites were searched in March 2013 for all three reviews and again in September 2013 for the diagnostic and management reviews (search terms used are provided in Appendix 1):

• ClinicalTrials.gov (http://clinicaltrials.gov/)

• metaRegister of Controlled Trials (www.controlled-trials.com/mrct/)

• US Food and Drug Administration Manufacturer and User Facility Device Experience (MAUDE) database (www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfmaude/search.cfm)

• EuroScan International Network (http://euroscan.org.uk/).

Management review searches

Searches for the management review were developed following the identification of a 2009 Cochrane review. 31 Study design filters were not applied to the strategy in case lower levels of evidence were needed for the subgroups defined a priori in the protocol. The strategy (Figure 9) was made up of (1) free-text terms for NIOX MINO and NObreath (including manufacturer names), (2) subject heading and free-text terms for asthma (e.g. respiratory hypersensitivity, bronchoconstriction) and (3) subject heading and free-text terms for lower respiratory tract symptoms (e.g. coughing, wheezing, chest pain). Search strings (2) and (3) were combined with subject heading and free-text terms for exhaled nitric oxide and the results were added to the results for search string (1). Searches were limited to publications since 2009.

FIGURE 9.

Management and diagnostic studies of FeNO devices.

A summary of the search records retrieved from the searches is provided in Table 6.

TABLE 6 - Search records retrieved by database: management review

Database	Number of records
MEDLINE and MEDLINE In-Process & Other Non-Indexed Citations	991
EMBASE	2269
CDSR	44
DARE	1
CCRCT	117
HTA database	8
SCIE	1387
CPCI-S	70
Total unique references	2747

Equivalence of devices review searches

The analytical validity study searches for NIOX MINO and NObreath were carried out using terms for NIOX MINO and NObreath and the manufacturer names without any application of filters and limits in the databases listed (Figure 10). The numbers of records retrieved by database are provided in Table 7 (final column).

FIGURE 10.

Equivalence studies of FeNO devices.

Study selection

Retrieved citations were considered for inclusion in several stages. First, titles were considered and any studies obviously not relevant were excluded. Second, abstracts were consulted. At this stage, tags were applied to studies in Reference Manager to identify the device used, the age group of the participants and the study design. In instances in which it was obvious which review the study was likely to inform, this tag was also applied. In the third stage, articles tagged as the highest levels of evidence for each review were retrieved and the full texts were compared against the inclusion and exclusion criteria.

Once the full-text selection process was complete, a decision was made whether there were gaps in the evidence that would require lower levels of evidence to be consulted. This was the case for the diagnostic review, in which no end-to-end studies were identified; for the management review, in which only limited evidence was identified for NIOX MINO and no evidence was identified for NObreath; and for some of the subgroups of interest to the review. For the diagnostic review, studies including any device were included rather than just those using NIOX MINO, NIOX VERO or NObreath (see Review of the diagnostic accuracy of FeNO testing for asthma) and, for the management review, studies using any FeNO device were included (see Review of the efficacy of FeNO-guided management of asthma); the rapid review of the equivalence of devices was conducted in full (see Review of the equivalence of devices). To retrieve relevant titles from the database for the subgroups of interest to the review, the keyword search facility in Reference Manager was used to search for the following keywords:

• elderly asthmatics: elderly, old, older and elderly care

• smokers: smoke, smoking, smoking.adverse effects, smoking.epidemiology, smoking cessation, smoking cessation programme, smoking habit, smoking/ae [adverse drug reaction] and smoking: epidemiology

• pregnant women: pregnant, pregnancy, expectant, pregnancy complication/co [complication], pregnancy complication/si [side effect], pregnancy complications, pregnancy diabetes mellitus, pregnancy diabetes mellitus/dt [drug therapy], pregnancy outcome, pregnancy test and pregnant women.

These titles were then sifted by title, abstract and full text for inclusion in the review with relation to criteria for population, intervention and comparator. Criteria on study design and specific outcomes were relaxed and studies of the next best level of evidence that provided data evaluating the use of FeNO measurements in appropriate subgroups were included. The hierarchy of evidence used was as described in the NICE guidelines methods guide. 34

Data extraction

A different standardised data extraction form was developed for each review following the guidelines given in the Cochrane Handbook for Systematic Reviews of Interventions36 and the Centre for Reviews and Dissemination (CRD) Guidance for Undertaking Reviews in Healthcare;37 these forms were piloted using two studies per review. Missing fields were added as appropriate and backfilled where necessary. Appendix 3 lists the fields that were data extracted for each review. Data were extracted from the studies by one of three reviewers and checked by a second reviewer (SH, ME, TG), except for the rapid review of the equivalence of devices for which a sole reviewer (SH) extracted all relevant data. Any discrepancies were resolved by discussion, with involvement of a third reviewer when necessary. When appropriate, authors were contacted for missing or unclear data. Data from multiple publications of the same study were extracted and quality assessed as a single study. In a change from the protocol, data were not extracted from existing systematic reviews, but directly from the primary research journal articles and conference abstracts.

Quality assessment

As it was a rapid review, quality assessment was not conducted for the review of the equivalence of devices.

Diagnostic cohort studies were assessed using the Quality Assessment of Diagnostic Accuracy Studies – second revision (QUADAS-2) tool. 38 The tool was adapted to the specifics of this appraisal and the scoring scheme can be found in Appendix 4. Because of the complexity of this assessment, items within QUADAS-2 that related to applicability were omitted and this was addressed in detail as follows:

1. Are there concerns that the included patients do not match the review question? – addressed through the subcategorisation of studies according to patient characteristics.

2. Are there concerns that the index test, its conduct or its interpretation differs from the review question? – addressed through a review of the equivalence of devices and through the selection of studies that recorded FeNO according to American Thoracic Society (ATS) guidelines. 35

3. Are there concerns that the target condition as defined by the reference standard does not match the review question? – addressed through the subcategorisation of studies according to the reference standard used.

Management RCT studies were assessed using domains listed in the Cochrane risk of bias tool. 36 The scoring scheme can also be found in Appendix 4.

Studies of lower quality were not formally quality assessed but were considered on their individual merits.

Quality assessment was conducted by one reviewer and checked by a second. A third reviewer was consulted in cases of disagreement.

Analysis and synthesis

A narrative synthesis was conducted for the rapid review of the equivalence of devices and no meta-analysis was planned or attempted.

A narrative synthesis was conducted for the review of diagnostic studies. A meta-analysis was planned if sufficient studies of acceptable clinical heterogeneity in terms of patient populations, devices, cut-off points and reference standards were available. A meta-regression to allow the use of multiple cut-off points in the modelling was planned, again if the necessary data were available with appropriate levels of heterogeneity between studies. However, data were not suitable for meta-analysis or meta-regression.

A narrative synthesis was conducted for the review of management studies. A meta-analysis was planned if enough studies of acceptable clinical heterogeneity in terms of patient populations, devices, cut-off points, interventions, comparators and outcomes were available. Clinical heterogeneity indicated that such an analysis was unlikely to produce meaningful results, but exploratory analyses and sensitivity analyses in relation to elements of study design were conducted in the review of adult studies, even though clinical heterogeneity was high. For rate outcomes, the generic inverse variance method was used in Review Manager version 5.3 (The Cochrane Collaboration, The Nordic Cochrane Centre, Copenhagen, Denmark) to meta-analyse rate ratios. For continuous outcomes, a standardised mean difference analysis was conducted as metrics for ICS use were different.

In all cases, fixed effects were used first and random effects were applied if the I² statistic indicated that heterogeneity was moderate or high. This was judged to be the case at > 40%.

Equivalence of devices (analytical validity)

A total of 27 studies [30 citations39–68 plus two sponsors’ submissions (Fukuhara 2013 and Hedlund 2013)] comparing the intervention devices (NIOX MINO, NIOX VERO and NObreath) with other devices were included in the review. One additional study127 was excluded as it compared NIOX MINO with another hand-held device (NoVario; FILT, Berlin, Germany) not in the scope of this appraisal. The studies have been categorised for presentation and discussion according to the devices compared and population age ranges as follows:

• NIOX MINO compared with the Niox chemiluminescent device (Aerocrine) in adults

• NIOX MINO compared with the Niox chemiluminescent device in children

• NIOX MINO compared with other stationary chemiluminescent devices in adults and/or children

• NIOX VERO compared with NIOX MINO

• NObreath compared with other stationary chemiluminescent devices in adults and/or children

• NIOX MINO compared with NObreath in adults and/or children

• area under the curve (AUC), cut-off points and correction equations

• test failure rates

• conclusions.

Three main comparisons were considered in this review:

• Comparison of means – comparison between reported mean FeNO values as measured by each device in the same cohort. This comparison may be confounded by natural within-patient variance between measurements by the two devices.

• Correlation coefficients – these show whether measurements by the two devices are correlated but not whether the actual values produced are the same (agreement). Highly correlated devices might produce slopes on a graph (plotting FeNO measurement against a known FeNO concentration) of the same gradient but at different heights, indicating that one device measures consistently higher or lower than another. Correlation coefficients can be confounded by the fact that comparison over wider ranges of values can lead to higher correlation values. 128

• Bland–Altman analysis – produces a number of useful comparison statistics that assess agreement between devices rather than just correlation. Bland–Altman plots128 plot the mean of two measurements by two devices (x-axis) against the difference between the measurements (y-axis). If the devices agreed perfectly across the whole range of measurements, all points would be at point zero on the y-axis across the range of measurements. However, if agreement is not perfect, the points will fall above and below zero. If there is a systematic bias in the results, such as one device consistently reading higher than the other, the mean of the points will be clustered either above or below zero on the y-axis and this will be evident both visually and by the mean difference value produced. If this deviation is consistent and can be relied on, readings between devices can be corrected by subtracting or adding the mean difference. However, if there is also variance in the difference between devices, points will be more scattered and there will be a wider ‘limit of agreement’, which is calculated as ± 2 standard deviations (SDs). If this limit of agreement is wide by clinical standards, it may be concluded that the devices are not clinically interchangeable, even if the mean difference is relatively small.

Diagnostic review

In the absence of an end-to-end study, the next best study design is a cohort study. The ideal cohort study would have recruited patients presenting to their GP with symptoms of asthma and would have assessed the standard UK diagnostic pathway8 as well as this pathway with the addition of FeNO against a reference standard of long-term follow-up. No studies of this design were found. Instead, studies that compared FeNO with or without another test against a reference standard of any test or combination of tests in the UK guidelines (see Figures 3 and 4) were included. UK guideline tests include:

• Spirometry and lung function tests (mostly FEV₁%, FEV₁/FVC, PEFR).

• Airway reversibility: airway obstruction that shows reversibility when a bronchodilator is taken.

• ICS responsiveness: response to a trial of treatment with ICSs.

• Airway hyper-responsiveness to methacholine, histamine, exercise or mannitol.

• Tests for airway inflammation (FeNO or sputum eosinophil counts), although these are currently restricted to a few specialist centres. Studies that use sputum eosinophils within the reference standard are not considered to be similar to UK practice as this test is not widely available.

Twenty-three cohort studies [across 26 publications54,69–79,82–91,93–96 and one sponsor’s submission (Fukuhara 2013)] that reported the sensitivity and specificity of FeNO testing (alone or in combination with another test or tests) compared with the sensitivity and specificity of some or all of the tests within the UK diagnostic pathway were identified and included in the review. A further three studies80,81,92 were identified from the update search. Four of these studies70,77,87,95 included data for FeNO testing in conjunction with another test.

Studies were not similar enough to warrant a meta-analysis, with substantial heterogeneity in populations, cut-off values, devices used and reference standards. We decided to instead focus on key studies that most closely resembled UK practice and resifted the included studies to separate out these studies. We did not, however, want to exclude completely the other studies in case they might prove useful to the committee in their decision-making, especially as some were studies that the SCMs had indicated might be of use when consulted during the clarification of the inclusion and exclusion criteria (see Appendix 2).

This review is subdivided into a number of sections and to aid reading a summary is given here:

• Studies including adults, adults plus adolescents, all age groups and unspecified age groups.

  • Studies meeting the inclusion criteria – this section tabulates all included studies and assesses their relevance to the decision problem.

  • Studies relevant to the decision problem using only FeNO as the index test – this section provides an appraisal of study quality, a narrative synthesis and greater detail relating to these studies, along with estimates of sensitivity and specificity.

  • Studies using FeNO in conjunction with another test as the index test – this section provides a narrative synthesis and greater detail relating to these studies, along with estimates of sensitivity and specificity.

• Studies including children or children and adolescents.

  • Studies using only FeNO as the index test – because of the smaller number of studies relating to children, all studies are included without selection based on their relevance to the decision problem. This section provides a narrative synthesis and greater detail relating to these studies, along with estimates of sensitivity and specificity.

  • Studies using FeNO in conjunction with another test as the index test – this section provides greater detail relating to one study, along with estimates of sensitivity and specificity.

• Studies providing data on subgroups of interest to the review:

  • adult smokers

  • children exposed to tobacco smoke

  • pregnant women

  • the elderly.

More detailed descriptions of the study characteristics for studies that were judged to be less relevant, along with sensitivity and specificity data, are included in Appendices 7 and 8 for reference.

Adults, adults plus adolescents, all age groups and unspecified age groups

Of the 26 studies included in the review, 22 (26 citations) were conducted in adults (13 studies, 16 citations54,69–81,91), adults plus adolescents (three studies, four citations82–85), all age groups (three studies, three citations86–88) or an unspecified age range (three studies, three citations89,90,92). Table 21 summarises the characteristics of the studies and provides a brief description explaining their relevance to the decision problem (note: Schneider et al. 71,72 appears twice in this table as this study reported two differently defined populations). This table should be read alongside Figure 8, which ascribes a letter to positions that already exist in the current diagnostic pathway and a number to the positions where FeNO could be added to the pathway, as agreed during the scoping workshop.

TABLE 21 - Diagnostic review: key study characteristics of the diagnostic cohort studies

IgE, immunoglobulin E; NA, not applicable; NR, not reported; WBP, whole-body plethysmography.

Army recruits, some of whom are thought to have lied about their existing asthma diagnosis.

Author, year	Population	Age group	Device	Cut-off values (ppb)	Reference standard	Position of FeNO in the pathway	Relevance to decision problem
Position A
Fortuna 200770	Position A	Adults	N-6008	20	FEV1, FEV1/FVC, airway responsiveness, WBP, airway hyper-responsiveness (MCT)	No equivalent position in UK pathway	Not relevant – uses WBP and device of unknown equivalence
Fukuhara 2011,54 Fukuhara 2013 (sponsor’s submission)	Position A	Adults	NA623	40	At least two of the three criteria of induced sputum eosinophilia, airway hyper-responsiveness and reversibility. Exclusion of other lung diseases	No equivalent position in UK pathway (sputum eosinophilia)	Not relevant – reference standard not similar enough to common UK practice (sputum eosinophilia)
Schneider 200971,72	Position A	Adults	NIOX MINO	12, 16, 20, 35, 46, 76	FEV1, FEV1/FVC, airway reversibility, airway hyper-responsiveness (MCT) in sequence similar to that in UK guidelines	FeNO replaces whole pathway (excluding trial of treatment)	Relevant
Schneider 201369	Position A	Adults	NIOX MINO	9, 12,16, 20, 25, 35, 41, 42, 43, 44, 45, 46, 71	FEV1, FEV1/FVC, airway reversibility, hyper-responsiveness (MCT)	FeNO replaces whole pathway (excluding trial of treatment)	Relevant
Smith 200486	Position A	All ages	NR	20	ATS 1987130 symptoms plus one of airway reversibility, positive response to ICSs, airway hyper-responsiveness (MCT)	FeNO replaces whole pathway	Relevant although uses unknown device of unknown equivalence
Smith 200583	Position A	Adults and adolescents	Niox	≥ 15, > 47, < 15	ATS 1987130 symptoms plus one of airway reversibility, positive response to ICSs, airway hyper-responsiveness (MCT)	FeNO replaces whole pathway	Relevant
de La Barra 201184				25, 40, 50, 70, 90, 110, 130, 150	Airway reversibility	FeNO replaces airway reversibility	Relevant
Subset of position A
Cordeiro 201187	Position A with high prevalence of atopy	All ages	NIOX Flex (Aerocrine, Solna, Sweden)	27	Airway reversibility, airway hyper-responsiveness (histamine)	FeNO replaces whole pathway	Relevant although population not whole spectrum
			NIOX Flex, airway reversibility	FeNO 27 and/or > 12% and 200-ml improvement in FEV1 with bronchodilator	Airway reversibility, airway hyper-responsiveness (histamine)	Combination replacing FEV1/FVC and airway hyper-responsiveness	Relevant although population not whole spectrum
Heffler 200682	Position A with rhinitis	Adults and adolescents	Niox	10, 15, 20, 25, 30, 34, 36, 40, 454, 50, 55, 60, 65, 75, 80, 85, 100	Airway hyper-responsiveness (MCT) or airway reversibility	FeNO replaces pathway (not including ICS responsiveness but including airway reversibility)	Relevant – rhinitis population generalisable to whole asthma population
Pizzimenti 200990	Position A with chronic cough	Unclear age group	NIOX MINO	55	Airway hyper-responsiveness (MCT)	No equivalent position in UK practice	Not relevant – not equivalent to UK practice
Difficult to diagnose
Bobolea 201288	Position H	All ages	NIOX MINO	30	Adenosine challenge test	FeNO at position 2	Relevant
Katsoulis 201381	Before position G (via negative bronchodilator response)	Adults	NIOX MINO	32	Hyper-responsiveness (MCT)	FeNO precedes MCT	Relevant
Mathew 201191	Patients at E or F	Unclear age group	NR	NR	Airway hyper-responsiveness (MCT)	Equivalent to FeNO at position 1 or 2	Not relevant – device and cut-off values not reported and uses unknown device of unknown equivalence
Pedrosa 201085	Before position G	Adults and adolescents	NIOX MINO	40	Airway hyper-responsiveness (MCT)	FeNO at position 1	Relevant
Schleich 201277	Before position G	Adults	Niox	34	Airway hyper-responsiveness (MCT)	FeNO at positions 1 and 2 (in place of MCT)	Relevant
			Niox, FEV1 101%	FeNO > 34 and FEV1 ≤ 101%	Airway hyper-responsiveness (MCT)	Combination replaces FEV1/FVC and airway hyper-responsiveness (MCT)	Relevant
Schneider 200971,72	Position Bii	Adults	NIOX MINO	12, 46	Airway hyper-responsiveness (MCT)	NA	Not relevant – in the UK these patients are more likely to receive trial of treatment than MCT
Difficult to diagnose with chronic cough
Hahn 200774	Position F with chronic cough	Adults	NOA 280i	35, 38	ICS responsiveness	FeNO replaces trial of treatment (whole treatment pathway)	Relevant although diagnosis is of ICS responsiveness not asthma and uses device of unknown equivalence
Hsu 201373	Position F with chronic cough	Adults	NOA 280i	30, 33.9	ICS responsiveness	FeNO replaces trial of treatment (whole treatment pathway)	Relevant although diagnosis is of ICS responsiveness not asthma and uses device of unknown equivalence
Prieto 200976	Position F with chronic cough	Adults	Niox	20	ICS responsiveness	FeNO replaces trial of treatment	Relevant although diagnosis is of ICS responsiveness not asthma
Sato 2008,75 Mathew 201191	Position F with chronic cough	Adults	Chemiluminescence analyser (Kimoto; Osaka, Japan)	38.8	Cough with/without wheeze, sputum eosinophilia, airway reversibility, airway hyper-responsiveness (MCT)	No equivalent position in UK pathway (sputum eosinophilia)	Not relevant – reference standard not similar to common UK practice (sputum eosinophilia) and uses device of unknown equivalence
Zhang 201189	Position F with chronic cough	Unclear age group	NIOX MINO	31, 40	Sputum eosinophilia, pulmonary function test, airway hyper-responsiveness, 24-hour oesophageal pH monitoring, skin prick test and serum IgE	No equivalent position in UK practice	Not relevant – reference standard uses tests not used in UK standard practice
EIB
El Halawani 200378	Suspected EIB	Adults	NOA/Sievers 280A	12	Exercise challenge	NA	Relevant but unclear if patients similar to those who would be referred to exercise challenge test in the UK and uses device of unknown equivalence
Other
Arora 200679	Mix of undiagnosed and diagnoseda	Adults	Niox	6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 25, 30, 40, 46	Airway hyper-responsiveness (histamine)	Equivalent to MCT or later?	Not relevant – too different from UK population; reference standard would not be applied to all UK patients
Brannan 201392	Unclear, referred for mannitol challenge in Australia	Unclear age group	HypAir (Medisoft, Leeds, UK) with 0.6-ppb correction to match Niox	47	Hyper-responsiveness – mannitol challenge	Unclear if comparable as population unknown	Unclear
Chancafe-Morgan 201380	Unclear, referred for bronchial hyper-responsiveness testing in Spain	Adults	NR	35	Hyper-responsiveness (MCT)	Unclear if comparable as population unknown	Unclear

Studies meeting the inclusion criteria

Of the 22 studies conducted in adults that met the inclusion criteria for the review:

• Nine studies54,69–72,82–84,86,87,90 recruited patients with symptoms of asthma who were broadly equivalent to patients entering the UK pathway at position A (see Figure 8). Of these, six69,71,72,82–84,86,87 were considered to be of most relevance to the decision problem, although the studies had not necessarily been conducted in the UK.

• Nine studies73–77,81,85,88,89,91 recruited patients who could be considered ‘difficult to diagnose’ and are located at other points along the diagnostic pathway. These patients had already undergone some of the tests in the UK pathway and had tested negative for asthma thus far. One further study71,72 reported a subset of difficult-to-diagnose patients from a larger cohort of patients at position A (see Figure 8). Of these, seven studies73,74,76,77,81,85,88 were considered to be of most relevance to the decision problem, although the studies had not necessarily been conducted in the UK.

• One study78 recruited patients with suspected EIB. This study was considered to be relevant to the decision problem, although it had not been conducted in the UK.

• One study79 recruited army recruits, amongst whom a high proportion are thought to have lied about their asthmatic status. This study was not considered to be relevant to the decision problem and it had not been conducted in the UK.

• Two studies80,92 did not describe the populations they included and their relevance to the decision problem is therefore unknown.

Reasons for considering a study not relevant to the decision problem are given in Table 21. A total of 14 studies69,71–74,76–78,81–88 were considered relevant and a further two80,92 were considered to be of unknown relevance and are considered in greater detail in the following section. Full study details and results for the six studies54,70,75,79,89–91 considered not relevant are provided in Appendices 7 and 8.

Studies relevant to the decision problem using only FeNO as the index test

From the initial 22 studies conducted in adults, adults plus adolescents, all age groups and unspecified age groups, 14 studies were considered to be of most or of some relevance to the decision problem. These studies are Schneider et al. ,69 Schneider et al. ,71,72 Schleich et al. ,77 Prieto et al. ,76 El Halawani et al. ,78 Hsu et al. ,73 Hahn et al. ,74 Pedrosa et al. ,85 Heffler et al. ,82 Smith et al. 83 (de la Barra et al. 84 reports additional analyses to Smith et al. 83) Smith et al. ,86 Bobolea et al. ,88 Cordeiro et al. 87 and Katsoulis et al. 81 Two further studies80,92 had unclear relevance to the decision problem and are also considered.

Table 22 groups the 14 studies considered to be of relevance to the decision problem according to the position on the pathway and the reference standard and tabulates the study and patient characteristics. Appendix 9 provides more detail about the specifics of the reference standards used and Appendix 10 provides more detail about the patient inclusion and exclusion criteria. There are several main sources of heterogeneity amongst these studies that preclude meta-analysis of the results. These include: the age groups recruited; the spectrum of patients in terms of their position in the pathway and other restrictions in recruitment such as having rhinitis or chronic cough; the device used to measure FeNO; the reference standards used; and the cut-offs reported.

TABLE 22 - Diagnostic review: study and patient characteristics of the 14 studies considered of relevance to the decision problem

CI, confidence interval; GM, geometric mean; GORD, gastro-oesophageal reflux disease; IQR, interquartile range; LTFU, lost to follow-up; NR, not reported; OAD, obstructive airway disease; UACS, upper airway cough syndrome.

Author, year	Study design, funding	Country, setting, recruitment dates	Population	Device	Reference standard	n analysed/N recruited (%), reasons for withdrawals	Mean (SD) age (years)	Sex male, n/N (%)	Severity, mean (SD) FEV1%	Mean (SD) FeNO (ppb)	Smokers, n/N (%)	Atopic, n/N (%)
Position A vs. whole pathway
Schneider 200971,72	Prospective, consecutive cohort study Funding: GOVERNMENT	Germany Primary care – 14 GPs across 10 practices February 2006–June 2007	Adults Position A	NIOX MINO	FEV1, FEV1/FVC, airway reversibility, airway hyper-responsiveness (MCT)	160/160 (100)	43.9	72/160 (45)	Asthmatics (n = 75) 100 (12.2); COPD (n = 25) 67.8 (18.5); overlap (n = 8) 68.8 (18.4); no OAD (n = 52)107.4 (12.8)	Asthmatics (n = 75) 42.6 (47.9); COPD (n = 25) 16.2 (11.1); overlap (n = 8) 20.4 (18.6); no OAD (n = 52) 24.7 (16.0)	Current and ex-smokers 86/160 (54)	NR
Schneider 201369	Prospective, consecutive cohort study Funding NR: authors report no conflicts	Germany Private practice run by five pneumologists June 2010–October 2011	Adults, position A	NIOX MINO	FEV1, FEV1/FVC, airway reversibility, hyper-responsiveness (MCT)	393/400 (98) Lack of data: n = 7	Asthma: 40.5 (15.4); COPD: 60.8 (17.0); no OAD 44.6 (16.5)	158/393 (40.2)	Asthmatics 101.3 (17.0); COPD 74.1 (12.3); no OAD 107.7 (16.3)	Asthmatics 42.4 (46.4); COPD 16.6 (6.8); no OAD 22.0 (16.5)	Current smoker 39/393 (9.9); ex-smoker 139/393 (35.4)	NR
Smith 200486	Prospective, consecutive cohort study Funding: mix of industry and non-industry but not device manufacturer	New Zealand Secondary care, one centre Dates NR	All patients	NR	Airway reversibility, positive response to ICSs, airway hyper-responsiveness (MCT)	44/51 (86) Withdrew: n = 7; withdrew consent (time): n = 4; technical difficulties n = 3	Asthmatics (n = 17) 41.6 (range 9–72); non-asthmatics (n = 30) 31.8 (range 9–64)	20 (42.6)	Asthmatics (n = 17) 90.5 (18.4); non-asthmatics (n = 30) 110.0 (13.5)	Asthmatics (n = 17) 52 (34.0); non-asthmatics (n = 30) 15.7 (12.9)	Ex-smokers 5/47 (10.6)	NR
Smith 200583	Prospective, consecutive cohort study Funding: mix of industry and non-industry but not device manufacturer	New Zealand Secondary care, one centre Dates NR	Adults and adolescents Position A	Niox	Airway reversibility, positive response to ICSs, airway hyper-responsiveness (MCT)	52/60 (87) Withdrew: n = 8; withdrew consent: n = 3; respiratory tract infection n = 1; acute rhinitis n = 1, LTFU n = 3	40.5 (range 14–71)	20/52 (38.5)	97.8 (14.2)	Range 6.3–242.084	Current smoker 3/52 (5.8); ex-smoker 10/52 (19.2)	40/52 (77)84
Position A vs. airway reversibility
de La Barra 201184	Prospective, consecutive cohort study Funding: Aerocrine and lottery grants New Zealand	New Zealand Secondary care, one centre Dates NR	Adults and adolescents Position A	Niox	Airway reversibility	52/60 (87) Withdrew: n = 8; withdrew consent: n = 3; respiratory tract infection: n = 1; acute rhinitis: n = 1; LTFU: n = 3	40.5 (range 14–71)83	20 (38.5)83	97.8 (14.2)83	Range 6.3–242.0	Current smokers 3/52 (5.8); ex-smokers 10/52 (19.2)	40/52 (77)
Subset of Position A vs. airway reversibility or airway hyper-responsiveness
Cordeiro 201187	Retrospective (analysis of prospective database) Funding NR, authors reported no conflicts	Netherlands Secondary care January 2007–September 2007	All ages with high prevalence of atopy Position A	NIOX Flex	Airway reversibility, airway hyper-responsiveness (histamine)	114/114 (100)	Median (range): asthmatics (n = 42) 39 (7–83); non-asthmatics (n = 72): 38 (7–87)	43/114 (37.7)	FEV1/FVC%, median (range): asthmatics 70 (42–95); non-asthmatics 77 (69–95)	Median (range): asthmatics (n = 42) 44 (6–290); non-asthmatics (n = 72) 17 (5–45)	11/114 (9.6)	81/114 (71.1)
Heffler 200682	Prospective, consecutive cohort study Funding: government/non-industry	Italy Allergy and immunity clinic Dates NR	Adults and adolescents with rhinitis Position A	Niox	Airway hyper-responsiveness (MCT) or airway reversibility	48/48 (100)	40.08 (SD NR)	21 (43.75)	89.2 (95% CI 80.1 to 98.4)	59.7 (95% CI 50.2 to 89)	0	35/38 (92.1)
Difficult to diagnose vs. airway hyper-responsiveness
Bobolea 201288	Prospective, consecutive cohort study Funding: NR	Spain Assuming secondary care Dates NR	All ages Position H	NIOX MINO	Adenosine challenge test	30/30 (100)	37.3 (range 13–69)	13/30 (43.3)	NR	NR	NR	NR
Katsoulis 201381	Assume prospective cohort Funding NR	Greece Army general hospital and local general hospital Dates NR	Adults Position G	NIOX MINO	Airway hyper-responsiveness (MCT)	112	Median (IQR) 25 (22 to 37)	95/112 (84.8)	Median (IQR) 89 (83 to 99)	Median (IQR) 20.5 (12 to 34)	NR Recent ex-smokers excluded	51/112 (45.5)
Pedrosa 201085	Prospective, consecutive cohort study Funding: NR, authors report no conflicts	Spain Secondary care Dates NR	Adults and adolescents Position G	NIOX MINO	Airway hyper-responsiveness (MCT)	114/115 (99) Withdrawal n = 1, reason NR	34 (13)	72/115 (62.6)	(n = 115) 104.29 (14.95)	(n = 115) 34	Current smokers 17/115 (14.8); ex-smokers 11/115 (9.6)	100/115 (87)
Schleich 201277	Prospective cohort study Funding: non-industry	Belgium Secondary care March 2009–December 2009	Adults with chronic cough Position G	Niox	Airway hyper-responsiveness (MCT)	174/237 (73) n = 63 did not meet inclusion criteria	41 (16)	72/174 (41)	97 (13)	Median (range) 17 (4–271)	59/174 (33.9)	84/174 (48)
Suspected EIB vs. exercise challenge test
El Halawani 200378	Prospective, consecutive cohort study Funding: NR	USA Naval medical centre Dates NR	Adults Suspected EIB	NOA/Sievers 280A	Exercise challenge	49/50 (98) Inability to complete spirometry n = 1	27.9 (SD NR)	35 (71.4)	NR	EIB group (n = 7) 41; non-EIB group (n = 42) 25.6	0	0
Position H with chronic cough vs. ICS responsiveness
Hahn 200774	Retrospective cohort study Funding: NR, authors reported no conflicts	Mayo Clinic, Rochester, MN, USA Secondary care December 2004–November 2005	Adults with chronic cough Position H	NOA 280i	ICS responsiveness	64/64 (100)	Pooled weighted mean 46.8 (NR)	26/64 (40.6)	ICS unresponsive group 98; ICS responsive group 94	ICS unresponsive group 26.0 ± 16.5; ICS responsive group 51.25 ± 20.1	Current smokers 0/64 (0); ex-smokers 10/64 (15.6)	NR
Hsu 201373	Retrospective cohort study Funding: NR	Taiwan Asthma and cough-specific clinic June 2007–May 2008	Adults with chronic cough Position H	NOA 280i	ICS responsiveness	81/114 (71) n = 33 (26 lost after first visit, 7 stopped coughing after 1–2 weeks of treatment for UACS and GORD)	49 (14)	33/81 (40.7)	91.8 (15.3)	Mean rank FeNO by Kruskal–Wallis test: 47	0/81 (0)	NR
Prieto 200976	Prospective cohort study, unclear if consecutive Funding: none	Spain Allergy or respiratory clinics Dates NR	Adults with chronic cough Position H	Niox	ICS responsiveness	43/43 (100)	48 (95% CI 43 to 52)	18/43 (41.9)	113.2 (95% CI 108.0 to 118.3)	GM (95% CI): responders 23.2 (17.5 to 30.7); non-responders 18.6 (14.7 to 24.0)	0/43 (0)	43/43 (100)
Unclear population vs. airway hyper-responsiveness
Brannan 201392	Retrospective cohort study Funding NR	Australia Pulmonary function laboratory (secondary care) Dates NR	Unclear, referred for mannitol challenge in Australia	HypAir with 0.6-ppb correction to match Niox	Mannitol challenge	401	NR	NR	NR	NR	NR	NR
Chancafe-Morgan 201380	Prospective, consecutive cohort study Funding NR	Spain Pulmonary function laboratory (secondary care) Dates NR	Unclear, referred for bronchial hyper-responsiveness testing in Spain	NR	Hyper-responsiveness (MCT)	30	44.2 (16.7)	10/30 (33)	NR	33.6 (18.7)	NR	NR

For each study we have selected and presented in tables three sets of sensitivity and specificity estimates. These are:

• The highest sum of sensitivity and specificity as reported by the authors of the study.

• The highest sensitivity – in this scenario a negative test result rules out a diagnosis of asthma (see Tables 3–5 for details). This was selected as the cut-off that provided the highest sensitivity. When 100% sensitivity was reported for more than one cut-off, the cut-off that maintained the highest specificity was selected. It should be noted that some studies did not report 100% sensitivity, although this may have been achievable at lower cut-off points. When the cut-off with the highest sensitivity was not also the cut-off with the highest positive predictive value (PPV), this latter cut-off was also presented.

• The highest specificity – in this scenario a positive test result rules in a diagnosis of asthma (see Tables 3–5 for details). Selected as for the highest sensitivity but for specificity. When the cut-off with the highest specificity was not also the cut-off with the highest negative predictive value (NPV), this latter cut-off was also presented.

It should be noted that superior sets of sensitivity and specificity values may in fact have been achieved but selection was limited to the range of cut-off points reported within studies.

Quality assessment of studies relevant to the decision problem

Sixteen studies (18 citations69,71–74,76–78,80–88,92) exploring FeNO measurement for the diagnosis of asthma in adults were assessed for quality according to QUADAS-2 criteria for diagnostic accuracy studies. 38 Although based on the same data as the study by Smith et al. ,83 the study by de la Barra et al. 84 was assessed separately as the analysis was different.

The overall study quality was variable, with the study by Smith et al. 83 scoring well on all of the domains and thus being at least risk of bias. The studies at highest risk appeared to be those by Hsu et al. 73 and Cordeiro et al. ,87 neither of which provided sufficient information on the nature of blinding for the index and reference standard tests. There were also some issues in terms of patient flow in both studies. In the study by Cordeiro et al. ,87 patients did not all receive the same reference test (MCT was provided only if asthma was suspected from other tests). Similarly, in the study by Hsu et al. ,73 the reference standard was allocated based on an algorithm rather than on an a priori set of tests (Figure 11). The studies by Katsoulis et al. 81 and Brannan et al. 92 were poorly reported and at unknown risk of bias. The risk of bias from the conduct of the index test scored worst overall, with only one study scoring positively for this domain. Studies scored poorly for risk of bias from the conduct of the reference standard, with 12 having a score of ‘unclear’ for this domain.

FIGURE 11.

Risk of bias summary: review authors’ judgements about each risk of bias item for each included study. Dark green circles with + sign, low risk of bias; light green circles with – sign, high risk of bias; medium green circles with ?, unclear risk of bias.

Risk of bias from patient selection

As far as we could ascertain, patient selection did not appear to be a source of bias in the body of literature. All studies avoided a case–control design and recruited appropriately (i.e. those patients presenting with clinical signs of asthma or a subset thereof or patients at a definable point in the UK pathway). However, it was unclear in seven cases73,74,76,77,81,87,92 whether a consecutive sample was recruited.

Risk of bias from the conduct of the index test

The conduct of the index test was a potentially important source of bias, with only the study by Smith et al. 83 being free from bias in this domain as a whole. There were two component questions for this domain, one relating to blinding and one to whether the study was a derivation study or a validation study. Ten studies were unclear whether the index test was interpreted blind to the reference standard. 71,73,74,77,80–82,85,88,92 The studies by de la Barra et al. 84 and Cordeiro et al. 87 were not explicit with regard to the blinding of the reference standard results; however, as the index test was performed before the reference standard, it would not have been possible for the investigator to be aware of the reference results at the time of the index test unless interpretation was not performed at the time of the test. This would seem unlikely as FeNO measurement carried out according to standardised protocols is objective and interpretation is not required.

Several further studies were at potential risk of bias in that they were derivation studies that fitted cut-off points to the data post hoc and were thus likely to overestimate accuracy. 69,74,76–78,82,84,85,87

Risk of bias from the conduct of the reference standard

The reference standard and its interpretation was a further source of bias among much of the literature, with only the study by Prieto et al. ,76 the two studies by Schneider et al. 69,71 and the study by Smith et al. 83 being free of bias. It was not possible to ascertain in any of the remaining literature whether the operator conducting the reference standard had been blinded to the results of the index test.

Risk of bias from patient flow and timing of the study

For the most part there was little concern about the patient flow and study timing. However, at least two studies did not provide an identical reference standard for all patients. In the study by Cordeiro et al. 87 patients received MCT only if asthma was suspected based on other tests, whereas in the study by Hsu et al. 73 reference standard provision was algorithm based, with some patients not receiving ICS treatment. It was not necessarily clear in all other studies whether patients received all reference standard tests or a sequence. In other respects, the patient flow and study timing were satisfactory. Dropout rates were low and, when dropouts occurred, these were adequately accounted for in the study reports.

Summary

The corpus of included literature was of variable quality, with the study by Smith et al. 83 being at the least risk of bias and the two Schneider studies69,71 also performing well. The conduct of the index test was identified as a potentially serious source of bias among the literature, with few studies providing adequate information on how blinding to the reference test results was achieved. Most studies were derivation studies and, in fitting cut-off points to the data post hoc, are likely to overestimate the accuracy of FeNO as a diagnostic test. In terms of the conduct of the reference standard, few studies provided satisfactory information on how operators were blinded to the results of the index test. However, it is important to stress that this may reflect lack of clarity in the study reports rather than in the conduct of the reference test itself. The likelihood of unblinding biasing the results is therefore unclear.

Studies recruiting patients at Position A

Position A is the start of the UK pathway. Patients will have undergone no other tests. The reference standard used in studies that recruit patients at this position will determine whether the results relate to a scenario in which FeNO is replacing the whole pathway or a scenario in which it is replacing just one test within the pathway. When replacing just one test it could be used as a rule-in scenario; patients testing positive would go on to be treated as asthmatic and patients testing negative would go on to have further tests for asthma. When a rule-out scenario is used, patients testing positive would go on to have further tests and patients testing negative would go on to be treated as not asthmatic.

Position A compared with the whole pathway

Population: Four studies69,71,72,83,86 recruited patients with symptoms of asthma who had not undergone any other tests. These studies are unlikely to have recruited the full spectrum of patients at this point in the pathway because of common exclusions such as those who had experienced a respiratory infection in the last month and those taking ICSs (see Appendix 10). As in many cases a GP may provide a patient with ICSs before confirmation of asthma; these exclusions may result in a patient spectrum that does not reflect UK practice.

Of the four studies, those by Schneider et al. 69,71,72 recruited adults, that by Smith et al. 83 recruited adults and adolescents and that by Smith et al. 86 recruited patients of any age. The largest study was that by Schneider et al. 69 with 393 participants and the smallest was that by Smith et al. 86 with 44 participants. Mean age and FEV₁% and FeNO values were not always reported for the whole cohort, making it difficult to compare across studies. All studies recruited more females than males, with the proportion of males ranging from 38.5% to 45%. Schneider et al. 69,71,72 and Smith et al. 83 recruited a mix of smokers, ex-smokers and non-smokers whereas Smith et al. 86 did not recruit any smokers, although this was not listed as an exclusion criteria and may be a result of the small sample size. Only Smith et al. 83 reported how many participants were atopic, with a high prevalence of 77%; the other three studies did not list atopy as an exclusion criterion, making it likely they included a proportion of atopic patients. The study by Schneider et al. 69 excluded pregnant women.

Intervention: The two Schneider studies69,71,72 both used NIOX MINO. The study by Smith et al. 83 used the Niox chemiluminescent device and that by Smith et al. 86 did not report the device used.

Reference standard: The reference standard for both studies by Schneider et al. 69,71,72 was airway reversibility or airway hyper-responsiveness (depending on spirometric test results), whereas in both studies by Smith et al. 83,86 the reference standard also incorporated ICS responsiveness. Although these reference standards do differ, bronchodilator reversibility and ICS responsiveness appear to be used interchangeably in the UK pathway and so both reference standards are equivalent to the whole pathway. However, it is likely that these reference standards may differentially influence estimates of FeNO diagnostic accuracy as FeNO would be expected to correlate better with ICS responsiveness than airway reversibility testing with a bronchodilator.

Study design and setting: All studies were prospective, consecutive cohort studies and none of the studies was funded by the manufacturers of a FeNO device. Both studies by Schneider et al. 69,71,72 were conducted in Germany in primary care or private practice whereas both studies by Smith et al. 83,86 were conducted in New Zealand in secondary care.

Estimates of diagnostic accuracy: Table 23 details the estimates of sensitivity and specificity for these studies. The results do not appear to be similar between studies. The cut-off for the highest sum of sensitivity and specificity varied from 20 ppb to 47 ppb and this did not appear to be dependent on any variable. The studies by Schneider et al. 69,71,72 and Smith et al. 83 all reported higher specificity values than sensitivity values, whereas the study by Smith et al. 86 reported the opposite. This study recruited a mixed population of adults and children and also did not report the device used to measure FeNO. Sensitivities varied greatly across studies, ranging from 32% to 88%. Specificities were more consistent across studies, ranging from 75% to 93%.

TABLE 23 - Diagnostic review: diagnostic accuracy of FeNO tests in adults, adults and adolescents and all agesa

NR, not reported.

A table of all results reported by these studies is given in Appendix 11.

Unclear if this is the best sum of sensitivity and specificity.

Calculated by reviewer.

Only reported cut-offs for 10–30 ppb.

Author, year	Population	Device	n	Reference standard	Highest sum of sensitivity and specificity					Rule out					Rule in
					Cut-off (ppb)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	Cut-off (ppb)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	Cut-off (ppb)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
Position A vs. whole pathway
Schneider 200971,72	Adults Position A	NIOX MINO	160	FEV1, FEV1/FVC, airway reversibility, airway hyper-responsiveness (MCT)	46	32	93	77.3	59.5	12	85	24	49.6	64.5	76	13	100	100	56.7
										16	69	53	56.5	66.2
Schneider 201369	Adults Position A	NIOX MINO	393	FEV1, FEV1/FVC, airway reversibility, hyper-responsiveness (MCT)	25	49	75	56.0	69.5	9	96	13	41.6	83.8	71	18	97	79.5	64.6
Smith 200486	All patients	NR	44	Airway reversibility, positive response to ICSs, airway hyper-responsiveness (MCT)	20	88	79	70	91.7	NR	NR	NR	NR	NR	NR	NR	NR	NR	NR
Smith 200583	Adults and adolescents Position A	Niox	52	Airway reversibility, positive response to ICSs, airway hyper-responsiveness (MCT)	47	55.6	92	88.2	65.7	15	81.5	48	29.4	37.1	As highest sum
Position A vs. airway reversibility
de la Barra 201184	Adults and adolescents Position A	Niox	52	Airway reversibility	41.7	NR	NR			25	83.3	57.5	37.0	92	110	25	95	60	80.9
															90	41.7	92.5	62.5	84.1
Subset of Position A vs. airway reversibility or airway hyper-responsiveness
Cordeiro 201187	All ages with high prevalence of atopy Position A	NIOX Flex	114	Airway reversibility, airway hyper-responsiveness (histamine)	27	78	92	84.6	88	NR	NR	NR	NR	NR	NR	NR	NR	NR	NR
		NIOX Flex, airway reversibility		Airway reversibility, airway hyper-responsiveness (histamine)	27	87	90			NR	NR	NR	NR	NR	NR	NR	NR	NR	NR
Heffler 200682	Adults and adolescents with rhinitis Position A	Niox	48	Airway hyper-responsiveness (MCT) or airway reversibility	36	77.8	60	53.8	81.8	25	100	46.7	52.9	100	100	27.8	100	100	69.8
Difficult to diagnose vs. airway hyper-responsiveness
Bobolea 201288	All ages Position H	NIOX MINO	30	Adenosine challenge test	30b	100	29.2	26	100	NR	NR	NR	NR	NR	NR	NR	NR	NR	NR
Katsoulis 201381	Adults Position G	NIOX MINO	112	Hyper-responsiveness (MCT)	32	47	85	70.1c	68.1c	10d	81	39	49.9c	73.2c	30d	49	82	67.1c	68.2c
	Atopics		51		26	55	85	NR	NR	10d	90	10%	NR	NR	30d	48	85	NR	NR
Pedrosa 201085	Adults and adolescents Position G	NIOX MINO	114	Airway hyper-responsiveness (MCT)	40	74.3	72.5	54.1	86.3	NR	NR	NR	NR	NR	NR	NR	NR	NR	NR
Schleich 201277	Adults with chronic cough Position G	Niox	174	Airway hyper-responsiveness (MCT)	34	35	95	87.8	62.4	NR	NR	NR	NR	NR	NR	NR	NR	NR	NR
		Niox, FEV1 ≤ 101%		Airway hyper-responsiveness (MCT)	34	24.4	98.9			NR	NR	NR	NR	NR	NR	NR	NR	NR	NR
Suspected EIB vs. exercise challenge test
El Halawani 200378	Adults Suspected EIB	NOA/Sievers 280A	49	Exercise challenge	12b	100	31	19.4	100	NR	NR	NR	NR	NR	NR	NR	NR	NR	NR
Position H with chronic cough vs. ICS responsiveness
Hahn 200774	Adults with chronic cough Position H	NOA 280i	64	ICS responsiveness	38	90	85	89.5	84.6	NR	NR	NR	NR	NR	NR	NR	NR	NR	NR
Hsu 201373	Adults with chronic cough Position H	NOA 280i	81	ICS responsiveness	33.9	94.7	76.3	80	94	NR	NR	NR	NR	NR	NR	NR	NR	NR	NR
Prieto 200976	Adults with chronic cough Position H	Niox	43	ICS responsiveness	20	53	63	52.6	62.5	NR	NR	NR	NR	NR	NR	NR	NR	NR	NR
Unclear position in the pathway vs. airway hyper-responsiveness
Brannan 201392	Unclear	HypAir	401	Mannitol challenge	47	30.2c	96.3c	65.7c	85.5c	NR	NR	NR	NR	NR	NR	NR	NR	NR	NR
Chancafe-Morgan 201380	Unclear	NR	30	Airway hyper-responsiveness (MCT)	35	75	83.3	75	83.3	NR	NR	NR	NR	NR	NR	NR	NR	NR	NR

Rule-out cut-off points varied from 9 ppb to 16 ppb with sensitivities between 69% and 96%, specificities between 13% and 53%, PPVs between 29.4% and 56.5% and NPVs between 37.1% and 83.8%. Rule-in cut-off points varied from 47 ppb to 76 ppb, with specificities between 92% and 100%, sensitivities between 13% and 55.6%, PPV between 79.5% and 100% and NPV between 56.7% and 65.7%. The two studies by Schneider et al. 69,71,72 reported very similar rule-in (71 ppb and 76 ppb respectively) and rule-out (9 ppb and 12 ppb respectively) cut-off points, but the cut-off providing the highest sum of sensitivity and specificity was not similar between these two studies (25 ppb and 46 ppb respectively). Smith et al. 83 reported a similar rule-out cut-off point (15 ppb) to these studies but a quite different rule-out cut-off point (47 ppb). Only Schneider et al. 71,72 reported a 100% PPV that would reliably rule patients in, with no studies reporting a 100% NPV.

Position A compared with airway reversibility

De la Barra et al. 84 performed a secondary analysis of the data from Smith et al. 83 against a reference standard of airway reversibility only. This is equivalent to replacing airway reversibility with FeNO, or placing FeNO before airway reversibility as a rule-in test or rule-out test, with patients going on to receive this and further tests as appropriate. The cut-off point with the best sum of sensitivity and specificity (41.7 ppb) seemed fairly similar to that reported in Smith et al. 83 (47 ppb). The rule-out cut-off point was somewhat higher at 25ppb, compared with 15 ppb in Smith et al. ,83 and the rule-in cut-off point was higher at 110 ppb (or 90 ppb if selecting the cut-off with the highest NPV) compared with 47 ppb respectively. Sensitivity and specificity values were also different (see Table 23).

Subset of patients at Position A compared with airway reversibility or airway hyper-responsiveness

Population: These studies recruited patients who may represent a narrower selection of the full spectrum of patients who present with symptoms of asthma than described in the previous studies. Heffler et al. 82 recruited 48 adults and adolescents with rhinitis and symptoms of asthma whereas Cordeiro et al. 87 recruited 114 patients with a ‘high prevalence of atopy’. However, it would appear that these two studies are in fact reasonably comparable to the studies that recruited a full spectrum of patients at position A. The prevalence of atopy in the study by Heffler et al. 82 was higher than that in the study by Cordeiro et al. ,87 at 92% compared with 71%. The study by Smith et al. 83 was the only study to report the prevalence of atopy among the studies that recruited the fuller spectrum of patients at position A and this study reported a similar prevalence of 77%. Similar to the previous studies, Heffler et al. 82 did not recruit any smokers whereas 9.6% of the participants in the study by Cordeiro et al. 87 were smokers. Mean age, severity and FeNO values were not reported in a way that allowed comparison between studies.

Intervention: Heffler et al. 82 used the Niox chemiluminescent device and Cordeiro et al. 87 used NIOX Flex.

Reference standard: Both studies used a combination of airway reversibility and airway hyper-responsiveness as the reference standard, which is equivalent to the whole UK pathway.

Study design and study setting: The study by Heffler et al. 82 was a prospective consecutive cohort study conducted in Italy in an allergy and immunity clinic. The study by Cordeiro et al. 87 was a retrospective analysis of a prospective database conducted in the Netherlands in secondary care. Neither study was funded by the manufacturers of the FeNO devices.

Estimates of diagnostic accuracy: Table 23 details the estimates of sensitivity and specificity for these studies. In Heffler et al. ,82 the sensitivity and specificity for the highest sum of sensitivity and specificity do not seem noticeably different from the values in studies recruiting the fuller spectrum of patients with symptoms of asthma (77.8% and 60% respectively), although the rule-in scenario achieved 100% for specificity and the rule-out scenario achieved 100% for sensitivity, which was not achieved by the studies with a full spectrum of patients at position A. This study also reported higher values for the paired sensitivity and specificity in the rule-in and rule-out scenarios than the studies with a fuller spectrum of patients at position A. In the study by Cordeiro et al. ,87 sensitivity and specificity were 78% and 92%, which has a similar sum to the highest pair of sensitivity and specificity values reported for studies recruiting the full spectrum of patients, reported by Smith 200486 at 88% and 79% respectively, but with the balance between sensitivity and specificity inverted.

Studies recruiting patients who are difficult to diagnose of relevance to the decision problem

The most appropriate reference standard in the difficult-to-diagnose population in relation to UK guidelines varies according to where in the pathway the group is recruited from, in other words, which tests they have already undergone and which test they would get next. As previously described, when a rule-in scenario is used, patients testing positive would go on to be treated as asthmatic and patients testing negative would go on to have further tests for asthma. When a rule-out scenario is used, patients testing positive would go on to have further tests and patients testing negative would go on to be treated as not asthmatic.

Population

Four studies recruited patients who fall into the difficult-to-diagnose category and were of relevance to the decision problem, but none recruited the same spectrum of patients at exactly the same point in the pathway (see Appendix 10 for more details of study inclusion criteria). Schleich et al. 77 recruited adults with chronic cough who had a negative test for airway reversibility and normal spirometry (position G in the UK pathway). Pedrosa et al. 85 and Katsoulis et al. 81 also recruited patients at position G but in the study by Pedrosa et al. 85 this was not restricted to those with chronic cough and included adolescents as well as adults. Bobolea et al. 88 recruited a somewhat different spectrum of patients who were of all ages and who had a negative test for airway reversibility, normal spirometry and a negative MCT. Mean ages were between 34 and 41 years and FEV₁% when reported was similar at 97% and 104%; however, FeNO values were not reported in a way that allowed comparison between studies. The studies by Schleich et al. ,77 Katsoulis et al. 81 and Pedrosa et al. 85 recruited smokers and atopic patients, although the prevalence of atopy was higher in the study by Pedrosa et al. 85 at 87%, compared with 48% and 45.5% in the studies by Schleich et al. 77 and Katsoulis et al. 81 respectively.

Intervention

The studies by Katsoulis et al. ,81 Pedrosa et al. 85 and Bobolea et al. 88 used NIOX MINO, whereas the study by Schleich et al. 77 used the Niox chemiluminescent device.

Reference standard

All four studies77,81,85,88 used airway hyper-responsiveness as the reference standard, which was appropriate to the UK pathway for the patients selected. The studies by Katsoulis et al. ,81 Schleich et al. 77 and Pedrosa et al. 85 used MCT whereas Bobolea et al. 88 used an adenosine challenge test as patients had already had a negative MCT test.

Estimates of diagnostic accuracy

Table 23 details the estimates of sensitivity and specificity for these studies. Katsoulis et al. ,81 Schleich et al. 77 and Pedrosa et al. 85 reported quite similar cut-offs for the highest sum of sensitivity and specificity at 32, 34 and 40 ppb respectively. The paired estimates of sensitivity and specificity were similar in Katsoulis et al. 81 and Schleich et al. ,77 with sensitivities of 47% and 35% and specificities of 85% and 95% respectively. Pedrosa et al. 85 reported a sensitivity of 74.3% and a specificity of 72.5%. There is no one obvious characteristic of the studies that correlates with the differences in results. Bobolea et al. 88 reported 100% sensitivity but only 29.2% specificity, indicating that FeNO measurement in this position would be most likely to be useful as a rule-out test. No data were available for other cut-off points.

In comparison to studies that recruited patients at position A in the pathway, patient populations are perhaps somewhat younger. Other patient spectrum characteristics look comparable. The range of estimates of sensitivity and specificity also look largely comparable. A sensitivity of 100% was achieved by Bobolea et al. ,88 although it is not clear if this was for the highest sum of sensitivity and specificity or if the cut-off was selected so that FeNO could perform as a rule-out test with high sensitivity.

Studies recruiting patients with chronic cough at position H

Population

The studies by Prieto et al. ,76 Hsu et al. 73 and Hahn et al. 74 all recruited adults with chronic cough who were negative for some other causes of cough (see Appendix 10 for more details on the inclusion and exclusion criteria). Prieto et al. 76 recruited patients with a FEV₁ of at least 80% predicted with chronic cough and no signs of other lung disease. Hsu et al. 73 recruited patients who were negative for upper airway cough syndrome and gastro-oesophageal reflux disease (GORD) and who had no obvious chest radiograph abnormalities. Hahn et al. 74 recruited patients with normal chest radiographs. All three patient groups appear to be equivalent to patients at position H in the UK pathway, who were classed as being at low risk for asthma, who have undergone tests for other conditions and for whom an asthma diagnosis is being reconsidered. All three studies recruited no current smokers and only the study by Prieto et al. 76 reported the prevalence of atopy and this was 100%. Cohorts were perhaps somewhat older than in other studies, with all averaging in the mid- to high 40s.

Intervention

The device used was the NOA 280i in Hsu et al. 73 and Hahn et al. 74 and the Niox chemiluminescent device in Prieto et al. 76

Reference standard

The reference standard was ICS responsiveness, which would be the next test in UK practice for some or all of these patients.

Estimates of diagnostic accuracy

Table 23 details the estimates of sensitivity and specificity for these studies. In the study by Prieto et al. ,76 sensitivity and specificity were poor at 53% and 63%, respectively, although the studies by Hsu et al. 73 and Hahn et al. 74 both report high sensitivities (94.7% and 90% respectively) and fairly high specificities (76.3% and 85% respectively), indicating that FeNO could be a useful rule-out test. It is not clear why the estimates reported by Prieto et al. 76 differ from those reported by the other two similar studies, but it may be the result of differences in patient selection. Although the device used by Prieto et al. 76 was Niox, it is not thought that this would alter estimates of diagnostic accuracy, rather the cut-off points derived.

Other studies of some interest to the assessment

Table 23 details the estimates of sensitivity and specificity for other studies of some interest to the assessment. The study by El Halawani et al. 78 recruited adults with suspected EIB. As in the study by Arora et al. ,79 which was not considered relevant to the review because of the reference standard used, this group of patients was made up of army recruits. It is not clear what previous tests these patients had undergone, if any. None of the patients was a smoker or atopic. The reference standard was an exercise challenge test, which will identify only patients with EIB rather than other forms of asthma. The device used (NOA/Sievers 280A) is of unknown equivalence to NIOX MINO and NObreath. The study reports 100% sensitivity and 31% specificity, indicating that this test could be used as a rule-out test.

The studies by Brannan et al. 92 and Chancafe-Morgan et al. 80 both used airway hyper-responsiveness (MCT or mannitol challenge) as the reference standard but it is not clear if the patient spectrum matches UK practice for patients being referred for such testing. The next most similar studies are those by Schleich et al. ,77 Pedrosa et al. 85 and Katsoulis et al. ,81 all of which use MCT challenge testing as the reference standard, in patients with asthma symptoms. In the study by Chancafe-Morgan et al. ,80 the cut-off point was comparable to that in the other studies at 35 ppb, whereas in the study by Brannan et al. 92 the cut-off point was higher than in any of the other studies at 47 ppb (next highest 40 ppb85), which may reflect the use of a device known to read higher than the Niox chemiluminescent device (despite a correction of 0.6 ppb). Brannan et al. ,92 Katsoulis et al. 81 and Schleich et al. 77 all report sensitivities and specificities in similar ranges, with low sensitivity but high specificity, whereas Chancafe-Morgan et al. 80 and Pedrosa et al. 85 report a more even split between sensitivity and specificity (sensitivities of 75% and 74.3% and specificities of 83.3% and 72.5% respectively).

Diagnostic accuracy meta-analysis

From Table 22 it can be seen that only two sets of two studies are similar enough to each other to warrant meta-analysis:

• Schneider et al. 71,72 and Schneider et al. 69 – two studies conducted by the same research group with populations recruited in 2006–7 and 2010–11 respectively

• Hsu et al. 73 and Hahn et al. 74 – these studies recruited in 2009–10 and 2004–5, respectively, and were conducted in China and the USA respectively.

However, the value of such a meta-analysis is limited given that these studies are no more or less relevant to the decision problem than any of the other studies found.

Studies using FeNO in conjunction with another test as the index test

From the initial 22 studies conducted in adults, adults plus adolescents, all age groups and unspecified age groups, two77,87 reported diagnostic accuracy data for FeNO in conjunction with another test as the index test. One further sudy70 did not report actual data but did state that the addition of certain tests (see next paragraph) did not increase accuracy.

The study characteristics of the studies by Schleich et al. ,77 Cordeiro et al. 87 and Fortuna et al. 70 are presented in Tables 21 and 22 and the diagnostic accuracies in the studies by Schleich et al. 77 and Cordeiro et al. 87 are presented in Table 23. Neither the study by Schleich et al. 77 nor that by Cordeiro et al. 87 reported a change in the optimum cut-off for FeNO when using it in conjunction with another test, but sensitivities and specificities did change. The study by Fortuna et al. 70 was not judged to be of high relevance to the decision problem as it used sputum eosinophilia as part of the reference standard. This test is not widely available in the UK and so this study has low generalisability. The study reported that the addition of sputum eosinophilia to FeNO measurements increased specificity from 64% to 76%; sensitivity was not reported for the two tests together. The authors also stated that the addition of lung function tests and bronchodilator tests did not increase accuracy, but actual data were not provided.

Cordeiro et al. 87 used FeNO with a cut-off of 27 ppb in conjunction with airway reversibility in a population of patients at position A in the UK pathway. If patients were positive by either test they were considered to have tested positive. Compared with using FeNO at a cut-off of 27 ppb alone, sensitivity increased from 78% to 87% whereas specificity decreased from 92% to 90%. However, it should be noted that the reference standard for this study was airway reversibility or airway hyper-responsiveness to histamine. As such, the study results are at high risk of incorporation bias as the reference standard incorporates some of the same results as the index test. This is likely to overestimate the actual diagnostic accuracy of this combination of tests. 131

Schleich et al. 77 used FeNO with a cut-off of 34 ppb in conjunction with a FEV₁% predicted of ≤ 101% in a population of patients with chronic cough and at position E (difficult to diagnose) in the diagnostic pathway. Patients were required to have both a FeNO > 34 ppb and an FEV₁% predicted ≤ 101% to be judged positive by this combination of tests. This resulted in an increase in specificity from 95% to 98.9%, but a decrease in sensitivity from 35% to 24.4%. In this case the reference standard was airway hyper-responsiveness to MCT and so incorporation bias was avoided.

Conclusions

In both cases the improvements in diagnostic accuracy are modest (or negative when considering the sum of sensitivity and specificity) and necessitate the usual trade-off between sensitivity and specificity. As both studies are derivation studies rather than validation studies (in which the cut-off points are pre-set), it is possible that the gains seen are an overestimate of increases in diagnostic accuracy. However, it would seem that using a combination of tests may have additional benefit to using FeNO on its own, and these studies equate more accurately to adding FeNO into the pathway than studies that do not use FeNO in conjunction with other tests.

Studies including children or children and adolescents

Studies using only FeNO as the index test

Four studies that recruited children (plus adolescents and/or young adults) and compared FeNO- guided diagnosis to non-FeNO-guided diagnosis were identified. 93–96 All of the studies were based in secondary care and each study was undertaken in a different country: Finland,93 Switzerland,94 Israel95 and Korea. 96 Funding sources were reported only in the studies by Linkosalo et al. 93 and Woo et al. :96 these were the Tampere Tuberculosis Foundation/Medical Research Fund of Tampere University Hospital and the National Research Foundation of Korea respectively. The study by Sivan et al. 95 declared that there were no conflicts of interest in the research.

Quality assessment

The four studies exploring FeNO measurement for the diagnosis of asthma in children were assessed for quality according to QUADAS-2 criteria for diagnostic accuracy studies. 38 The overall quality was variable, with no one study being free from potential bias in all domains and no single domain being free from bias in all studies. The Woo et al. study96 appeared to be at the lowest risk of bias whereas the studies by Ramser et al. 94 and Linkosalo et al. 93 displayed the highest risk of bias (Figure 12).

FIGURE 12.

Risk of bias summary: review authors’ judgements about each risk of bias item for each included study. Dark green circles with + sign, low risk of bias; light green circles with – sign, high risk of bias; medium green circles with ?, unclear risk of bias.

Risk of bias from patient selection

The studies by Sivan et al. 95 and Woo et al. 96 both appeared to be free of bias in terms of patient selection. Both studies enrolled consecutive samples, avoided a case–control design and recruited appropriately (i.e. those patients presenting with clinical signs of asthma). However, there were potential sources of bias in the studies by Linkosalo et al. 93 and Ramser et al. 94 in that neither study explicitly clarified whether it had enrolled patients consecutively.

Risk of bias from the conduct of the index test

There was potential bias in the conduct of the index test throughout the corpus of literature. All four studies93–96 were derivation studies; hence, in fitting the cut-off points to the data post hoc they are likely to provide liberal estimates of diagnostic accuracy. There was an additional source of possible bias in the Ramser et al. study94 in that it was not clear whether the index test was interpreted blind to the results of the reference standard.

Risk of bias from the conduct of the reference standard

The studies by Sivan et al. 95 and Woo et al. 96 both appeared to provide a satisfactory reference standard in that both adhered to all or part of the UK guidelines and clearly stated that the results were interpreted by a blinded investigator. Neither the study by Linkosalo et al. 93 nor that by Ramser et al. 94 provided sufficient information to confirm whether the result was interpreted blind to the index test results.

Risk of bias from patient flow and timing of the study

The patient flow and test timing appeared to be broadly satisfactory. The studies by Linkosalo et al. ,93 Ramser et al. 94 and Woo et al. 96 each conducted tests consecutively, provided the same reference standard to all patients and included all enrolled patients in the final analysis. The one study that did display a potential source of bias in this domain was that by Sivan et al. 95 These investigators provided a list of criteria that may have been used to confirm a diagnosis of asthma, but it was not clear precisely which of these tests in which combination(s) were given to which patients.

Summary

The small body of research was of variable quality, with the study by Woo et al. 96 displaying the least risk of bias and that by Ramser et al. 94 being at the highest risk of bias. The most important source of potential bias in this literature is concerned with the conduct and interpretation of the index test. All studies fitted FeNO cut-off points to the data post hoc and are thus likely to overestimate diagnostic accuracy. In addition, the study by Ramser et al. 94 did not provide sufficient information to judge whether the index test results had been interpreted blind to the reference standard. Study flow and timing was the least likely domain to contain sources of bias in that only the study by Sivan et al. 95 did not provide sufficient clarity on whether all patients received the same reference standard.

All studies included in the review of diagnostic accuracy of FeNO in children

Study design and timeline of studies

Study characteristics and timelines are provided in Tables 24 and 25. All four studies had a prospective cohort design and, with the exception of the study by Linkosalo et al. 93 (when the relevant information was not reported), they each enrolled consecutive patients. The timing of diagnostic procedures among the studies also appeared broadly comparable. Linkosalo et al. 93 performed FeNO measurement before an exercise challenge test, after which spirometric testing was used at 4, 10 and 15 minutes. Final spirometry occurred at 20 minutes, after salbutamol inhalation. Ramser et al. 94 likewise performed FeNO measurement before pulmonary function assessment and spirometric testing. In addition, patients who did not react to the exercise testing were provided with an additional MCT at 1 hour. Sivan et al. 95 also assessed FeNO first and followed up with spirometry and sputum induction 1–2 hours later, although sputum induction did not contribute to the diagnosis of asthma, which was based on assessment by a certified paediatric pulmonologist after at least 18 months’ follow-up and treatment. Finally, Woo et al. 96 asked all participants to fill in an ISAAC (International Study of Asthma and Allergies in Childhood) questionnaire132 and undergo clinical assessment. FeNO measurements were then taken, followed by spirometry and a MCT.

TABLE 24 - Diagnostic review: study and patient characteristics in studies recruiting children and adolescents

+ve, positive; –ve, negative; CI, confidence interval; GM, geometric mean; NR, not reported; SE, standard error.

Author, year	Study details	Age group	Inclusion/exclusion criteria	No. analysed/no. recruited	Age (years), sex	FEV1% predicted	FeNO (ppb)	Atopic, n/N (%)
Linkosalo 201293	Setting: paediatric allergist, Finland Funding: non-industry Design: prospective cohort study, unclear if consecutive	Children and adolescents	Children and adolescents aged 6–19 years; those with confirmed atopy referred to an allergist with asthma-like symptoms	30/30	Mean age (range): EIB +ve: 10.7 (8–19); EIB –ve: 9.6 (6–13) Code for population: children and adolescents Male, n (%): 20 (66.7)	Mean ± SE: EIB +ve: 97 ± 2; EIB –ve: 96 ± 3 (p = 0.723)	Mean ± SE: EIB +ve: 31.3 (SD 4.1); EIB –ve: 15.6 (SD 3.6)	30/30 (100)
Ramser 200894	Setting: secondary care, Switzerland Funding: NR Design: prospective, consecutive cohort study	Children and adolescents	Children aged 6–16 years referred to an outpatient clinic for diagnostic work on possible reactive airway disease. SABAs must have been withheld on the day of testing and long-acting beta2-agonists withheld for at least 24 hours before testing	169/169	Mean age: NR Code for population: young children Male, n/N (%): 96/169 (57)	Mean (SD): atopic (n = 104): 97 (12); non-atopic (n = 57): 102 (13)	Mean (SD): atopic (n = 104): 35 (36); non-atopic (n = 57): 13 (16)	104/169 (61.5)
Sivan 200995	Setting: secondary care, outpatient clinic, Israel Funding: authors declared no conflict of interests Design: prospective, consecutive patients	Children and adolescents	Children and adolescents Inclusion criteria: (1) non-specific respiratory symptoms suggestive of asthma for ≥ 3 months’ duration, including cough, wheezing, and shortness of breath with or without trials of treatment with bronchodilators and ICSs; (2) children were co-operative and successfully completed all three tests; (3) follow-up at clinic for at least 1 year Exclusion criteria: patients with other conditions that could affect FeNO or sputum eosinophil count, including subjects with symptoms of unresolved respiratory tract infection, with systemic clinical manifestations of atopy such as anaphylaxis, angioedema, food allergy or urticaria or with an underlying systemic or inflammatory disease	150/156 (n = 6 unable to produce sputum)	Mean age (range): steroid-naive asthmatics (n = 69): 12.6 (5–18); asthma treated with ICSs (n = 37): 12.3 (6–18); non-asthmatics (n = 44): 12.0 (7–18) Code for population: children and adolescents Male, n/N (%): steroid-naive asthmatics: 40/69 (58); asthma treated with ICSs: 19/37 (51); non-asthmatics: 24/44 (55)	Mean (SD): steroid-nave asthmatics: 79.3 (44.4); asthma treated with ICSs: 75.0 (16.0); non-asthmatics: 86.1 (17.1)	Mean (SD): steroid-naive asthmatics: 69 (17); asthma treated with ICSs: 36 (57); non-asthmatics: 12.6 (9)	NR
Woo 201296	Setting: secondary care (outpatient clinic), Korea Funding: non-industry Design: prospective, consecutive cohort study	Children and adolescents	Children and adolescents aged 8–16 years Inclusion criteria: children presenting with non-specific respiratory symptoms suggestive of asthma, including cough, wheezing, and shortness of breath. All included patients did not receive inhaled SABAs in the 8 hours before the measurements and were also not receiving a regular treatment with controller medications for ≥ 3 months before evaluation of FeNO and lung function	245/245	Mean (SD) age: 11.7 ± 2.2 Non-atopic asthmatic (n = 38): 11.6 (2.7); non-atopic non-asthmatic (n = 18): 11.4 (2.0); atopic asthmatic: 11.7 (2.4) (n = 129); atopic non-asthmatic (n = 60): 12.6 (2.6) Code for population: children and adolescents Male, n/N (%): non-atopic asthmatic 20/38 (52.6); non-atopic non-asthmatic 9/18 (50.0); atopic asthmatic 92/129 (71.3); atopic non-asthmatic 42/60 (70.0)	Mean (SD): 87.6 (11.6)	GM (95% CI): asthmatic: 23.4 (20.9 to 26.2); non-asthmatic: 12.6 (10.9 to 14.5)	189/245 (77)

TABLE 25 - Diagnostic review: description of interventions in studies recruiting children and adolescents

BHR, bronchial hyper-responsiveness; ERS, European Respiratory Society; NR, not reported; PD20, dose of methacholine needed to cause a 20% fall from baseline in FEV₁.

Author, year	Population	Age group	Device	Cut-off values (ppb)	Reference standard	Details of reference standard	Position of FeNO in the pathway	Relevance to decision problem
Linkosalo 201293	Position A with confirmed atopy	Children and adolescents	NOA 280i (chemiluminescence)	10, 20, 30, 40, 50	Airway hyper-responsiveness to exercise	EIB – free-running test with goal of 80% maximum heart rate according to age. Spirometry 4, 10 and 15 minutes after exercise and after salbutamol inhalation given 20 minutes after exercise. EIB positive if maximal decrease in FEV1 ≥ 12%	FeNO at position 1	Relevant
Ramser 200894	Position A	Children and adolescents	CLD 77 AM (chemiluminescence)	10, 20, 30, 40, 50	Airway hyper-responsiveness (MCT or exercise)	Spirometry, body pleythysmography and MTC challenge according to ATS/ERS guidelines.35 EIB was defined by decrease in FEV1 by ≥ 15% of baseline. MCT challenge was carried out using a panel of incremental dosages of MCT and a dose of 1.8 mg was defined as the threshold of PD20 to differentiate normal airway hyper-responsiveness from BHR	FeNO replaces whole pathway	Not relevant – uses reference standard not used in the UK
Sivan 200995	Position A	Children and adolescents	CLD 88 (chemiluminescence)	15, 18, 19, 25, > 20 or < 15	Exacerbation history, airway reversibility, airway hyper-responsiveness	Patient’s history of two or more clinical exacerbations of wheezing documented by a physician, dyspnoea or cough relieved by bronchodilators, documented variability in FEV1 ≥ 15% in response to bronchodilators at any time during the follow-up period (reversibility), documented variability in FEV1 ≥ 15% over time with or without controller medications (ICSs or montelukast). Results of provocation tests were included when available. Children in whom asthma did not manifest within 18 months of follow-up were considered as not having asthma	FeNO replaces whole pathway	Relevant – uses long-term follow-up
Woo 201296	Position A	Children and adolescents	NIOX MINO	5, 10, 15, 20, 25, 30, 34, 40, 45, 50 (optimum at 22)	Airway reversibility, airway hyper-responsiveness (MCT)	Relevant symptom history and reversible airflow obstruction (≥ 12% improvement in FEV1 in response to inhaled beta2-agonist) and/or airway hyper-responsiveness	FeNO replaces whole pathway	Relevant

Population

The study populations were broadly similar in terms of their position on the diagnostic pathway and all recruited children and adolescents, although upper and lower age cut-offs varied a little, with the most inclusive being those in the study by Sivan et al. 95 at 5–18 years and Linkosalo et al. 93 at 6–19 years and the least inclusive being those in the study by Ramser et al. 94 at 6–16 years. There were some further differences in the inclusion criteria. Linkosalo et al. 93 included only children and adolescents with confirmed atopy whereas Ramser et al. 94 and Woo et al. 96 included a mix of atopic and non-atopic patients. Sivan et al. 95 did not report the number of patients with atopy but did not specifically include on this basis and the study is therefore likely to have included a mix of atopic and non-atopic patients. All studies recruited patients at position A in the UK pathway; Linkosalo et al. 93 recruited patients who had been referred to an allergist with asthma-like symptoms (position A); Ramser et al. 94 included children in position A who had been referred to an outpatient clinic for diagnostic assessment of possible reactive airway disease; Woo et al. 96 included children presenting with non-specific respiratory symptoms suggestive of asthma and who had not been receiving controller medications for at least 3 months prior to FeNO testing (position A); and Sivan et al. 95 also recruited from position A – in this case those with non-specific respiratory symptoms suggestive of asthma for at least 3 months. This study also excluded patients with any other conditions that may have interfered with FeNO testing or sputum eosinophil count, especially unresolved respiratory tract infection or underlying systemic or inflammatory disease.

Sample size ranged from 3093 to 24596 and the mean age (often reported for subgroups rather than whole cohorts) ranged from 9.693 to 12.6 years,95 although mean age was not provided by Ramser et al. 94 Unlike adult studies in which male participants were in the minority, there was a preponderance towards male participants in all four studies, with the lowest percentage being observed in the Sivan et al. 95 study (55.3%).

Interventions

Three of the four studies measured FeNO via chemiluminescence, although each used a different device: Linkosalo et al. 93 used the NOA280i; Ramser et al. 94 used the CLD 77 AM; and Sivan et al. 95 used the CLD 88. Woo et al. 96 was the only study to use NIOX MINO for FeNO evaluation. In terms of FeNO cut-off points, Linkosalo et al. 93 and Ramser et al. 94 both used the same prespecified cut-off points of 10, 20, 30, 40 and 50 ppb. Sivan et al. 95 used cut-offs of 15, 18, 19, 25 and > 20/< 15 ppb. Woo et al. 96 reported a large number of cut-off values, ranging from 5 ppb to 50 ppb.

Reference standard

None of the studies fully replicated the UK guidelines. Linkosalo et al. 93 used an exercise challenge test (free-running test) with spirometric tests before and after exercise and after salbutamol inhalation. Sivan et al. 95 based the diagnosis of asthma on a history of two or more exacerbations, evidence of airway reversibility in response to ICSs or bronchodilators or airway hyper-responsiveness at any time during a period of 18 months’ follow-up. Woo et al. 96 performed a battery of tests similar to those in the UK treatment pathway (spirometry, MCT and atopy assessment), with FeNO being measured before these other tests. Ramser et al. 94 reported results against a reference standard of MCT or exercise challenge testing.

Summary

The study of greatest relevance for this assessment is that by Woo et al.,96 which recruited patients in position A on the pathway and used the NIOX MINO device compared with a reference standard that roughly equates to UK practice. In this study FeNO replaces the whole pathway prior to ICS use.

The three remaining studies were of varying relevance to the UK context:

• Sivan et al. 95 used an ECO MEDICS CLD 88 device in patients at position A in the pathway compared with a reference standard similar to UK practice. In this study FeNO replaces the whole pathway prior to ICS use.

• The study by Ramser et al. ,94 used an Eco Physics CLD 77 AM device for patients in position A on the pathway with a reference standard of airway hyper-responsiveness to exercise or methacholine. As MCT is a very good test for asthma, this study can be seen as similar to testing FeNO against the whole UK diagnostic pathway.

• Linkosalo et al. ,93 used a Sievers NOA280i chemiluminescence device for patients at position A on the pathway with a reference standard of an exercise challenge test. In the UK this test is reserved for those with symptoms of EIB.

No studies in children were identified that incorporated ICS responsiveness in the reference standard.

Estimates of diagnostic accuracy

The sensitivity and specificity values for each of the studies are presented in Appendix 12.

Table 26 also displays three sets of sensitivity and specificity values for each of the studies. These are:

• The highest sum of sensitivity and specificity as reported by the authors of the studies.

• The highest sensitivity – in this scenario a negative test result rules out a diagnosis. This was selected as the cut-off that provided the highest sensitivity. When 100% sensitivity was reported for more than one cut-off, the cut-off that maintained the highest specificity was selected. When the cut-off with the highest sensitivity was not also the cut-off with the highest PPV, this latter cut-off was also presented.

• The highest specificity – in this scenario a positive test result rules in a diagnosis of asthma. Selected as for the highest sensitivity but for specificity. When the cut-off with the highest specificity was not also the cut-off with the highest NPV, this latter cut-off was also presented.

TABLE 26 - Diagnostic review: diagnostic accuracy of FeNO tests in children and adolescents

Author, year	Population	Device	Reference standard	No. analysed	Highest sum of sensitivity and specificity					Rule out					Rule in
					Cut-off (ppb)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	Cut-off (ppb)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	Cut-off (ppb)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
Position A vs. whole pathway
Linkosalo 201293	Position A with confirmed atopy	NOA280i (chemiluminescence)	Airway hyper-responsiveness to exercise	30	20	72	83	86.7	66.7	10	89	33	66.7	66.7	30	50	92	90	55
Ramser 200894	Position A	CLD 77 AM (chemiluminescence)	Airway hyper-responsiveness (MCT or exercise)	169	20	49	76	74	51	10	76	36	63	51	50	20	93	80	45
										20	49	76	74	51
Sivan 200995	Position A	CLD 88 (chemiluminescence)	Exacerbation history, airway reversibility, airway hyper-responsiveness	150	19	86	89	92.2	79.6	15	90	70	82.7	81.6	As highest sum
					> 20 or < 15	89	88	93.5	82.1
Woo 201296	Position A	NIOX MINO	Airway reversibility, airway hyper-responsiveness (MCT)	245	21	56.9	87.2	90.5	50.0	5	94	14.1	70.0	50	41	23.4	100.0	100	37.9

It should be noted that superior sets of sensitivity and specificity values may have in fact been achieved but selection was limited to the range of cut-off points reported within studies.

There was a high degree of agreement as to the cut-off point that produces the highest sum of sensitivity and specificity, despite the heterogeneity in devices and reference standards, with values between 19 and 21 ppb (see Table 26). However, estimates of sensitivity at these cut-off points were not similar across studies, ranging from 49% to 86%; specificity was more similar between studies, ranging from 76% to 89%. Rule-out cut-off points were not similar and varied from 5 to 20 ppb; rule-in cut-off points similarly ranged from 30 to 50 ppb. For ruling out, the highest sensitivity was reported by Woo et al. ,96 at 100%, with a paired specificity of 14.1%, PPV of 70% and NPV of 50%. Sensitivities ranged from 76% to 94%. For ruling in, the highest specificity was also reported by Woo et al. 96 at 100%, with a paired sensitivity of 23.4%, PPV of 100% and NPV of 37.9%. Specificities varied less than sensitivities, from 89% to 100%. It should be noted that superior rule-in and rule-out sets of sensitivity and specificity may in fact have been achieved but selection was limited to the range of cut-off points reported within the studies.

No meta-analysis was performed on these data because of heterogeneity in FeNO measurement devices and reference standards.

Studies using FeNO in conjunction with another test as the index test

One study recruiting children reported estimates of diagnostic accuracy for FeNO in conjunction with another test. 95 This study was described in more detail in the previous section but in summary recruited children at position A in the pathway and used FeNO in conjunction with sputum eosinophilia against a reference standard of evidence of airway reversibility in response to ICSs or bronchodilators or airway hyper-responsiveness at any time during 18 months’ follow-up. Sputum eosinophilia is not a test in widespread use in the UK and the combination of it and FeNO as a diagnostic test is of low relevance to the UK pathway. The results showed that improvements in diagnostic accuracy were very small: sensitivity increased from 86% to 87% and specificity remained the same at 89%.

Management review

This section is broken down into a number of subsections by population age and subgroup. Briefly, these are:

• FeNO-guided management in adults:

  • quality assessment

  • study details

  • estimates of efficacy.

• FeNO-guided management in children:

  • quality assessment

  • study details

  • estimates of efficacy.

• FeNO-guided management in subgroups defined in the scope:

  • pregnant women

  • the elderly

  • adult smokers

  • children exposed to tobacco smoke.

Adults

Four studies that recruited adults and compared FeNO-guided management with non-FeNO-guided management were included in the review. 97–100 One additional study101 was identified from the update search. The study by Shaw et al. 98 was based in the UK, that by Smith et al. 97 was based in New Zealand, that by Syk et al. 99 was based in Sweden, that by Calhoun et al. 100 was based in the USA and that by Honkoop et al. 101 was based in the Netherlands. The studies by Smith et al. ,97 Syk et al. ,99 Calhoun et al. 100 and Honkoop et al. 101 were at least partly supported by Aerocrine and the study by Syk et al. 99 was submitted as part of Aerocrine’s sponsor submission. An additional study by Powell et al. 102 was conducted in adult pregnant women and is discussed separately as this group was defined a priori as a distinct group.

Quality assessment

The quality of the five adult management studies was assessed according to criteria proposed in the Cochrane Handbook for Systematic Reviews of Interventions36 and Systematic Reviews – CRD’s Guidance for Undertaking Reviews in Healthcare. 37 The studies by Powell et al. 102 and Shaw et al. 98 appeared to be the highest-quality articles, with each containing only one potential source of bias (industry sponsorship and uncertain outcome assessor blinding respectively) (Figure 13). The study at highest risk of bias was the unpublished study by Syk et al. ;99 this was because of the lack of participant/personnel blinding, incomplete outcome data and selective reporting. In addition, Calhoun et al. 100 was largely at unclear or high risk, making this study a potential source of bias. The study by Honkoop et al. 101 was published as an abstract and some methodological data were available from the published protocol. However, it is unclear if the execution of the study was per protocol and so quality assessment items were scored as unclear.

FIGURE 13.

Methodological quality summary: review authors’ judgements about each methodological quality item for all included studies. Dark green circles with + sign, low risk of bias; light green circles with – sign, high risk of bias; medium green circles with ?, unclear risk of bias.

Risk of selection bias

All of the included studies were described as randomised and two of the five studies provided satisfactory information on both random sequence generation and allocation concealment. In the study by Shaw et al. ,98 allocation was performed by an independent individual; Syk et al. 99 drew lots from sealed envelopes. There may have been adequate randomisation procedures in the remaining three studies97,100,101 but this could not be confirmed on the basis of the reports.

Risk of performance bias

The study at highest risk from lack of blinding was that by Syk et al. ,99 which was described as an open-label study. The study by Smith et al. 97 was rated as ‘unclear’ on this item as the study was single blind (participants only). As many of the outcomes were patient reported, patient blinding may have been the most important source of bias to avoid, although the blinding of other study personnel who were deciding whether to step patients up or down may also have been important. The study by Honkoop et al. 101 was also scored as ‘unclear’. The remaining two studies98,100 were double-blind and therefore at low risk of performance bias.

Risk of detection bias

The poor reporting of outcome assessment blinding in the studies means that unblinded outcome assessment may be a potentially important source of bias throughout this body of literature. However, as outcome assessment blinding often goes unreported in journal articles, it was unclear whether any potential bias was the result of reporting practices or methodological shortcomings in the conduct of the studies themselves.

Risk of attrition bias

The studies by Shaw et al. 98 and Smith et al. 97 appeared to be at low risk of attrition bias. Dropout rates from these studies were low, adequately reported and corrected for in the statistical analyses. In addition, Smith et al. 97 performed analysis by intention to treat and extrapolated missing data. There was a potentially high risk of bias in the Syk et al. study99 in that patients were missing and not corrected for in multiple analyses. However, Syk et al. 99 did consistently report the number of patients included in each analysis and so the attrition rate was transparent. There were two possible sources of bias in the Calhoun et al. study:100 it was unclear how missing data were corrected for and there were more dropouts in the intervention arm. If these patients were dropping out because of unsatisfactory outcomes (which was not clear from the report), this could skew the results in favour of FeNO. Finally, the study by Honkoop et al. 101 was scored as ‘unclear’ for this item.

Risk of reporting bias

Two of the five studies appeared to have provided data on all of the prespecified outcomes. 97,98 However, there was some evidence of selective reporting in Calhoun et al. 100 and Syk et al. 99 Calhoun et al. 100 failed to report oral prednisone levels, although this had been specified as an outcome in the study protocol, and Syk et al. 99 did not report the number of severe exacerbations. Syk et al. 99 also used medians rather than means in several of the outcomes, precluding these from meta-analysis. However, these data were supplied by the manufacturer (Aerocrine) on request. Finally, the study by Honkoop et al. 101 was scored as ‘unclear’ for this item.

Risk of other bias

There were a number of further potential sources of bias in the studies. Smith et al. 97 reported receipt of commercial sponsorship whereas Syk et al. ,99 Calhoun et al. 100 and Honkoop et al. 101 reported at least partial commercial funding. Three studies97,99,100 also conducted a run-in period before randomisation. It is unclear whether this may have introduced bias to the results in the studies by Smith et al. 97 and Syk et al. 99 In the study by Calhoun et al. 100 patients were excluded if their asthma did not remain controlled when administered two puffs twice a day of beclomethasone HFA (40 µg/puff). This is likely to have influenced the spectrum of patients recruited to this trial towards those with less severe asthma. However, this is likely to affect external validity rather than internal validity as both arms are subject to the same run-in period. No further sources of bias were identified in the remaining studies.

Summary

The quality of the sampled literature was variable, with the study by Shaw et al. 98 being at the lowest risk of bias. The only potential source of bias that we identified in this study pertained to the failure to explicitly state the blinding of outcome assessors. However, it was unclear whether this was an actual methodological flaw or merely inadequate reporting, and at least some of the outcomes were patient reported (patients were blinded). Among the remaining literature, the most important potential source of bias was selective reporting in the Calhoun et al. 100 and Syk et al. 99 studies, both of which failed to report some prespecified outcomes. The study at highest overall risk of bias was the open-label Syk et al. 99 investigation. In addition to lack of blinding, and the aforementioned selective reporting, this study may have been subject to attrition bias. In the absence of information on why data were missing from this study, it is difficult to ascertain how this may have biased the results and in what direction. However, there may be debate about the impact of blinding as a source of bias in these studies. For example, patients could use knowledge of their FeNO readings to guide their self-management, which may capture real-world clinical benefits that would not be observable in double-blind studies.

Study details

Unlike other reviews of FeNO for asthma management, in this review our primary analysis of studies that assess the efficacy of guiding treatment by FeNO measurement in adults considers the study in pregnant women102 separately, as this subgroup of patients was defined a priori as a separate group. This study is described and discussed later. This current section considers the other five studies in adults.

Study design and timeline of studies

Table 28 provides details of study design and the timelines of the studies. All five studies were RCTs. The studies by Smith et al. 97 and Shaw et al. 98 were both single blind whereas that by Syk et al. 99 was an open-label study. The study by Calhoun et al. 100 was described as ‘multiply blinded’, although it is not entirely clear who was blinded. The study by Honkoop et al. 101 did not report blinding. No two studies followed the same timeline exactly. Smith et al. ,97 Syk et al. 99 and Calhoun et al. 100 had a run-in period pre randomisation in which LABAs were reduced or withdrawn and/or doses of ICSs were standardised. Post randomisation, all studies except that by Honkoop et al. 101 had an initial period of time in which visits were more frequent. In the study by Calhoun et al. ,100 visits were made every 2 weeks for the first 6 weeks post randomisation and then were 6-weekly after that. In Shaw et al. 98 initial visits were monthly for 4 months and then every 2 months up to 12 months. In the study by Smith et al. ,97 treatment consisted of two phases: a optimisation phase of 3–12 months and a titration phase of a further 12 months. Patients were randomised before both phases and both phases managed patients according to protocols that either did or did not incorporate FeNO measurements. However, data on exacerbations were reported only for the titration phase and it was these data that were incorporated into the analysis. In the study by Syk et al. 99 participants had an initial visit 2–4 weeks after the initial titration visit, followed by every 2 months up to 4 months and then every 4 months up to 12 months. All studies titrated doses for at least 12 months except that by Calhoun et al. 100 in which doses were titrated for 9 months only. Calhoun et al. 100 included a third intervention arm that was not relevant to this review in which the ICS dose was controlled by matching ICS use on a puff-by-puff basis to the rescue use of albuterol in response to the occurrence of symptoms.

TABLE 28 - Adult management review: study design and timelines

ACQ, Asthma Control Questionnaire; AQLQ, Asthma Quality of Life Questionnaire; ASUI, Asthma Symptoms Utility Index; b.i.d., twice a day; BMQ, Beliefs about Medicines Questionnaire; CostQ, cost questionnaires; EQ-5D, European Quality of Life-5 Dimensions; FACCT, Foundation for Accountability; ICQ, undefined in source document; IgE, immunoglobulin E; IPQ, Illness Perception Questionnaire; MARS, Medication Adherence Report Scale; PC₂₀, provocative concentration that causes a positive reaction; SF-36, Short Form questionnaire-36 items; TTO, time-trade-off.

Author, year	Study design	Timeline of study
Calhoun 2012100	RCT – multiply blinded, multicentre study	Visit 1 (week 0): consent and start of run-in period of 2 weeks – two puffs b.i.d. of beclomethasone HFA (40 µg/puff). If asthma acceptably controlled at this level, enrolled in trial. Visits 2 and 3 (weeks 2–8): pre-randomisation period – patients given two pairs of inhalers to facilitate blinding, one with beclomethasone (2 × 40 µg b.i.d.) and a placebo counterpart and one with albuterol and a placebo counterpart (taken together on demand). Visit 4 (week 8): randomisation to group 1 or group 2. Visits 5–12 (titration): 2, 4, 6, 12, 18, 24, 30 and 36 weeks post randomisation – dose adjustments made at time of clinic visits, monitoring of secondary outcomes
Honkoop 2013101	RCT – cluster design	From protocol: Visit 1: introduction session run by practice nurses; randomisation by GP cluster with stratification on postcode; baseline measurement of IPQ, MARS, BMQ, ICQ, FACCT, SF-36, AQLQ, TTO, ASUI, EQ-5D and CostQ, all at home. Visit 2 (titration): GP measures ACQ, FEV1 and FeNO and titrated dose according to algorithm. Visit 3 (titration): 3 months – GP measures ACQ, FEV1 and FeNO; patients measure AQLQ, EQ-5D and CostQ at home. Visit 4 (titration): 6 months – GP measures ACQ, FEV1 and FeNO; patients measure IPQ, MARS, BMQ, ICQ, FACCT, SF-36, AQLA, TTO, ASUI, EQ-5D and CostQ, all at home. Visit 5 (titration): 9 months – as visit 3. Visit 6 (titration): 12 months – as visit 4
Shaw 200798	RCT – single blind, parallel group	Titration at each visit Visit 0: randomisation – FeNO, FEV1, FVC, PC20, induced sputum analysis, skin prick test, Juniper score. Visit 1: 2 weeks after visit 0 – FEV1, FeNO, Juniper score. Visits 2–5: monthly visits to 4 months – FEV1, FeNO, Juniper score. Visit 6: at 6 months – FEV1, FeNO, Juniper score, PC20, sputum analysis. Visits 7 and 8: at 8 and 10 months respectively – FEV1, FeNO, Juniper score. Visit 9: 12 months – FEV1, FeNO, Juniper score, PC20, sputum analysis
Smith 200597	RCT – single blind, single centre, placebo controlled	Visit 1: enrolment, start of 2-week run-in period in which LABA withdrawn, reinstated at fixed dose if not tolerated. Visit 2 (week 2): FeNO and spirometry; patients begin 4 weeks of 750 µg/day fluticasone or 500 µg/day if previous dose < 200 µg/day. Visit 3 (week 6): randomisation and start of phase 1 Titration phase 1 (3–12 months after randomisation): visits every 4 weeks, FeNO and spirometry, dose adjustment to optimal dose by downwards titration until FeNO ≥ 15 ppb (equivalent to 35 ppb at 50-ml flow rate) or uncontrolled, then uptitrated until controlled/≤ 15 ppb. This dose deemed the ‘optimal dose’ Titration phase 2 (12 months after completion of phase 1): visits every 2 months, upwards adjustments when control lost/FeNO > 15 ppb, downwards adjustment if controlled/FeNO ≤ 15 ppb for two consecutive visits, but not below optimal dose. Treatment orders assigned by blinded investigator. Compliance assessed by inhaler weight
Syk 201399	RCT – open label, parallel group, multicentre	Visit 1: eligibility and consent – capillary blood for IgE confirmation, LABA withdrawn, ICSs continued (salbutamol inhaler with dose counter). Visit 2 (titration): 2–4 weeks later – FeNO, spirometry, reversibility, Juniper mini-AQLQ, generic quality of life, Juniper six-item ACQ, questionnaire on allergen exposure, venous blood for IgE analysis. ICSs and LTRAs altered according to (a) FeNO levels and six fixed treatment steps in FeNO group and (b) usual care (patient report, SABA use, physical examination, pulmonary function tests) in the control group, with FeNO measured but not revealed to treatment decision-maker or patient. Visit 3 (titration): 2 months – ACQ, FeNO and treatment altered. Visit 4 (titration): 4 months – mini-AQLQ, ACQ, FeNO and treatment altered. Visit 5 (titration): 8 months – as visit 3. Visit 6 (titration): 12 months – identical to visit 2. Outcomes recorded at visits 2–6

Population

Table 29 provides details of patient characteristics across studies. All studies were of a moderate size, with numbers analysed ranging from 9497 to 611. 101 All patients were recruited from primary care, except in the study by Calhoun et al. ,100 in which it was not clear whether patients were recruited from primary or secondary care settings. All had either a doctor’s diagnosis of asthma or asthma diagnosed according to guidelines. In the study by Calhoun et al. ,100 the doctor’s diagnosis was confirmed with either a positive MCT or demonstration of airway reversibility. Inclusion and exclusion criteria varied across studies but, when compatible data are reported, study populations seem broadly similar in terms of age (mean ranged from around 34.5100 to 4597 years), FEV₁% (mean ranged from 81.4% to 87.7%) and FeNO values (range of geometric means 18.88–29.0 ppb). It is difficult to determine the comparability of study populations in terms of severity at baseline as different scales for severity and different metrics for medication use have been used. Inclusion and exclusion criteria suggest that at least three studies97,99,100 recruited populations with mild to moderate asthma. Smith et al. 97 excluded those with four or more severe exacerbations in the previous 12 months and those ever admitted to intensive care for asthma, whereas Syk et al. 99 and Calhoun et al. 100 stated that all patients were mild to moderate asthmatics. Smith et al. 97 and Syk et al. 99 also required patients to have been receiving ICS treatment for > 6 months and Calhoun et al. 100 recruited only patients who were well controlled when prescribed two puffs twice a day of beclomethasone HFA (40 µg/puff) and who were ≥ 75% compliant with medication during the 2-week run-in period. Shaw et al. 98 on the other hand may have recruited patients with a wider spectrum of severity. Patients were required only to have had one prescription for asthma medication in the previous 12 months, making it possible that patients with comparatively less severe or less well-documented asthma were included, but excluded only those with severe exacerbations in the previous 4 weeks, making it possible that severe asthmatics were included. Honkoop et al. 101 included patients with a broad spectrum of severity and excluded only very mild asthmatics. Unlike other studies, that by Honkoop et al. 101 had an upper age limit of 50 years.

TABLE 29 - Adult management review: study and population characteristics

ACQ, Asthma Control Questionnaire; AQLQ, Asthma Quality of Life Questionnaire; ASUI, Asthma Symptoms Utility Index; b.i.d., twice a day; C, control group; CI, confidence interval; GM, geometric mean; I, intervention group; IC, intensive care; IgE, immunoglobulin E; ITT, intention to treat; LOCF, last observation carried forward; NR, not reported; WBR, withdrew before randomisation.

Mix of industry and non-industry funding, e.g. research council grants.

37 withdrew, imputation method NR.

13 withdrew, imputation method NR.

Daily score, previous 7 days. Asthma symptoms were scored for each 24-hour period as follows: 0 – no symptoms, 1 – symptoms for one short period, 2 – symptoms for two or more short periods, 3 – symptoms most of the time that did not affect normal daily activities, 4 – symptoms most of the time that did affect normal daily activities, 5 – symptoms so severe as to disrupt daily activities.

FeNO measured at 250 ml/second gives lower values than FeNO measured at 50 ml/second.

Author, year	Study details	Inclusion/exclusion criteria	No. analysed/no. recruited	Age (years), sex, n/N (%)	Spirometry, mean (SD)	Severity, mean (SD)	FeNO (ppb)	Smokers, atopic	Medication use
Calhoun 2012100	Setting: care level NR, USA Funding: mixed;a equipment from Aerocrine	Patients with mild to moderate, well-controlled persistent asthma with compliance rates ≥ 75% who could tolerate treatment of two puffs b.i.d. of beclomethasone HFA (40 µg/puff) during the 2-week run-in period	363 recruited to trial WBR: 21; I: 115/115;b C: 114/114c Other study arm (not included in review): 113/113	Adults (assumed from mean age) Mean (SD) age: I: 34.8 (11.3); C: 34.2 (11.9) Male 75/229 (32.8)	FEV1%: I: 86.3 (10.4); C: 87.7 (12.1)	ACQ: I: 0.79 (0.54); C: 0.72 (0.50) AQLQ: I: 6.16 (0.77); C: 6.27 (0.76) ASUI: I: 0.88 (0.12); C: 0.90 (0.10)	FeNO, GM (SD): I: 18.88 (0.66); C: 21.38 (0.62)	Smokers NR Atopic: 196/229 (85.6%)	Albuterol rescue use, median (IQR): I: 0.07 (0 to 0.43); C: 0.04 (0 to 0.29)
Honkoop 2013101	Setting: Netherlands Funding: mixeda	From protocol: age 18–50 years; doctor’s diagnosis of asthma; patients who need ICSs as controller medication (steps 2–4 GINA guidelines112); ICSs for ≥ 3 months in the previous year; written informed consent; no exacerbation of asthma within 1 month before entry. Exclusions: daily or alternate day OCS therapy for at least 1 month before entering into the study; inability to understand written and oral Dutch instructions; active diseases likely to interfere with the purpose of the study, such as end-stage disease or inability to visit GP	611 randomised Other data NR	43 Male: 32%	NR	NR	NR	NR	NR
Shaw 200798	Setting: recruited from primary care UK Funding: Asthma UK grant; speakers’ fees but not from Aerocrine	Included: patients with GP diagnosis of asthma who received one or more prescriptions for antiasthma medication in the last 12 months; current non-smokers with a past smoking history of < 10 pack-years. Excluded: those poorly compliant; those with a severe asthma exacerbation (needing prednisolone) in the last 4 weeks	118 (ITT LOCF)/119 WBR: 1; I: 58; C: 60	Adults > 18 years Mean age NR Male 54/118 (46)	FEV1%: I: 81.4 (20.9); C: 84.9 (20.1) FEV1/FVC: I: 71 (10.7); C: 72 (9.9)	Juniper score: I: 1.32 (0.65); C: 1.26 (0.75)	Log FeNO, GM (68% CI): I: 29.2 (14.0 to 61.0); C: 31.2 (13.3 to 73.1)	Ex-smokers: I: 22%; C: 25% Atopic: 78/118 (66.1%)	Mean (SD) daily dose of ICS: I: 697 (708) µg; C: 652 (533) µg
Smith 200597	Setting: primary care, New Zealand Funding: mixed;a equipment from Aerocrine	Included: chronic asthma130 managed in primary care; regular ICSs for ≥ 6 months, no dose change in previous 6 weeks. If could not tolerate removal of LABA during run-in allowed to participate if could tolerate a fixed dose. Excluded: four or more courses of oral prednisone in previous 12 months, admission to hospital for asthma in previous 6 months, ever admitted to IC unit for asthma, smokers (current or ex) with history of > 10 pack-years	94/110 WBR: 13; I: 46/48; C: 48/49	Adolescents and adults 12–75 years Mean age (range): 44.8 (12–73) Male 41/110 (37.3)	FEV1%, mean (95% CI): I: 86.4 (80.6 to 92.2); C: 83.1 (76.5 to 89.7)	Symptom score,d mean (95% CI): I: 0.6 (0.4 to 0.8); C: 0.8 (0.6 to 1.1)	FeNO 250 ml/second,e GM (95% CI): I: 7.8 (6.6 to 9.3); C: 6.4 (5.5 to 7.5)	Smokers NR Atopic NR	Bronchodilator use, mean per day previous 7 days (95% CI): I: 0.5 (0.2 to 0.8); C: 0.6 (0.3 to 0.8) ICS NR
Syk 201399	Setting: primary care, Sweden Funding: mixed;a some from Aerocrine	Doctor’s diagnosis of asthma and ICS treatment for ≥ 6 months; IgE sensitisation to at least one major airborne perennial allergen (dog, cat or mite); non-smoker for ≥ 1 year and with smoking history of < 10 pack-years. Patients all had mild to moderate asthma	165/187 WBR: 6; I: 87/93; C: 78/88	Adults (18–64 years) Mean (SD) age 41 (12.4) Male 94/181 (51.9)	FEV1%, mean (SD): I: 84.3 (14.1); C: 83.7 (12.5) FEV1/FVC, mean (SD): I: 0.78 (0.08); C: 0.79 (0.08)	NR	FeNO, GM (95% CI): I: 22.0 (19.3 to 25.2); C: 21.6 (18.7 to 25.0)	Smokers: 0/165 (0%) Atopic: 165/165 (100%)	Median (IQR) budesonide-equivalent ICS dose (µg/day): 400 (400 to 800) LABA before study entry: 54/180 (30.0%)

The study by Smith et al. 97 included smokers (current or ex) with a history of < 10 pack-years whereas the studies by Shaw et al. ,98 Syk et al. 99 and Calhoun et al. 100 all excluded current smokers but included ex-smokers with a past smoking history of < 10 pack-years. Smith et al. ,97 Shaw et al. 98 and Calhoun et al. 100 all included a mix of atopic and non-atopic patients whereas Syk et al. 99 included only atopic patients. Honkoop et al. 101 included atopic patients and smokers. It is unclear whether studies in atopic patients will over- or underestimate efficacy or have no impact at all, although clinical input to the assessment suggested that it would be expected to increase estimates of efficacy as atopy is correlated with ICS responsiveness.

Overall, patient populations recruited by Honkoop et al. ,101 Smith et al. 97 and Shaw et al. 98 are likely to be the most representative of the general asthma population in the UK as these studies included atopic and non-atopic patients. Honkoop et al. 101 included smokers and a broad spectrum of patients from mild to severe, Smith et al. 97 included some smokers and Shaw et al. 98 included a potentially broader spectrum of patients than Smith et al. 97 Calhoun et al. 100 also recruited a mix of atopic and non-atopic asthmatics, but the run-in requirements for treatment tolerance and compliance may mean that generalisation to a wider population is difficult. However, were the application of FeNO management to be limited in the UK to certain populations (e.g. only atopic patients, only stable patients, only mild to moderate patients), data from Calhoun et al. 100 or Syk et al. 99 may be more appropriate.

Interventions

Table 30 provides details of the interventions used in each study. Syk et al. 99 and Honkoop et al. 101 used NIOX MINO. It is not possible to determine whether the other studies used the same devices as this information is not clearly reported. 97,98,100 Smith et al. 97 used an unusual flow rate but justified their conversion to 35 ppb equivalent at 50 ml/second. None of the studies used the same protocol or cut-off points for the management of asthma with FeNO. Syk et al. 99 and Calhoun et al. 100 used FeNO only to guide management, Smith et al. 97 used FeNO only, with a safety measure based on symptoms, bronchodilator use and spirometry, and Shaw et al. 98 used FeNO in addition to the Juniper score, which gauges control through symptoms. The study by Honkoop et al. 101 was similar to that by Shaw et al. 98 in that it used FeNO in conjunction with symptoms, spirometry and medication use [all captured in the Asthma Control Questionnaire (ACQ)]. Doses and medications used also varied from study to study, with Smith et al. 97 and Calhoun et al. 100 titrating only ICSs, Shaw et al. 98 titrating ICSs, LTRAs and bronchodilators and Syk et al. 99 titrating ICSs and LTRAs. Honkoop et al. 101 controlled multiple treatment doses rather than just ICSs. There were some differences between the doses and combinations of treatments indicated and the study allowed for a step-down in treatment on the basis of low FeNO in the presence of moderate symptoms (when FeNO was low for > 3 months) but not in the presence of a high ACQ score. This is in contrast to Shaw et al. 98 and Smith et al. 97 which did not allow step-down if moderate symptoms were present,98 or did not allow the dose to fall below the optimum derived in the first titration phase,97 thus placing a limit on how far ICS could be decreased. The number of cut-off points also varied. Smith et al. 97 used only one cut-off of 35 ppb (equivalent). Honkoop et al. ,101 Shaw et al. 98 and Calhoun et al. 100 each used two cut-offs but at different cut-points, one for titrating down (< 25 ppb, < 16 ppb and < 22 ppb respectively) and one for titrating up (50 ppb, > 26 ppb and > 35 ppb respectively), with an intermediate area in between where symptoms also guided treatment98,101 or the dose remained the same. 100 Syk et al. 99 used three cut-offs, with different values for men and women (< 19 ppb, ≥ 24 ppb and ≥ 30 ppb for men; < 21 ppb, ≥ 25 ppb and ≥ 32 ppb for women). Given the uncertain comparability in FeNO measurements between devices, it is difficult to assess how similar these cut-off points may in fact be.

TABLE 30 - Adult management review: description of the intervention management strategies

BDP, beclomethasone dipropionate; b.i.d., twice per day; IgE, immunoglobulin E; NR, not reported; PRN, not defined in source article; q.d., once per day.

Donated by Aerocrine.

Discussed and supported in journal article. 97

Author, year	Decisions based on flow rate, device and cut-off points	Step-up/step-down protocol	Doses
Calhoun 2012100	Based on FeNO only Flow rate; device: flow rate NR; device NR (protocol states Niox) Cut-offs: well controlled < 22 ppb; controlled 22–35 ppb; undercontrolled > 35 ppb	< 22 ppb = well controlled = down one level; 22–35 ppb = controlled = maintain current level; > 35 ppb = undercontrolled = up one level	Dosing beclomethasone HFA: level 1 = 0 µg/day; level 2 = 80 µg q.d.; level 3 = 160 µg b.i.d.; level 4 = 320 µg b.i.d.; level 5 = 640 µg b.i.d.
Honkoop 2013101	Based on ACQ and FeNO Flow rate; device: protocol states NIOX MINO Cut-offs: FeNO low ≤ 25 ppb, FeNO intermediate > 25 ppb and < 50 ppb, FeNO high ≥ 50 ppb; ACQ strictly controlled ≤ 0.75, ACQ sufficiently controlled > 0.75 and < 1.50, ACQ uncontrolled ≥ 1.50	When ACQ ≤ 0.75: if FeNO ≤ 25 ppb, step down if FeNO > 25 ppb and < 50 ppb, no change if FeNO ≥ 50 ppb, step up When ACQ > 0.75 and < 1.50: if FeNO ≤ 25 ppb and time < 3 months, no change or change to LABA; if time > 3 months, step down ICS if FeNO > 25 ppb and < 50 ppb, step up (treatment choice) if FeNO ≥ 50 ppb, step up ICS by one level When ACQ ≥ 1.50: if FeNO ≤ 25 ppb, step up LABA if FeNO > 25 ppb and < 50 ppb, step up (treatment choice) if FeNO ≥ 50 ppb, step up ICS by two levels	Step 1: SABA as needed; step 2: low-dose ICS or LTRA; step 3: low-dose ICS + LABA or medium- or high-dose ICS or low-dose ICS + LTRA; step 4: add one or both of medium- or high-dose ICS + LABA and LTRA; step 4: add one or both of OCS (lowest dose) and anti-IgE treatment
Shaw 200798	Based on FeNO plus symptoms (Juniper score) Flow rate; device: 50 ml/second; device NR Cut-offs: intermediate 16–26 ppb; high > 26 ppb	Exhaled NO < 16 ppb on first occasion or exhaled NO 16–26 ppb on second occasion: and Juniper score ≤ 1.57 = step down anti-inflammatory treatment, step down bronchodilator treatment once off steroids and Juniper score > 1.57 = step down anti-inflammatory treatment, step up bronchodilator treatment Exhaled NO > 26 ppb: and Juniper score ≤ 1.57 = step up anti-inflammatory treatment, no change in bronchodilator treatment and Juniper score > 1.57 = step up anti-inflammatory treatment, step up bronchodilator treatment once on maximum anti-inflammatory treatment Safety measure: patients on 2000 µg/day of beclomethasone with > 26 ppb FeNO and had not fallen to 60% of baseline had sputum checked. If no eosinophilic inflammation, treatment reduced stepwise unless FeNO increased by > 60% of baseline	Hierarchy of anti-inflammatory treatment: (1) low-dose inhaled steroid (100–200 µg BDP b.i.d.), (2) moderate-dose inhaled steroid (200–800 µg BDP b.i.d.), (3) high-dose inhaled steroid (800–2000 µg BDP b.i.d.), (4) high-dose inhaled steroid (800–2000 µg BDP b.i.d.) plus LRTA, (5) higher-dose inhaled steroid (2000 µg BDP b.i.d.) plus leukotriene antagonist, (6) higher-dose inhaled steroid (2000 µg BDP b.i.d.) plus leukotriene antagonist plus oral prednisolone 30 mg for 2 weeks, then titrate dose, reducing by 5 mg/week Hierarchy of bronchodilator treatment: (1) PRN SABAs, (2) LABAs, (3) LABAs plus theophylline, (4) LABAs plus theophylline plus nebulised bronchodilator
Smith 200597	Based on FeNO, with a safety measure based on symptoms, bronchodilator use and spirometry Flow rate; device: 250 ml/second according to ATS 1999 guidelines;133 assume Niox devicea Cut-offs: equivalent to 35 ppb at 50 ml/second (≥ 15 ppb at 250 ml/second)b	FeNO < 35 ppb (equivalent at 50 ml/second) = asthma controlled FeNO ≥ 35 ppb = asthma uncontrolled. Safety measure: if one or more of the following clinical criteria are met, increase one step: (1) symptom score for previous 7 days ≥ 1 point more than mean run-in and minimum score of 2/5; (2) nocturnal wakening on ≥ 3 nights/week more than mean run-in; (3) mean daily bronchodilator use three or more times that of mean run-in and minimum use 15 occasions during previous 7 days; (4) diurnal peak flow variation ≥ 30% and/or FEV1 of < 85% of baseline	Dose steps: placebo and inhaled fluticasone at 100 µg, 250 µg, 500 µg, 750 µg and 1000 µg Phase 1: until optimal dose reached. Optimal dose = one step higher than that at which control lost Phase 2: up-titrate one step at a time; down-titrate if controlled for two visits but not lower than optimal dose Patients had a personalised self-management plan that instructed them to take oral prednisone 40 mg/day when the morning peak flow fell below 70% of the mean run-in value, until it reached > 85%, at which time they took 20 mg/day for the same number of days
Syk 201399	Based on FeNO only Flow rate; device: according to 2005 guidelines;35 NIOX MINO Cut-offs: < 19 ppb (men), < 21 ppb (women); 19–23 ppb (men), 21–25 ppb (women); ≥ 24 ppb (men), ≥ 26 ppb (women); ≥ 30 ppb (men), ≥ 32 ppb (women)	FeNO < 19 ppb (men), < 21 ppb (women): decrease one step; FeNO 19–23 ppb (men), 21–25 ppb (women): no change; FeNO ≥ 24 ppb (men), ≥ 26 ppb (women): increase one step (no change in treatment step if on step 4 or 5 and using two or fewer inhalations of SABA per week); FeNO ≥ 30 ppb (men), ≥ 32 ppb (women): increase two steps (only if on treatment step 1). Grey zone of 5 ppb applied to avoid frequent dose changes	Steps 1–6: budesonide (µg/day): 0, 200, 400, 800, 800 + LTRA, 1600 + LTRA respectively; fluticasone (µg/day): 0, 100, 250, 500, 500 + LTRA, 1000 + LTRA respectively; mometasone (µg/day): 0, 100, 200, 400, 400 + LTRA, 800 + LTRA respectively

Control

Table 31 provides details of the control interventions used in each study. As with the interventions, none of the studies used the same criteria, protocols or treatment doses for the management of asthma in the control arm of the studies. Generally speaking, the control arms considered symptoms, self-reported medication use and sometimes lung function to guide titration. In terms of similarity to UK practice, Shaw et al. 98 state that BTS/SIGN guidelines8 were followed, using the Juniper scale to score symptoms. It is not clear how similar to UK practice other studies may be.

TABLE 31 - Adult management review: description of the control group management strategies

BDP, beclomethasone dipropionate; NHLBI, National Heart, Lung, and Blood Institute.

Author, year	Decisions based on	Step-up/step-down protocol	Doses
Calhoun 2012100	NHLBI guidelines134 (US version of the SIGN guidelines)	Use severity classification chart, assessing both domains of impairment and risk, to determine initial treatment. Use asthma control chart, assessing both domains of impairment and risk, to determine if therapy should be maintained or adjusted (step up if necessary, step down if possible). Use multiple measures of impairment and risk: different measures assess different manifestations of asthma; they may not correlate with each other; and they may respond differently to therapy. Obtain lung function measures by spirometry at least every 1–2 years, more frequently for not well-controlled asthma. Asthma is highly variable over time and periodic monitoring is essential. In general, consider scheduling patients at 2- to 6-week intervals while gaining control; at 1–6 month intervals, depending on step of care required or duration of control, to monitor if sufficient control is maintained; at 3-month intervals if a step down in therapy is anticipated. Assess asthma control, medication technique, written asthma action plan, patient adherence and concerns at every visit	As for intervention
Honkoop 2013101
Strict strategy	ACQ scores135	When ACQ ≤ 0.75: if < 3 months, no change; if > 3 months, step down When ACQ > 0.75 and < 1.50: step up (choice of treatments) When ACQ ≥ 1.50: step up (choice of treatments)	As for intervention
Sufficient strategy		When ACQ ≤ 0.75: step down When ACQ > 0.75 and < 1.50: no change When ACQ ≥ 1.50: step up (choice of treatments)	As for intervention
Shaw 200798	BTS/SIGN guidelines8 using the Juniper scale to score symptoms	Scored by Juniper scale. Treatment doubled if score > 1.57, treatment halved if score < 1.57 for 2 consecutive months	Step 1: SABA as required; step 2: add inhaled steroid 200–800 µg/day BDP equivalent; step 3: add inhaled LABA; step 4: increase ICS up to 2000 µg/day and addition of fourth drug, e.g. LTRA, theophylline, LABA; step 5: oral prednisolone, high-dose ICS, refer to specialist care
Smith 200597	GINA guidelines;136 symptoms, bronchodilator use, spirometry	GINA uncontrolled asthma criteria:125 (1) symptoms present on > 2 days/week with 24-hour asthma score ≥ 2/5; (2) More than one night-time waking/week; (3) bronchodilator use on more than four occasions/week or on > 2 days per week; (4) variation in PEFR > 20 (amplitude % of mean over previous 7 days); (5) FEV1 < 90% of baseline	As for intervention but without the personalised management plan
Syk 201399	Symptoms, lung function, beta-agonist use	Usual care (patient symptom report, SABA use, physical examination, pulmonary function tests)	Assume same doses as for intervention

Estimates of efficacy

Exacerbations

Exacerbations were reported in all studies but definitions varied (Table 32) and results were not entirely consistent across studies.

TABLE 32 - Adult management review: exacerbation and OCS use rates in adult patients with or without FeNO-guided management

CI, confidence interval; NR, not reported.

At-home measurements: any of the following three criteria, when not associated with the increased asthma symptoms, satisfies treatment failure criteria: (i) pre-bronchodilator morning PEFR of < 65% of the baseline on two consecutive mornings, scheduled measurements; (ii) post-bronchodilator PEFR of < 80% of baseline despite 60 minutes of rescue beta-agonist treatment; (iii) post-bronchodilator PEFR may be taken at any time of day. An increase in albuterol use of more than eight puffs per 24 hours over baseline use for a period of 48 hours or more than 16 puffs per 24 hours for more than 48 hours. In-clinic measurements: (i) pre-bronchodilator FEV₁ values from two consecutive sets of spirometric determinations measured 24–72 hours apart that are < 80% of the baseline pre-bronchodilator value (baseline value for adherence period: FEV₁ value at visit 3; baseline for randomisation period: FEV₁ value at visit 4); all participants found to have a FEV₁ of < 80% of baseline at any centre visit but who are not considered to meet treatment failure or exacerbation criteria must be seen again within 72 hours to have FEV₁ measured or (ii) physician judgement for patient safety or (iii) patient dissatisfaction with asthma control achieved by study regimen or (iv) requirement for open-label ICS or another (non-systemic corticosteroid) new asthma medication (e.g. montelukast) without the addition of systemic corticosteroids.

Asthma scores: 0 (stable): morning PEFR > 75% of best PEFR in 14-day run-in period without deterioration in any symptom scores; 1 (mildly unstable): one or more of the following: (i) bronchodilator use on two or more occasions in 24 hours more than the rounded mean number of occasions during the run-in period, (ii) increase in symptom score of ≥ 1 point compared with the rounded mean during the run-in period, (iii) onset of or increase in nocturnal waking: one or more times in the previous 7 nights more than the rounded mean number of times during the run-in period or morning PEFR of 61–75% without deterioration in any of the above categories; 2 (minor deterioration): morning PEFR of 61–75% of the best PEFR during the run-in period and one or more criteria for an asthma score of 1 or morning PEFR of 41–60% without deterioration in any criteria for an asthma score of 1; 3 (major deterioration): morning PEFR of 41–60% of the best PEFR during the run-in period and one or more criteria for an asthma score of 1; 4 (major exacerbation or medical emergency): morning PEFR of ≤ 40% of the best PEFR during the run-in period regardless of symptoms or attendance at a clinician’s office or emergency department because of severe asthma.

Estimated from graph.

ATS/ERS Task Force Criteria 2009. 137

Author, year, time of outcome	Definition of outcomes	n	Intervention per person-year	Control per person-year	Between-group comparison
Calhoun 2012100	Exacerbation: unscheduled medical contact for increased asthma symptoms that results in the use of OCSs, increased ICSs or additional medication for asthma	229	0.21 (97.5% CI 0.1 to 0.32)	0.23 (97.5% CI 0.1 to 0.37)	‘Did not differ’
	Treatment failure defined as exacerbation or loss of controla		0.27 (97.5% CI 0.14 to 0.39)	0.43 (97.5% CI 0.23 to 0.64)	‘Were not different’
Honkoop 2013101 12 months	Course of oral prednisone	611	0.20 per person-year	Strict: 0.29 per person-year; sufficient: 0.29 per person-year	Odds ratio: vs. strict: 0.52 (95% CI 0.20 to 1.30); vs. sufficient: 0.73 (95% CI 0.28 to 1.85)
Shaw 200798 12 months	Course of oral steroids or antibiotics	118	0.33 (SD 0.69)	0.42 (SD 0.79)	–21% (95% CI –57% to 43%; p = 0.43)
Smith 200597 3–12 months optimisation (exacerbation rates not reported for this period) plus 12 months titration	Minor: global daily asthma scoreb of 2 on ≥ 2 consecutive days	94	Minor:c 0.36	Minor:c 0.75	Minor: p = 0.24
	Major: global daily asthma scoreb of 3 on ≥ 2 consecutive days (or in one day, in the context of a minor exacerbation). Major exacerbation or medical emergency: global daily asthma scoreb of 4 in one day		Major:c 0.13	Major:c 0.14	Major: p = 0.91
	Any minor or major exacerbation		0.49 (95% CI 0.20 to 0.78)	0.90 (95% CI 0.31 to 1.49)	–45.6% (95% CI –78.6 to 54.5; p = 0.27)
	Course of oral prednisone		22 events in 46 patients (0.48 events per patient)	29 events in 48 patients (0.60 events per patient)	p = 0.60
Syk 201399 End points analysed from visit 2 to visit 6 (2–4 weeks to 12 months)	Moderate exacerbationd – need to step up controller treatment for at least 2 days with or without clinic visit. Prophylactic use before pollen season excluded	165	0.1	0.325	NR
	Severe exacerbationd – worsening requiring a course of OCSs		0.113	0.0875	Not significant
	Moderate or severe exacerbation		0.22	0.41	Total: p = 0.024

Major or severe exacerbations

This outcome was defined differently across studies. Smith et al. 97 reported two such outcomes: ‘major exacerbations’ defined according to global daily asthma scores and exacerbations leading to a course of oral prednisone. A similar outcome, ‘worsening requiring a course of oral prednisone’, was also reported in Syk et al. 99 Shaw et al. 98 did not report rates of oral prednisone use alone but did report a composite outcome of ‘exacerbations resulting in the use of oral prednisone or antibiotics’. Calhoun et al. 100 reported an outcome called ‘exacerbations’, which included exacerbations leading to oral prednisone use, increased ICS use or additional medication for asthma. This last definition may incorporate exacerbations that other studies would have classified as moderate or minor, although the study does define an additional outcome called ‘treatment failure’, which is likely to incorporate minor, moderate and major exacerbations. As such, the outcome ‘exacerbations’ in the study by Calhoun et al. 100 will be considered in this analysis. Honkoop et al. 101 reported courses of oral prednisone, as in Smith et al. 97 and Syk et al. 99

Honkoop et al. 101 and Shaw et al. 98 reported lower rates per person-year of major/severe exacerbations in the intervention arm and Smith et al. 97 reported lower rates per person (data per person-year not reported) but the difference did not reach statistical significance compared with the control arm in any of the studies (these data were available only as odds ratios for Honkoop et al. 101). Syk et al. 99 reported higher rates per person-year of oral prednisone use in the intervention arm, but the level of significance was not reported. Calhoun et al. 100 showed very similar rates per person-year of exacerbations in both arms of the trial, with no statistically significant difference between them. The best improvement in major/severe exacerbations per person-year was seen in the study by Shaw et al. ,98 at –21% (95% CI –57% to 43%; p = 0.43) (reviewer-calculated rate ratio 0.79, 95% CI 0.66 to 0.94), and the worst improvement was seen in the study by Syk et al. ,99 which reported a higher rate per person-year, although not statistically significantly so, in the intervention arm (0.113) than in the control arm (0.0875) (p-value not reported) (reviewer-calculated rate ratio 1.29, 95% CI 0.83 to 2.03).

Despite the high level of heterogeneity in study characteristics, an exploratory meta-analysis of the rates of major/severe exacerbations was performed. As data per person-year were not reported in or calculable for Smith et al. ,97 this study was excluded from the meta-analysis. The standard error (SE) was not reported in the study by Honkoop et al. 101 and analyses were performed without these data and with various imputed SEs. The pooled estimate of the rate ratio without the data from Honkoop et al. ,101 using random effects methods (Figure 14a), was 0.94 (95% CI 0.66 to 1.34), with a p-value of 0.73. This indicated no difference between the two intervention groups in major or severe exacerbations. The I² statistic was 52%, however, indicating moderate heterogeneity between studies.

FIGURE 14.

Random effects meta-analysis of the effects of FeNO-guided asthma management on major/severe exacerbation rates. IV, inverse variable. (a) All studies with data available, excluding Honkoop et al. 101 (unknown SE); and (b) all studies, imputing the SE for Honkoop et al. 101 at 0.2.

A SE of 0.1, 0.2, 0.3 or 0.4 was imputed for Honkoop et al. ,101 based on the range of errors observed in other studies. Depending on the error imputed, rate ratios ranged from 0.82 (95% CI 0.64 to 1.05; p = 0.11) to 0.89 (95% CI 0.67 to 1.17; p = 0.40), which is not statistically significant. Figure 14b presents the analysis imputing an error of 0.2.

In a prespecified sensitivity analysis, only studies that reported data relating to exacerbations resulting in the use of OCSs were included. Only Syk et al. 99 and Honkoop et al. 101 reported this outcome although Honkoop et al. 101 did not report SEs, leaving only the study by Syk et al. 99 This study reported a rate ratio of 1.29 (95% CI 0.83 to 2.03), indicating no significant difference between the intervention groups (p = 0.26) (Figure 15).

FIGURE 15.

Sensitivity analysis removing studies with wider definitions of major/severe exacerbations: all studies with data available, excluding Honkoop et al. 101 (unknown SE). IV, inverse variable.

In further sensitivity analyses, SEs of 0.1, 0.2, 0.3 and 0.4 were imputed for Honkoop et al. 101 based on the range of errors observed in other studies. Depending on the error imputed, rate ratios ranged from 0.91 (95% CI 0.47 to 1.77; p = 0.79) to 1.00 (95% CI 0.53 to 1.90; p = 1.00), indicating no significant differences between the intervention groups. Heterogeneity statistics were high, ranging from 80% to 53% and reflecting the opposite direction of effect reported in these two studies.

Sensitivity analyses including imputed data for Smith et al. 97 were not conducted and it is unclear how the exclusion of these data affects the meta-analyses.

All exacerbations or treatment failures

When considering other, wider definitions of exacerbation, as described in Table 32, three studies report composite outcomes that can be considered to be broadly similar and which represent what may be termed ‘treatment failure’. In the studies by Smith et al. 97 and Syk et al. 99 this was ‘any major or minor exacerbation’, whereas in the study by Calhoun et al. 100 it was exacerbation or any loss of control by a variety of measures (see footnotes to Table 32 for details). In the studies by Smith et al. 97 and Calhoun et al. ,100 FeNO-guided groups showed numerically but not statistically significantly lower rates of treatment failure. In the study by Syk et al. 99 the improvement was statistically significant, with a rate of 0.22 in the intervention arm compared with 0.41 in the control arm (p = 0.024) (reviewer-calculated rate ratio 0.52, 95% CI 0.44 to 0.61).

Despite the high level of heterogeneity in study characteristics, an exploratory meta-analysis of these rates using random effects methods (the I² statistic was 0%) was conducted (Figure 16). The pooled relative risk (RR) was 0.53 (95% CI 0.46 to 0.61), which represents a statistically significant effect in favour of using FeNO-guided management in asthmatics for this outcome (p < 0.00001).

FIGURE 16.

Meta-analysis of the effects of FeNO-guided asthma management on the composite outcome of major/severe, moderate and minor exacerbation rates/treatment failures. IV, inverse variable.

Moderate and/or minor exacerbations

Smith et al. 97 and Syk et al. 99 both reported the rates of less severe exacerbations separately from the rates of all exacerbations and from the rates of major/severe exacerbations (see Table 32). In both cases the point estimate reduction in minor/moderate exacerbations was far greater than the reduction in severe/major exacerbations. Smith et al. 97 reported 0.36 minor exacerbations per person-year in the intervention arm and 0.75 per person-year in the control arm, with a p-value of 0.24. Syk et al. 99 reported 0.1 moderate exacerbations per person-year in the intervention arm and 0.325 in the control arm. The p-value was not reported. When considering the results reported by Calhoun et al. 100 for exacerbations alone and the composite outcome treatment failure, it can be seen that the larger difference in rates of treatment failure in favour of the intervention arm is not driven by the exacerbation rates, which are very similar at 0.21 (97.5% CI 0.1 to 0.32) and 0.23 (97.5% CI 0.1 to 0.37), and it must therefore be due to a decrease in less severe exacerbations/loss of control in the intervention arm. The impact on quality of life and the costs of such exacerbations are much lower than for major/severe exacerbations.

Inhaled corticosteroid use

All studies except that by Honkoop et al. 101 provided some data on ICS use and these are presented in Table 33. Smith et al. 97 and Shaw et al. 98 reported ICS use as a mean per day at the end of the study, with mean differences of –270 µg/day (95% CI –430 µg/day to –112 µg/day; p = 0.003) and –338 µg/day (95% CI –640 µg/day to –37 µg/day; p = 0.028), respectively, in favour of FeNO-guided management. Syk et al. 99 reported median values that were not statistically significantly different. Means were supplied on request but without significance tests and showed a small increase in ICS use in the intervention arm [586 µg (SE 454 µg) vs. 540 µg (SE 317 µg) in the control arm]. Calhoun et al. 100 reported mean use per month, although it is unclear if this was an average over the whole course of the study or the mean for the final month of the study. The means were very similar at 1617 µg/month in the intervention group and 1610 µg/month in the control group. It should also be noted that this study managed and followed patients for only 9 months whereas the other studies did so for 12 months.

TABLE 33 - Adult management review: ICS use

CI, confidence interval; NR, not reported.

Negative values indicate that the FeNO group had lower ICS use.

Beclomethasone dipropionate or equivalent.

Fluticasone or equivalent.

The equivalent dose for budesonide when a different drug (e.g. fluticasone) has been used.

Author, year	ICS measurement	Intervention	Control	Between-group difference expressed as intervention minus controla
Calhoun 2012100	ICS use (unclear if mean over whole study or final value)b	Mean 1617 µg/month	Mean 1610 µg/month	NR
Shaw 200798	Final value ICS useb	557 µg	895 µg	Mean difference –338 µg/day (95% CI –640 µg to –37 µg; p = 0.028)
	Total used in study (AUC)			11% greater use in FeNO group (95% CI –15% to 37%)
Smith 200597	Final value ICS usec	Baseline: mean 411 µg/day (95% CI 344 µg to 478 µg); end of phase 2: mean 370 µg/day (95% CI 263 µg to 477 µg)	Baseline: mean 491 µg/day (95% CI 403 µg to 579 µg); end of phase 2: mean 641 µg/day (95% CI 526 µg to 756 µg)	Mean difference –270 µg/day (95% CI –112 µg to –430 µg; p = 0.003)
Syk 201399	ICS used	Median 0 (IQR –400 to 400) µg; baseline: mean 604 (SE 370) µg; final value: 586 (SE 454) µg	Median 0 (IQR –200 to 200) µg; baseline: mean 626 (SE 391) µg; final value: 540 (SE 317) µg	0.945

When looking at mean use over time (graphical data not reproduced here) in Smith et al. 97 and Syk et al. ,99 ICS use fell initially in the FeNO arm (both when compared with baseline and in comparison to the control arm) and then rose at the final measurement to a level above that in the control arm in Syk et al. 99 but staying below that in the control arm in Smith et al. 97 Conversely, in the Shaw et al. 98 study, ICS use initially rose and then fell at the final two measurement points to below the baseline level and below the control arm level. Only Shaw et al. 98 reported the AUC for ICS use and this showed an 11% greater use of ICSs in the FeNO group. Based on the ‘mean use over time’ figures, this is unlikely to be true for the study by Syk et al. ,99 in which a visual interpretation of the AUC would suggest very similar levels of total ICS use in both arms, with little change over time. Appropriate data were not available for the study by Calhoun et al. 100 or that by Smith et al. 97 These differences may be the result of the different titration protocols and cut-off values used in the studies and it is difficult to draw a generalised conclusion as to the direction of effect and the trends over time for ICS use. However, it would seem most likely that ICS use will either remain the same or fall in FeNO-managed groups when taken as an average over the course of the first year. The first year of titration is likely to be when the greatest gains are made, as patients reach a stable dose. It is unclear how ICS use will change in the following years as no study reported results beyond 1 year of follow-up, as the severity of the disease may progress, stay stable or remiss over time.

Despite the high level of heterogeneity in study characteristics, an exploratory meta-analysis of ICS use incorporating data from all four studies was conducted (Figure 17). As studies reported values for different ICSs (fluticasone, beclomethasone and budesonide), a standardised mean difference analysis was performed. A random-effects model was used as both clinical and statistical heterogeneity were high, but the I² statistic remained high at 75%. SDs for Calhoun et al. 100 were imputed based on consideration of the other three studies. Sensitivity analyses in which the imputed SDs were altered by an order of magnitude in either direction, and in which a value of 10,000 was used for the intervention arm and 5000 for the control arm (to mirror the SDs of Syk et al. 99), did not have a big effect, with the pooled-analysis CIs crossing the line of no effect in every case and the pooled mean value ranging from –0.25 to –0.23 standardised mean difference. The results of the meta-analysis agree with the conclusions drawn from the narrative consideration of the data; it would seem most likely that ICS use will either remain the same or fall in FeNO-managed groups, probably depending on factors such as step-up/step-down protocols, cut-off values selected, treatments incorporated in the treatment protocol and comparator interventions.

FIGURE 17.

Meta-analysis of the effects of FeNO-guided asthma management on mean ICS use (standardised mean difference analysis). IV, inverse variable.

Health-related quality of life

Syk et al. 99 Calhoun et al. 100 and Honkoop et al. 101 reported quality of life data. This was measured by the mini Asthma Quality of Life Questionnaire (mAQLQ) in Syk et al. 99 and the Asthma Quality of Life Questionnaire (AQLQ) in Calhoun et al. 100 and Honkoop et al. 101 In all studies the overall score, and in Syk et al. 99 three of four domains, did not show a statistically significant change over time. The symptoms domain did, however, show a relatively small but statistically significant between-group difference in change from baseline of 0.10 (Table 34) in Syk et al. 99 An exploratory meta-analysis of the overall scores (Figure 18) showed no effect, with a standardised mean difference of 0.00 (95% CI –0.20 to 0.20). In this case, data from the study by Honkoop et al. 101 were not included as there was not enough information provided to calculate a mean AQLQ score across the two control groups.

TABLE 34 - Adult management review: HRQoL

CI, confidence interval; GQLI, Gothenburg Quality of Life Instrument.

Author, year	Intervention	Control	Between-group difference
Calhoun 2012100	AQLQ change from baseline 0.02 (97.5% CI –0.14 to 0.18), p = 0.75	AQLQ change from baseline 0.02 (97.5% CI –0.14 to 0.17), p = 0.80	AQLQ between-group difference 0.00 (97.5% CI –0.22 to 0.23), p = 0.96
Syk 201399	Appears to be some data missing (n = 78–86) Total change over time in mAQLQ (n = 80/87), median (IQR): 0.23 (0.07 to 0.73); final mean (SE) value 6.07 (0.90) Visit 2 and visit 6 data, median (IQR): mAQLQ symptoms – visit 2: 5.60 (4.80 to 6.20), visit 6: 6.00 (5.60 to 6.60); activity limitation – visit 2: 6.50 (5.75 to 6.75), visit 6: 6.75 (6.00 to 7.00); emotional function – visit 2: 6.00 (4.67 to 6.67), visit 6: 6.33 (5.67 to 7.00); environmental stimuli – visit 2: 6.00 (5.00 to 6.67), visit 6: 6.33 (5.67 to 6.67) GQLI change (n = 85/88): 0.06 (–0.22 to 0.28)	Appears to be some data missing (n = 77–85) Total change over time in mAQLQ (n = 77/78), median (IQR): 0.07 (–0.20 to 0.80); final mean (SE) value 5.98 (0.83) Visit 2 and visit 6 data, median (IQR): mAQLQ symptoms – visit 2: 5.70 (4.80 to 6.40), visit 6: 6.00 (5.20 to 6.40); activity limitation – visit 2: 6.25 (5.50 to 7.00), visit 6: 6.50 (5.75 to 7.00); emotional function: visit 2: 6.00 (4.67 to 6.67), visit 6: 6.00 (5.33 to 6.67); environmental stimuli: visit 2: 5.67 (5.00 to 6.67), visit 6: 6.33 (5.33 to 6.67) GQLI change (n = 78/78): 0 (–0.39 to 0.39)	Analyses of median (IQR) change between visit 2 and visit 6: mAQLA overall: p = 0.197; mAQLQ symptoms: p = 0.041; activity limitation: p = 0.544; emotional function: p = 0.596; environmental stimuli: p = 0.193; GQLI: p = 0.666

FIGURE 18.

Meta-analysis of HRQoL outcomes. IV, inverse variable.

Asthma control and other medication use

Four studies reported data for asthma control. 97–100 Smith et al. 97, Calhoun et al. 100 and Shaw et al. 98 reported no change in asthma control, whereas Syk et al. 99 reported a statistically significant difference in change in ACQ score from visit 2 to visit 6 between the two trial arms (Table 35). This matches the change seen in the AQLQ symptoms domain previously mentioned. Smith et al. ,97 Calhoun et al. 100 and Syk et al. 99 reported use of other medications; Smith et al. 97 and Calhoun et al. 100 reported no significant difference between groups for bronchodilator use, although in the study by Calhoun et al. 100 there was a trend towards less use in the intervention arm, and Syk et al. 99 reported non-significant trends towards greater numbers using LTRAs and higher mean use of LTRAs and SABAs (significance not reported) in the FeNO-controlled arm.

TABLE 35 - Adult management review: other outcomes

ASUI, Asthma Symptoms Utility Index; CI, confidence interval; NA, not applicable; NR, not reported.

Author, year	Outcome	Intervention	Control	Between-group difference
Asthma control
Calhoun 2012100	Night-time symptoms, difference from beginning to end of treatment period using model-based estimates (97.5% CI)	0.01 (–0.00 to 0.02), p = 0.07	0.01 (–0.00 to 0.02), p = 0.11	0.00 (–0.02 to 0.02), p = 0.86
	Daytime symptoms, difference from beginning to end of treatment period using model-based estimates (97.5% CI)	–0.00 (–0.02 to 0.02), p = 0.86	0.01 (–0.00 to 0.03), p = 0.06	–0.01 (–0.04 to 0.01), p = 0.17
	ACQ score, difference from beginning to end of treatment period using model-based estimates (97.5% CI)	–0.01 (–0.15 to 0.12), p = 0.81	0.03 (–0.10 to 0.16), p = 0.64	–0.04 (–0.23 to 0.15), p = 0.62
	ASUI score, difference from beginning to end of treatment period using model-based estimates (97.5% CI)	0.01 (–0.02 to 0.04) p = 0.40	0.01 (–0.02 to 0.03) p = 0.64	0.00 (–0.04 to 0.04) p = 0.79
Shaw 200798	Asthma control	Data NR. No difference between groups in Juniper score throughout the study; however, in both groups the score decreased from baseline. Significance NR
Smith 200597	Symptom score (daily score previous 7 days): final scores, mean (95% CI)	0.4 (0.1 to 0.7)	0.6 (0.4 to 0.9)	p = 0.23
	Nocturnal waking (nights/week, previous 7 days): final scores, mean (95% CI)	0.2 (0.0 to 0.6)	0.2 (0.0 to 0.4)	p = 0.89
	Asthma score (% of days), mean (95% CI)	Score 0: 85.2 (78.4 to 92.0); score 1: 14.0 (7.4 to 20.6); score ≥ 2: 0.8 (0.3 to 1.3)	Score 0: 78.5 (70.4 to 86.6); score 1: 19.9 (12.3 to 27.5); score ≥ 2: 1.7 (0.3 to 3.1)	Between-group final scores (not change from baseline) p = 0.19
Syk 201399	ACQ score change between visit 2 and 6, median (IQR)	–0.17 (–0.67 to 0.17) (n = 81/88)	0 (–0.33 to 0.50) (n = 74/78)	p = 0.045
Other medication use
Calhoun 2012100	Albuterol rescue use (puffs/day), difference from beginning to end of treatment period using model-based estimates (97.5% CI)	–0.04 (–0.10 to 0.02), p = 0.15	0.02 (–0.03 to 0.08), p = 0.30	–0.06 (–0.14 to 0.02), p = 0.08
Shaw 200798	Medication use	NR	NR	NR
Smith 200597	Bronchodilator use (occasions/day, previous 7 days), mean (95% CI)	0.4 (0.1 to 0.7)	0.4 (0.1 to 0.6)	p = 0.98
	Safety buffer criteria used	16/436 assessments	NA	NA
Syk 201399	LTRA use, n/N (%)	33/92 (35.9)	19/85 (22.4)	p = 0.069
	Mean months on LTRA	2.87 (4.42)	1.81 (3.89)	p = 0.094
	SABA use between visit 5 and visit 6 (8–12 months), median (IQR)	1.56 (0.06 to 5.18)	0.94 (0.03 to 2.81)	NR

Adverse events, mortality, compliance and test failure rates

No data were reported in the four studies for adverse events or mortality, although Calhoun et al. 100 reported one unrelated adverse event (hip surgery) in the control arm. Compliance was reported by Smith et al. 97 and Calhoun et al. 100 and was 85% and 89% in the intervention and control arms, respectively, in Smith et al. 97 and ≥ 95% (median) in both groups in Calhoun et al. 100 No test failure rates for NIOX MINO or NObreath were reported.

Children

Five studies103–107 that recruited children (plus adolescents and/or young adults) and compared FeNO- guided management to non-FeNO-guided management were identified from the initial search and a further two studies108,109 were identified during the update search. The study by Fritsch et al. 103 was based in Vienna, Austria; that by Szefler et al. 104 was based in the USA; that by Verini et al. 105 was based in Italy; that by Pijnenburg et al. 106 was based in Rotterdam, the Netherlands; that by Petsky et al. 107 was based in Australia; that by Peirsman et al. 109 was conducted in Belgium; and that by Pike et al. 108 was the first UK-based study identified on this topic in children. Fritsch et al. 103 received technical and analytical support from Aerocrine; one of the authors in the study by Szefler et al. 104 received speaker fees from Aerocrine; the study by Pijnenburg et al. 106 was supported by the Kroger Foundation/Sophia Children’s Hospital Foundation, although Aerocrine had provided a grant to the department; the study by Petsky et al. 107 was funded by the Royal Children’s Hospital Foundation, Asthma Foundation of Queensland; the study by Pike et al. 108 was funded by Sparks charity; the study by Peirsman et al. 109 was funded by Merck & Co., with equipment provided by Aerocrine; and Verini et al. 105 did not report their source of funding.

Quality assessment

Study quality varied, with no one study scoring well on every item and no item scoring well in every study (Figure 19). Of the studies included in the review, the study with the highest overall quality appeared to be that by Szefler et al. ,104 in which the only potential source of bias identified was study funding by a company with a commercial interest in FeNO measurement. The study with the lowest quality appeared to be that by Petsky et al. ,107 which was not scored as ‘low risk’ on any of the quality assessment items. As it was a conference abstract, Petsky et al. 107 was at especially high risk of selective reporting, whereas Verini et al. 105 was at risk of bias as the statistical comparison data were presented poorly (as discussed in the following sections). Potential sources of bias for the evidence base as a whole are discussed in the following sections.

FIGURE 19.

Methodological quality summary: review authors’ judgements about each methodological quality item for all included studies. Dark green circles with + sign, low risk of bias; light green circles with – sign, high risk of bias; medium green circles with ?, unclear risk of bias.

Risk of selection bias

All of the included studies were described as randomised. However, only the study by Szefler et al. 104 provided sufficient information on sequence generation and allocation concealment. Pike et al. 108 did demonstrate random sequence generation but the method of allocation concealment was unclear. In all other studies the method of sequence generation and allocation concealment was not reported.

Risk of performance bias

In terms of blinding, Szefler et al. ,104 Pijnenburg et al. 106 and Pike et al. 108 appeared to have performed adequate blinding for both participants and study personnel. Fritsch et al. 103 blinded participants but did not report whether this was the case for study personnel and so this study was rated as ‘unclear’ on this item. Peirsman et al. 109 blinded participants but not personnel. Neither Petsky et al. 107 nor Verini et al. 105 provided sufficient information to make a judgement on participant and personnel blinding and so these studies were rated as ‘unclear’. As many of the step-up/step-down protocols and criteria for exacerbations were based on participant symptom reporting and physician judgement, any potential lack of blinding in the studies could significantly affect the direction and size of the outcomes.

Risk of detection bias

Szefler et al. ,104 Peirsman et al. 109 and Pike et al. 108 clearly stated that outcome assessment blinding had been performed. The poor reporting of outcome assessment blinding in the other studies means that unblinded outcome assessment may be a potentially important source of bias throughout this body of literature.

Risk of attrition bias

In terms of outcome data completeness, Pijnenburg et al. ,106 Szefler et al. 104 and Peirsman et al. 109 appeared to be at low risk of bias. There may have been some bias in terms of outcome data completeness in the remaining four studies. Fritsch et al. 103 did not report the reasons for participant withdrawal or correction for missing FeNO values and there may have been missing outcome data in the study by Petsky et al. ,107 but this is unknown as only a conference abstract of this study was identified. In the study by Verini et al. 105 it was reported that 64 patients were recruited; however, it was unclear whether this was the total number after dropout or whether no participants dropped out. Consequently, the study was rated as ‘unclear’ on outcome data completeness. Pike et al. 108 reported 23% of dropouts in the intervention arm but only 7% in the control arm and was scored as being at high risk of attrition bias.

Risk of reporting bias

Selective reporting risk may also have been present in some of the data. The study by Petsky et al. 107 was reported in a conference abstract and so this was rated as ‘high risk’ for this item. In the study by Fritsch et al. ,103 medication usage was reported as median (IQR) values rather than mean values and so these values could not be used in the planned meta-analysis; however, although planned, a meta-analysis was not possible because of study heterogeneity and so this bias did not affect the synthesis of data. The study by Verini et al. 105 was rated as having a low risk of bias for selective reporting; respiratory function and immunoallergological parameters were inadequately reported (i.e. no numerical data were provided; it was stated only that there were no significant between-group differences on these outcomes), but these outcomes were not of relevance to this review. Szefler et al. ,104 Pijnenburg et al. 106 Pike et al. 108 and Peirsman et al. 109 appear to have reported all of the outcomes that they set out to measure and so these studies were rated as ‘low risk’.

Risk of other bias

There were a number of further potential sources of bias in each of the studies. Fritsch et al. 103 and Szefler et al. 104 were both in receipt of sponsorship from the pharmaceutical industry and there was evidence of some such sponsorship in the studies by Peirsman et al. 109 and Pijnenburg et al. 106 The statistical comparison data reported in Verini et al. 105 was of poor quality in that most comparisons were presented as within-group longitudinal trends. Pike et al. ,108 Pijnenburg et al. ,106 Szefler et al. 104 and Fritsch et al. 103 all conducted run-in periods before randomisation, which have an unknown risk of bias attached to them. Finally, it was unclear whether there may have been additional sources of bias in the study by Petsky et al. ,107 as this research was presented as a conference abstract only.

Summary

The quality of the sampled literature was variable, with selective reporting being the most common potential source of bias. The studies by Fritsch et al. ,103 Petsky et al. 107 and Verini et al. 105 were rated as ‘unclear’ on the majority of quality items. Such ratings were given for those aspects of study design that were not clearly presented within the articles themselves, meaning that it was unclear whether there is likely to be bias in the conduct of the studies themselves or whether the lack of clarity was a result of inadequate reporting. Other common potential sources of bias were random sequence generation, allocation concealment and outcome assessor blinding. Pharmaceutical industry sponsorship was declared as the source of funding by Fritsch et al. 103 and Szefler et al. 104 and at least partially funded the activities of one author in the studies by Pijnenburg et al. 106 and Peirsman et al. 109 The studies at lowest risk of overall bias appeared to be those by Szefler et al. 104 and Pike et al. 108 (low risk on six and five of seven items respectively) and the studies by Szefler et al. ,104 Pike et al. 108 and Pijnenburg et al. 106 were the only studies in which it was possible to confirm blinding of both participants and study personnel. Possible lack of participant blinding may be a particularly important source of bias given that the study outcomes were largely based on subjective measurements (i.e. self-reporting of symptoms); blinding of study personnel may also be an important source of bias as they decide whether a patient’s medication step should be changed and there is some degree of interpretation in this decision. Finally, there was the possibility of selective reporting in Petsky et al. ,107 which may predispose the results to favour the intervention over the control.

Study details

Study design and timeline of studies

All seven studies were RCTs, with varying degrees of blinding (see previous section). Study timelines are presented in Table 36. Study duration ranged from 6 months103 to 12 months. 105,107–109 Pijnenburg et al. ,106 Szefler et al. ,104 Fritsch et al. 103 and Pike et al. 108 reported run-in periods of 2, 3, 4 and 4–16 weeks respectively. In Pijnenburg et al. 106 and Fritsch et al. ,103 details of the run-in period were not provided. In Szefler et al. 104 patients were provided with a treatment programme based on previous treatment, adherence and control and in Pike et al. 108 patients were stabilised when necessary. Verini et al. 105 and Petsky et al. 107 reported no run-in. In addition, the frequency of visits varied from study to study. Fritsch et al. 103 and Szefler et al. 104 reported visits every 6–8 weeks, Pike et al. 108 every 2 months, Pijnenburg et al. 106 and Peirsman et al. 109 every 3 months and Verini et al. 105 every 6 months. Petsky et al. 107 did not report the frequency of visits and provided outcomes for 12 months only.

TABLE 36 - Children management review: study timelines

ASS, Asthma Severity Score; NHLBI, National Heart, Lung, and Blood Institute; PD20, dose of methacholine causing a 20% fall in FEV₁ from baseline; QoL, quality of life.

Author, year	Timeline of study	Final assessment
Fritsch 2006103	Visit 1: 4-week run-in. Visit 2: randomisation. Visit 3: visits at 6, 12, 18 and 24 weeks; symptoms, SABA use, anti-inflammatory treatment, FeNO and spirometry recorded. Bronchial challenge test (4.5% hypertonic saline) was carried out between the first and second visit	24 weeks
Peirsman 2013109	Visit 1: baseline. Visits 2–5: visit every 3 months; assessed symptom-free days (assessed daily), exacerbations, unscheduled asthma-related visits, hospital or emergency admissions, non-attendance at school and need for caregiver to take time off; FeNO recorded in all participants	12 months
Petsky 2010107	Spirometry, FeNO, QoL and asthma/cough diary every visit	12 months
Pijnenburg 2005106	Visit 0: 2-week run-in. Visit 1: randomisation, FeNO, FEV1, PD20, diary card. Visits 2–4: visit every 3 months; FeNO and symptoms (diary card previous 2 weeks) recorded at each visit. Visit 5: FeNO, FEV1, PD20, diary card	9 months
Pike 2013108	Unclear when baseline measurements performed. Visit 0: start of run-in period (4–16 weeks). Visit 1: randomisation. Visits 2–7: visit every 2 months for assessment – FeNO, exacerbations, symptoms and reliever use over previous 2 months recorded, treatment adherence assessed	12 months
Szefler 2008104	Visit 1: assessed asthma symptoms, pulmonary function, skin test, sensitivity, adherence and level of asthma control (NHLBI guidelines134). Run-in period of 3 weeks on a regimen based on standard treatment – physicians selected a treatment programme (one of six) based on previous treatment, adherence and asthma control. Included 10-minute session about adherence. Adherence measured during run-in period with Diskus inhaler and questionnaire. Centralised block randomisation and visit every 6–8 weeks for 46 weeks. Each visit: FeNO, days of symptoms, use of rescue drugs, pulmonary function, use of health care, adherence to treatment, missed days of school from asthma. Data were entered into computer and treatment option computed based on random allocation. In total, 546 were then randomly assigned to 46 weeks of either standard treatment or standard treatment modified on the basis of measurements of FeNO	46 weeks
Verini 2010105	Visit 1: baseline. Visit 2: 6 months. Visit 3: 12 months. ASS, asthma exacerbation frequency, asthma therapy score and immunoallergological and functional data recorded at each visit	12 months

Population

The study and population characteristics are provided in Table 37. Eligibility criteria varied from study to study. With the exception of Verini et al. ,105 all studies included children with confirmed or persistent asthma. It is difficult to determine whether severity was comparable in the studies as scores have not been reported in a way that allowed comparison. Fritsch et al. 103 did not report severity; Szefler et al. 104 reported ACT scores; Verini et al. 105 and Peirsman et al. 109 classified participants on the basis of GINA scores; Pijnenburg et al. 106 reported mean daily symptom scores; and Pike reported exacerbations and OCS use in the previous year. Some insight into severity can be gained from considering the inclusion criteria and setting of each study. Three studies appeared to recruit patients who were uncontrolled. 104,105,107 Verini et al. 105 and Petsky et al. 107 recruited patients attending a specialist clinic, perhaps suggesting difficult to control patients, although in Verini et al. 105 patients had not yet started ICS therapy; an alternative explanation would be that all patients are sent to a specialist clinic before starting ICS therapy in this region and the patients were therefore not necessarily uncontrolled. Szefler et al. 104 recruited only patients with evidence of persistent or uncontrolled disease. Studies recruiting patients who are difficult to control are likely to capture the efficacy of FeNO for increasing control but not for decreasing ICS use in patients who are well controlled. Pike et al. 108 aimed to recruit patients with moderate to severe asthma and specified that children must be receiving therapy equivalent to stage 4 in the BTS/SIGN guidelines,8 as well as a requirement to demonstrate responsiveness to a bronchodilator (i.e. an increase in FEV₁ of 15%). Fritsch et al. 103 recruited mild to moderate persistent asthma patients at an outpatients clinic. Pijnenburg et al. 106 recruited atopic asthma patients who were attending a children’s hospital, although it is not clear if this was a scheduled appointment, an emergency admission or just the location of the study follow-up and it is therefore unclear what level of severity of asthma these patients may have. As they had all had a stable dose of ICS for the previous 3 months, it may be reasonable to assume that these patients were reasonably well controlled. Peirsman et al. 109 recruited the broadest spectrum of patients, including mild to severe persistent asthmatics.

TABLE 37 - Child management review: study and population characteristics

C, control group; CI, confidence interval; G, geometric mean; I, intervention group; IgE, immunoglobulin E; ITT, intention to treat; NR, not reported; NS, not significant; PEF, peak expiratory flow.

Author, year	Study details	Inclusion/exclusion criteria	No. analysed/no. recruited	Age (years), sex	Spirometry, mean (SD)	Severity	FeNO (ppb)	Atopic, smokers	Medication use
Fritsch 2006103	Setting: outpatient clinic, Vienna, Austria Funding: technical and analytical support from Aerocrine Study design: RCT, single blind, parallel groups	Patients with mild to moderate persistent asthma with a positive skin prick test (SPT) or radioallergosorbent test (RAST > 1) to at least one of seven common aeroallergens (cat, dog, house dust mite, alternaria, birch pollen, hazelnut pollen and mixed grass pollen) in past medical history or at the time of recruitment	47/52 I: 22 analysed, no. recruited unclear; C: 25 analysed, no. recruited unclear	Children and adolescents (6–18 years) Mean (SD) age: I: 11.3 (3.4); C: 12.1 (2.8) Male 28/47 (59.6%)	FEV1%, median (IQR): I: 101 (91.1 to 107.5); C: 93.7 (83.8 to 99.6)	NR	Median (IQR): I: 34.6 (17.5 to 58.6); C: 31 (20.8 to 54.8); difference NS	Atopic: assume 100% Smokers: NR	ICS dose (µg/day), median (25%–75%): I: 230 (100 to 400); C: 140 (0 to 400) Beta-agonist puffs/day, median (IQR): I: 1 (0 to 7); C: 0 (0 to 2)
Peirsman 2013109	Setting: Belgium (seven sites) Funding: Investigator Initiated Studies Program of Merck & Co. Study design: multicentre, single-blind, parallel-group RCT	Mild to severe persistent asthma according to GINA guidelines125 for ≥ 6 months, allergic sensitisation (i.e. positive skin prick test and/or specific IgE antibodies against inhaled allergens), elevated FeNO values in the presence of allergic airway inflammation. Exclusion: significant comorbidity, acute exacerbation or administration of experimental medication 4 weeks before screening, hospitalisation and/or systemic corticosteroids in 12 weeks before screening or OCS dependence	93/99 I: 47/49; C: 46/50	Children and young adolescent (5–14 years) Mean (SD) age: I: 10.6 (2.2); C: 10.7 (2.1) Male 66/99 (66.6%)	FEV1% predicted: I: 92.9 (12.2); C: 89.0 (16.2); t-test: p = 0.18	Mild/moderate/severe, n (%): I: 14 (28.6)/32 (65.3)/2 (4.1) (one missing value); C: 14 (28)/33 (66)/3 (6); chi-square: p = 0.92	Median (IQR): I: 32.5 (13.8 to 72.0); C: 27.5 (16.3 to 47.3)	Atopic: assume 100% Smokers (household exposure): I: 5/48 (one missing value); C: 8/49 (one missing value)	Daily ICS dose (µg), median (IQR): I: 320 (160 to 400); C: 320 (145 to 400) LABA use (n/N): I: 25/49; C: 23/50
Petsky 2010107	Setting: Australia (clinical setting NR) Funding: non-industry Study design: RCT	Children with asthma and on ICS attending a paediatric specialist clinic. Excluded if other cardiorespiratory illness, if poorly compliant or if unable to take ICS or LABA	63/63 I: 31/31; C: 32/32	Children (min. 4 years, max. NR) Mean age NR Male NR	FEV1% predicted (95% CI): I baseline: 101.3 (90.9 to 111.7); I at 12 months: 106.1 (91.9 to 120.2); C baseline: 84.28 (63.9 to 104.7); C at 12 months: 91.16 (84.7 to 97.7)	NR	NR	Atopic: NR Smokers: NR	NR
Pijnenburg 2005106	Setting: children’s hospital, Rotterdam, the Netherlands Funding: some non-industry plus grant from Aerocrine to department Study design: parallel-group, double-blind RCT	Children with atopic asthma who had been using ICS at a constant dose for previous 3 months	85/89 I: 39/42; C: 46/47	Children and adolescents (6–18 years) Mean (SD) age: I: 11.9 (2.9); C: 12.6 (2.8) Male n = 55/85 (64.7%)	FEV1% predicted: I 96 (14); C 99 (20)	Mean (SD) daily symptom score: I: 1.4 (2.0); C: 2.0 (2.4)	GM (range): I: 26.4 (5.6–134.9); C: 29.8 (3.1–117.5)	Atopic: I: 39/39 (100%); C: 46/46 (100%) Smokers: NR	Mean (SD) daily beta2-agonist puffs: I: 0.4 (0.6); C: 0.4 (0.5) Initial ICS dose (µg), mean (SD): I: 762 (335); C: 746 (410)
Pike 2013108	Setting: multiple outpatient clinics in the UK (n = 4) Funding: NR; authors state no conflict of interest Study design: RCT (assessor blinded)	Age 6–17 years, treatment with 400 µg/day beclomethasone/budesonide or 200 µg/day fluticasone. Clinical asthma diagnosis based on symptoms including 15% increase in FEV1 with bronchodilator or diurnal PEF variability ≥ 15%. Exclusion criteria: inability to perform spirometry or FENO, cigarette smoking, poor treatment adherence, life-threatening exacerbation or need for maintenance oral prednisolone	90/90 I: 34/44 completers (all analysed by ITT); C: 43/46 completers (all analysed via ITT)	Children and adolescents (6–17 years) Mean (SD) age: I: 10.51 (2.6); C: 11.42 (2.69) Male 51/90 (56.7%)	NR	Median (IQR) exacerbations in past year: I: 3.5 (2 to 8); C: 4.5 (2 to7) Median (IQR) OCSs in past year: I: 1 (0 to 3.5); C: 2 (0 to 3)	Median (IQR): NR	Atopic: I: 81.1%; C: 88.4% Smokers (household exposure): I: 9.1%; C: 13.0%	ICS dose (µg/day), median (25%–75%): I: 750 (400 to 1000); C: 800 (400 to 1000) Beta-agonist puffs/day, median (IQR): NR
Szefler 2008104	Setting: 10 sites, USA Funding: non-industry, although one author received speaker fees from Aerocrine Study design: multicentre, double-blind, parallel-group RCT	Eligibility restricted to residents of urban census tracts in which at least 20% of households had incomes below the federal poverty threshold. Eligible participants had been diagnosed with asthma by a physician. Those on long-term control included only if they had persistent asthma or evidence of uncontrolled disease; all others must have both symptoms of persistent asthma and evidence of uncontrolled disease. Excluded if adherence < 25% or a smoker	534/546 I: 267/276; C: 267/270	Children and young adults (12–20 years) Mean (SD) age: 14.4 (2.1) Male 288/546 (53%)	FEV1 (proportion of best FEV1): I: 95.9 (15.5); C: 95.7 (15.9) FEV1/FVC: I: 79.8 (9.0); C: 80.4 (8.3)	ACT score in last month (range 5–25), mean (SD): I: 21.1 (3.6); C: 21.3 (3.2)	Median (25%–75%): at enrolment: 31.7 (14.1–65.4); at randomisation: I: 20.5 (11.5–45.3), C: 19.7 (10.9–38.0)	Atopic: NR Smokers: 0%	NR
Verini 2010105	Setting: secondary care hospital, Italy Funding: NR Study design: single-centre, parallel-group RCT	Patients referred for allergic asthma – not all already on ICS	64/64 I: 32/32; C: 32/32	Children and adolescents (6–17 years) Mean (SD) age: I: 10.7 (2.4); C: 11.3 (2.1) Male 36/64 (56.3)	NR	GINA score 0: I: 7; C: 7 GINA score 1: I: 18; C: 19 GINA score 2 or 3: I: 7; C: 6	Baseline, mean (SD): I: 13.78 (12.31); C: NR	Atopic: 64/64 (100%) Smokers: NR	NR

Interestingly, atopy is a known confounder to FeNO measurements as atopic subjects have raised FeNO levels regardless of asthma status. Asthmatic atopic patients are thought to have the highest levels overall and so FeNO is theoretically able to distinguish between controlled and uncontrolled atopic asthmatics as well as between controlled and uncontrolled non-atopic asthmatics. 138 However, atopic asthma patients tend to have ICS-responsive asthma more often than non-atopic patients and so studies recruiting atopic patients will have limited generalisability. It is unclear whether studies in atopic patients will over- or underestimate efficacy or have no impact at all, although clinical input to the assessment suggested that it would be expected to increase estimates of efficacy as atopy is correlated with ICS responsiveness. Atopic patients were recruited by Peirsman et al. ,109 Fritsch et al. ,103 Verini et al. 105 and Pijnenburg et al. 106 Pike et al. 108 recruited only patients who demonstrated responsiveness to bronchodilators, but not all were atopic.

Petsky et al. 107 was the only study to include young children only, with all other studies including adolescents and/or young adults as well as young children, or adolescents only in the case of Szefler et al. 104 The studies also varied in terms of size (range 52103–546104) and baseline FeNO was inconsistently reported.

Interventions

Table 38 describes the interventions used in each study. NIOX MINO was used in the two most recent studies;108,109 the NIOX device was used in three of the studies;103,104,106 Verini et al. 105 used the ECO MEDICS CLD 88 device; and it was unclear what device was used by Petsky et al. 107

TABLE 38 - Child management review: description of the intervention management strategies

b.i.d., twice per day; p.o., by mouth.

Author, year	Decisions based on flow rate, device and cut-off points	Step-up/step-down protocol	Doses
Fritsch 2006103	Based on FeNO readings only Flow rate; device: 50 ml/second with the single-breath online method; NIOX device Cut-off: FeNO > 20 ppb	In patients with stable asthma, an increased FeNO level was considered a sign of insufficient anti-inflammatory treatment (either because of insufficient dosing or because of low adherence to prescribed therapy); hence, aimed to improve adherence to therapy in patients on anti-inflammatory treatment with raised FeNO. These patients were provided with 2-week diary cards to record daily symptoms, beta-agonist use and controller medication requirements and telephone calls were made regularly to check adherence to therapy. Asymptomatic patients on therapy with beta-agonists on demand only, with normal lung function but increased FeNO levels, were prescribed low-dose steroids. Step up was performed irrespective of FeNO level if FEV1% predicted was < 80% and/or there were severe symptoms over the last 4 weeks and/or beta-agonist use involved six or more puffs over the last 14 days. If FeNO was raised in these patients they received 2-week diary cards as well. Step down was performed if FEV1% predicted was ≥ 80% and there were no or mild symptoms over the last 4 weeks and beta-agonist use involved fewer than six puffs over the last 14 days and FeNO was ≤ 20 ppb	Low-dose ICS: 2 × 100 µg fluticasone or 2 × 200 µg budesonide Low-dose ICS + LTRAs: 2 × 100 µg fluticasone or 2 × 200 µg budesonide + 5 mg montelukast once daily p.o. Low-dose ICS + LABAs: 2 × 100 µg fluticasone + 2 × 50 µg salmeterol or 2 × 200 µg budesonide + 2 × 12 µg formoterol High-dose ICS + LTRAs: 2 × 250 µg fluticasone or 2 × 400 µg budesonide + 5 mg montelukast once daily p.o. High-dose ICS + LABAs: 2 × 250 µg fluticasone + 2 × 50 µg salmeterol or 2 × 400 µg budesonide + 2 × 12 µg formoterol
Peirsman 2013109	Based on FeNO readings and symptoms Flow rate; device: 50 ml/second; NIOX MINO Cut-off: 20 ppb	Asthma classed as ‘controlled’ (≤ 20 ppb and symptoms controlled); ‘partly controlled’ (≤ 20 ppb and partly controlled symptoms); or uncontrolled (FeNO > 20 ppb regardless of symptoms). Medication changes were guided by participants’ baseline therapies	Controlled: if on ICS only: step down ICS 100 µg/day, if already < 100 µg stop and add LTRA; if on LTRA only, no change; if on ICS + LTRA, step down ICS 100 µg/day, if already < 100 µg, stop ICS; if on ICS + LABA, stop LABA Partly controlled: if on ICS only, consider adding LTRA; if on LTRA only, consider adding ICS 100 µg/day (max. 200 µg/day); if on ICS + LTRA, consider ICS step up by 100 µg/day (max. 400 µg/day, then add LABA); if on ICS + LABA, consider adding LTRA Uncontrolled: if on ICS only, add LTRA; if on LTRA only, add ICS 100 µg/day (max. 200 µg/day); if on ICS + LTRA, step up ICS by 100 µg/day (max. 400 µg, then add LABA); if on ICS + LABA, replace LABA with LTRA
Petsky 2010107	Based on FeNO and atopy Flow rate; device: NR Cut-offs: NR	NR (treatment adjusted according to exhaled nitric oxide result, monthly for 4 months, then every second month for 8 months)	NR
Pijnenburg 2005106	Based on FeNO and symptoms Flow rate; device: presume 50 ml/second; NIOX analyser Cut-offs: > 30 ppb; ≤ 30 ppb + symptoms > 14 ≤ 30 ppb + symptoms ≤ 14	FeNO > 30 ppb = ICS increased; FeNO ≤ 30 ppb and symptoms > 14 = ICS stays the same; FeNO ≤ 30 ppb and symptoms ≤ 14 = ICS decreased	ICS doses: 100 µg: increase to 200 µg, decrease to 0 µg 200 µg: increase to 400 µg, decrease to 100 µg 400 µg: increase to 800 µg, decrease to 200 µg 500 µg: increase to 1000 µg, decrease to 250 µg 800 µg: increase to 1200 µg, decrease to 400 µg 1000 µg: increase to 1500 µg, decrease to 500 µg 1200 µg: increase to 1600 µg, decrease to 800 µg 1600 µg: increase to 2000 µg, decrease to 1200 µg 2000 µg: no further increase, decrease to 1000 µg
Pike 2013108	Based on FeNO readings and symptoms Flow rate; device: flow rate NR; NIOX MINO Cut-offs: ≤ 15 ppb; > 15 ppb and ≤ 25 ppb; ≥ 25 or FeNO doubled from baseline	FeNO ≥ 25 ppb or more than twice baseline: poorly controlled: increase ICS or add LTRA if already at SIGN/BTS guidelines8 step 4; if after increasing by two SIGN/BTS steps FENO remains high do not increase therapy further asthma controlled or well controlled: increase ICS or add LTRA if already at SIGN/BTS guidelines8 step 4 FeNO > 15 ppb and ≤ 25 ppb poorly controlled: increase LABA therapy; if dose maximal, increase ICS or add LTRA if already at SIGN/BTS guidelines8 step 4 asthma controlled or well controlled: continue current treatment FeNO ≤ 15 ppb poorly controlled: increase LABA; if dose maximal, increase ICS or add LTRA if already at SIGN/BTS guidelines8 step 4 asthma controlled or well controlled: if asthma controlled for 3 months, reduce ICS; if dose ≤ 400 µg, reduce LABA	Step 1: no ICS Step 2: beclomethasone 50 µg b.i.d. via spacer or budesonide 50 µg b.i.d. via spacer or turbohaler or fluticasone 50 µg once a day via spacer (or accuhaler) Step 3: beclomethasone 100 µg b.i.d. via spacer or budesamide 100 µg b.i.d. via spacer or turbohaler or fluticasone 100 µg once a day via spacer (or accuhaler) Step 4: beclomethasone 200 µg b.i.d. via spacer or budesamide 200 µg b.i.d. via spacer or turbohaler or fluticasone 100 µg once a day via spacer (or accuhaler) Step 5: trial of LABA. If ineffective, consider trial of LTRA Step 6: fluticasone 125 µg b.i.d. via spacer Step 7: fluticasone 250 µg b.i.d. via spacer Step 8: consider short course of prednisolone or other therapeutic options
Szefler 2008104	Based on days/nights of symptoms (patient recall over past 2 weeks), FEV1 as percentage of personal best, FeNO Flow rate; device: 50 ml/second; NIOX device Cut-offs: 0–20, 20.1–30, 30.1–40, > 40 ppb	Step up/down based on predefined levels of control: Level 1 – days of symptoms in past 2 weeks 0–3; nights of symptoms in past 2 weeks 0–1; % of personal best FEV1 ≥ 80%; FeNO 0–20 ppb. Medication would not change at this level or if at level 1 for two consecutive visits it may be stepped down Level 2 – days of symptoms in past 2 weeks 4–9; nights of symptoms in past 2 weeks 2; % of personal best FEV1 ≥ 80%; FeNO 20.1–30 ppb. Increase medication by one step Level 3 – days of symptoms in past 2 weeks 10–13; nights of symptoms in past 2 weeks 3–4; % of personal best FEV1 70–79%; FeNO 30.1–40 ppb. Increase medication by two steps Level 4 – days of symptoms in past 2 weeks 14; nights of symptoms in past 2 weeks 5–14; % of personal best FEV1 < 70%; FeNO > 40 ppb. Increase medication by three steps or two steps + prednisone	Step 0 – no controller medication; rescue treatment with salbutamol as needed Step 1 – fluticasone by dry powder inhaler 100 µg/day Step 2 – fluticasone by dry powder inhaler 100 µg b.i.d. Step 3 – fluticasone by dry powder inhaler 100 µg/day and salmeterol 50 µg b.i.d. Step 4 – fluticasone by dry powder inhaler 250 µg/day and salmeterol 50 µg b.i.d. Step 5 – fluticasone by dry powder inhaler 500 µg/day and salmeterol 50 µg b.i.d. Step 6 – fluticasone by dry powder inhaler 500 µg/day and salmeterol 50 µg b.i.d. plus either low-dose theophylline or montelukast every day
Verini 2010105	Based on GINA guidelines136 plus FeNO values Flow rate; device: flow rate NR; CLD 88 Cut-off: 12 ppb	Values > 12 ppb lead to increased medication. Values < 12 ppb lead to a reduction in or maintenance of amount of drugs. Changes in drugs not reported. Unclear whether FeNO used at visit 2 only to guide therapy	NR

None of the studies used the same protocol or cut-off points for the management of asthma with FeNO. The protocol of Fritsch et al. 103 was based on FeNO readings only. Other studies used a combination of FeNO levels and symptoms, with various cut-off points and numbers of cut-off points: Szefler et al. 104 specified three levels of cut-off, Pike et al. 108 specified two cut-off points, Petsky et al. 107 did not report cut-offs and all other studies used just one cut-off point. These cut-off points ranged from 12 ppb to 40 ppb. Treatments indicated at each step were also highly heterogeneous across studies, with Pike et al. ,108 Peirsman et al. ,109 Fritsch et al. 103 and Szefler et al. 104 indicating doses for ICSs, LTRAs and LABAs and Pijnenburg et al. 106 indicating doses only for ICSs. Verini et al. 105 and Petsky et al. 107 did not report doses. The treatment in the study by Pike et al. 108 was most similar to that in the BTS/SIGN guidelines,8 with some minor modifications. Importantly, in the studies by Szefler et al. ,104 Pike et al. 108 and Peirsman et al. ,109 step down could not occur on the basis of low FeNO levels if patients were still experiencing symptoms. This may have limited any potential reduction in ICS use for patients who were non-ICS responsive.

Control

Table 39 provides details of the control group interventions. As with the interventions, none of the studies used the same criteria, protocols or treatment doses for the management of asthma in the control arm of the study, but, when reported, they all used the same treatment steps in both arms of the study. Management was typically guided by symptom severity and/or FEV₁. Verini et al. 105 and Peirsman et al. 109 used GINA guidelines136 as a control and Pike et al. 108 used BTS/SIGN guidelines8 with some minor modifications.

TABLE 39 - Child management review: description of the control group management strategies

b.i.d., twice per day; NR, not reported.

Author, year	Decisions based on	Step-up/step-down protocol	Doses
Fritsch 2006103	Asthma control (symptoms, SABA use, lung function), as recommended in current (German) asthma guidelines139	A step down in therapy was performed if FEV1% predicted was ≥ 80% and there were no or mild symptoms over the last 4 weeks and beta-agonist use involved fewer than six puffs over the last 14 days. A step up was performed in every other case	As intervention
Peirsman 2013109	Symptoms, need for rescue treatment in past 2 weeks, spirometry (FEV1) based on GINA guidelines136	GINA guidelines136 (specific step-up/step-down protocol not described) to determine if controlled, partly controlled or uncontrolled	As intervention
Petsky 2010107	Symptoms/FEV1	Unclear	NR
Pijnenburg 2005106	Symptoms	Symptom score > 14 = ICS increase; symptom score ≤ 14 for first time = ICS stays the same; symptom score ≤ 14 for second time = ICS decrease Symptom score calculated as mean of daily scores for dyspnoea, wheezing and cough, during daytime and night-time, with each scored from 0 to 3, giving a max. score of 18, as well as use of beta2-agonists and percentage of symptom-free days. Calculated over previous 2 weeks for monitoring and over previous 4 weeks for end-point evaluation	As intervention
Pike 2013108	Symptom control	Therapy was stepped up if symptoms were poorly controlled and decreased if well controlled for ≥ 3 months, according to BTS/SIGN guidelines8	As intervention
Szefler 2008104	National Asthma Education and Prevention Program (symptoms, treatment use, lung function)140	Control group received standard treatment based on the guidelines of the National Asthma Education and Prevention Program140 (i.e. as intervention but without FeNO measurements)	As intervention
Verini 2010105	GINA guidelines136 (symptoms, SABA use, lung function)	GINA guidelines136 (specific step-up/step-down protocol not described)	NR

Estimates of efficacy

Exacerbations

All seven studies provided some data on asthma exacerbations, although it was unclear in some cases what the precise definition of an exacerbation was (Table 40).

TABLE 40 - Child management review: exacerbation and OCS use rates in children and adolescents with or without FeNO-guided management

NS, not significant.

p-value for number of patients having one or more events, not number of events.

Unable to calculate rate per person-year as unclear how many patients were included in the analysis.

Unable to calculate rate per person-year because of lack of data on absolute number of exacerbations per group.

Author, year	Definition of outcomes	Intervention	Control	Between-group comparison
Hospital admission
Peirsman 2013109	One or more hospital admission	1/43 (2.3%)	1/43 (2.3%)	Chi-square test: p = 1.00
Pike 2013108	Requiring ≥ 8 hours of admission	5 patients (11.4%)	3 patients (6.5%)	p = 0.420
Szefler 2008104	One or more hospital admissions	3.3% (SD 1.78%)	4.1% (SD 1.98%)	Mean difference –0.8 (95% CI –4.0 to 2.3), p = 0.61
Unscheduled use of health care
Peirsman 2013109	One or more unscheduled uses of health care	6/44 (13.6%)	15/43 (34.9%)	Chi-square test: p = 0.02
Szefler 2008104	One or more unscheduled uses of health care	21.3% (SD 4.09%)	22.7% (SD 4.19%)	Mean difference –1.4 (95% CI –9.3 to 6.7), p = 0.74
OCS use
Fritsch 2006103	OCS use (no. of patients/group)	n = 2	n = 2	p = NS
Pijnenburg 2005106	Prednisone courses	Eight events in 1 year = 8/39 = 0.21 per patient	18 events in 1 year = 18/46 = 0.39 per patient	p = 0.60a
Szefler 2008104	Prednisone courses	Mean 0.66 (SE 0.085)	Mean 0.84 (SE 0.085)	Mean difference 0.17 (95% CI –0.08 to 0.41), p = 0.14
Any/wide definition of exacerbation
Fritsch 2006103	Exacerbation defined as oral steroid courses because of asthma symptoms over the last 4 weeks and/or unscheduled visit because of asthma symptoms over the last 4 weeks and/or increase in asthma symptoms from a symptom score of 0 or 1 to a score of 2 and/or decline in FEV1 > 10% compared with last visit (no. of patients/group)	17/88 observations (18.2%) [sic]b	22/99 observations (21.2%) [sic]b	p = NS
Peirsman 2013109	Exacerbation as per GINA guidelines:136 an episode of progressive increased shortness of breath, coughing, wheezing or chest tightness or a combination of these symptoms	18 per yearb	35 per yearb	Mann–Whitney test: p = 0.02
Petsky 2010107	Asthma exacerbations (severity not described)	6/31 (19.4%)	15/32 (46.9%)	p = 0.021a
Pike 2013108	Patients experiencing an exacerbation	37 patients (84.1%)c	38 patients (82.6%)c	p = 0.850
Szefler 2008104	Exacerbation defined as a composite outcome consisting of admissions to hospital, unscheduled visits and prednisone use	37.0% (SD 4.83%)	43.6% (SD 4.96%)	Mean difference –6.5% (95% CI –14.4 to 1.4), p = 0.11
Verini 2010105	ATS 2005 definition:35 number of episodes of coughing, dyspnoea and wheezing requiring SABAs	0.83 (SD 0.98) per person-year	1.85 (SD 1.34) per person-year	Between-group comparison NR
Other outcomes
Peirsman 2013109	Number of children with one or more exacerbation	11/46 (23.9%)	22/46 (47.8%)	Chi-square test: p = 0.02
Peirsman 2013109	One or more emergency room admission	2/45 (4.4%)	4/46 (8.7%)	Chi-square test: p = 0.41
Szefler 2008104	One or more prednisone course	32.1% (SD 4.67%)	42.0% (SD 4.94%)	Mean difference –10.3% (95% CI –18.5 to –2.2), p = 0.01

Hospitalisations

Three studies104,108,109 reported the number of patients (but not the rate per person-year) requiring hospitalisations. All three studies reported no difference between groups (see Table 40).

Exacerbations resulting in oral corticosteroid use

Data on exacerbations resulting in OCS use were reported in three studies. 103,104,106 Szefler et al. 104 reported rates per year and Pijnenburg et al. 106 reported the number of courses per group. In both cases rates were lower in the intervention arm. In the study by Szefler et al. 104 the rate per year was 0.66 (SE 0.085) in the FeNO group and 0.84 (SE 0.085) in the control group. It is not clear if this analysis calculated rates per person-year to account for missing data points. The mean difference was not statistically significant (0.17, 95% CI −0.08 to 0.41; p = 0.14). In the study by Pijnenburg et al. 106 the reviewer-calculated mean number of exacerbations per person was 0.21 (eight courses in 39 patients) in the FeNO group and 0.39 (18 courses in 46 patients) in the control group. The difference in the number of people experiencing an exacerbation in the study by Pijnenburg et al. 106 was non-significant. Fritsch et al. 103 reported no significant difference in the number of people requiring OCSs between groups.

Any/wide definition of exacerbation

Fritsch et al. ,103 Szefler et al. ,104 Verini et al. ,105 Petsky et al. ,107 Pike et al. 108 and Peirsman et al. 109 all reported this outcome using a broad definition of exacerbation (sometimes called treatment failure), but used different definitions to one another. Pijnenburg et al. 106 did not report this outcome. Lack of data allowing calculation of rates per person-year precluded meta analysis.

Five studies reported numerically fewer exacerbations or treatment failures in the intervention arm, but these differences were statistically significant in only two studies. Fritsch et al. 103 used a composite outcome that appeared to include moderate to severe exacerbations. There were 17 exacerbations out of 88 observations in the intervention group (reported as 18.2% in Fritsch et al. 103) compared with 22 exacerbations out of 99 observations in the control group (reported as 21.2% in Fritsch et al. 103). These data were not convertible to rates as some data points were missing in the study. The difference was not significant at the p < 0.05 level. Szefler et al. 104 also used a composite outcome, which appeared also to relate to moderate to severe exacerbations. This study reported the percentage of patients in each group with more than one exacerbation; these were 37.0% in the intervention group and 43.6% in the control group (risk ratio −6.5, 95% CI −14.4 to 1.4; p = 0.11). Verini et al. 105 reported events leading to the use of SABAs, which appeared to be likely to incorporate minor to major exacerbations. Rates per person-year were 0.83 (SD 0.98) and 1.85 (SD 1.34) in the intervention arm and control arm respectively. No between-group comparisons were presented for this outcome. Pike et al. 108 reported the number of patients who had ≥ 48 hours of increased asthma symptoms or therapy and showed no difference between the groups, with 37 (84.1%) patients in the intervention arm and 38 (82.6%) in the control arm experiencing an exacerbation (p = 0.850).

The two studies that reported a significant between-group difference were those by Petsky et al. 107 and Peirsman et al. 109 In the study by Petsky et al. ,107 exacerbations were not clearly defined but occurred in six out of 31 participants in the intervention group (19.4%) and 15 out of 32 participants in the control group (46.9%; p = 0.021). Peirsman et al. 109 reported statistically significantly fewer exacerbations of any severity (as defined using GINA guidelines136) in the intervention arm (18 events) than in the control arm (35 events; p = 0.02), although rates were not calculable because of missing information about the number of participants included in the analysis.

Inhaled corticosteroid use

Table 41 provides details of ICS use in each study. Fritsch et al. 103 and Szefler et al. 104 reported statistically significantly higher ICS use in the intervention group, Pijnenburg et al. 106 reported very similar levels and the values in the remaining study105 (in terms of absolute numbers using ICSs) were difficult to interpret. Fritsch et al. 103 reported median (IQR) end-point values for ICS use in the intervention and control groups as 316 (200 to 500) µg and 241 (26 to 607) µg respectively (β = 0.20, p < 0.01). Pijnenburg et al. 106 reported the mean (standard error of the mean) cumulative ICS dose from visit 1 to visit 5 as 4407 (367) µg in the intervention group and 4332 (383) µg in the control group (p = 0.73). Szefler et al. 104 reported the between-group difference in use of fluticasone, which was 119 µg/day greater in the FeNO group by the final visit (95% CI 49 µg to 189 µg; p = 0.001). Finally, Verini et al. 105 reported the absolute number of participants using ICSs from each group at each time point (intervention group: T1 20, T2 19, T3 19; control group: T1 15, T2 22, T3 19). However, the baseline values for the groups in this study were not comparable and the absolute numbers of participants using ICSs, without concomitant data on dosages used, provide little understanding of between-group ICS use.

TABLE 41 - Child management review: ICS use

CI, confidence interval; NR, not reported; T, time.

Author, year	ICS type and measurement definition	Intervention	Control	Between-group difference
Fritsch 2006103	Fluticasone and budesonide permitted. Data reported as median ICS dose (µg/day); unclear which ICS type the doses refer to	Baseline median (IQR) dose: 230 (100 to 400) µg; end point median (IQR) dose: 316 (200 to 500) µg	Baseline median (IQR) dose: 140 (0 to 400) µg; end point median (IQR) dose: 241 (26 to 607) µg	Repeated measures analysis: β = 10.20, p < 0.01
Peirsman 2013109	Budesonide or equivalent:
	Median (IQR) cumulative ICS dose (calculated by summing daily ICS dose from visit 1 to visit 5)	Cumulative dose: 1280 (800 to 1800) µg	Cumulative dose: 1200 (675 to 1600) µg	p = NS
	Median (IQR) change in daily ICS dose from baseline	+100 (0 to + 400) µg	0 (–200 to + 80) µg	p = 0.016
Petsky 2010107	NR	NR	NR	NR
Pijnenburg 2005106	Budesonide (2 mg max. daily dose permitted). Cumulative steroid dose (sum of mean daily steroid doses from visit 1 to visit 5)	Cumulative end point: 4407 (367) µg	Cumulative end point: 4332 (383) µg	p = 0.73
Pike 2013108	Beclomethasone, fluticasone and budesonide permitted. Data reported as median (IQR) ICS dose (µg/day) in terms of baseline, end point, change and total dose. Unclear which ICS type the doses refer to	Baseline dose: 750 (400 to 1000) µg/day; end point dose: 800 (400 to 1000) µg/day; dose change: 0 (–200 to 300) µg/day	Baseline dose: 800 (400 to 1000) µg/day; end point dose: 500 (400 to 1000) µg/day; dose change: 0 (–300 to 0) µg/day	Mann–Whitney rank-sum tests: baseline dose: p = 0.629; end point dose: p = 0.543; dose change: p = 0.297
Szefler 2008104	Fluticasone	NR	NR	Difference 119 µg/day (95% CI 49 to 189 µg; p = 0.001) (higher in intervention group)
Verini 2010105	Unclear what ICS used. Measured in terms of absolute number of patients using per group at each time point	T1: 20, T2: 19, T3: 19	T1: 15, T2: 22, T3: 19	NR

When studies recruiting patients who are or who are likely to be difficult to control103,104 were compared with the study that recruited patients who had been on a stable dose of ICS for 3 months,106 it was seen that the two studies recruiting the difficult to control groups saw an increase in ICS usage whereas the study that recruited stable patients saw similar levels of ICS use across both arms. This would be expected as the difficult to control group of patients is unlikely to need a dose reduction whereas patients who are stable may be eligible for such a reduction. In addition to this, the Szefler et al. 104 protocol did not allow step-down of ICSs on the basis of low FeNO levels if symptoms were still present, making step-down less likely to occur.

The two studies identified from the search update reported results that agree with previous findings. The study by Pike et al. ,108 which recruited more severe patients, saw a higher final median (IQR) ICS usage [800 (400 to 1000) µg vs. 500 (400 to 1000) µg] and higher upper and lower ends of the CIs [0 (–200 to 300) vs. 0 (–300 to 0) µg] for the median change from baseline in the intervention group, indicating that there was an increase in ICS use. However, the differences were not statistically significant. The study by Peirsman et al. ,109 which recruited a wider spectrum of patients but which also included severe asthmatics, also saw numerically higher doses in the intervention arm but, again, not statistically significantly so.

Health-related quality of life

Table 42 provides HRQoL data. Only Petsky et al. 107 provided data on HRQoL and it was unclear which quality of life tool was used. In the intervention group, the baseline mean was 84.38 (95% CI 77.27 to 91.48), which rose to 86 (95% CI 74.84 to 97.1) at 12 months. Conversely, in the control group, the baseline mean of 86 (95% CI 81.49 to 90.51) dropped to 83.75 at 12 months (95% CI 78.6 to 88.9). If quality of life was measured with the European Quality of Life-5 Dimensions (EQ-5D), which seems likely, then higher values would indicate better quality of life and thus FeNO would be favoured. The end-point difference was statistically significant (p = 0.042), although it was unclear whether this comparison was for the change or for absolute end values.

TABLE 42 - Child management review: other outcomes

CI, confidence interval; NS, not statistically significant; T, time.

Author, year	Outcome	Intervention, mean (SD)	Control, mean (SD)	Comparison
Asthma control
Fritsch 2006103	Increase in symptoms to a score of 2 (severe), n/N (%)	10/88 (11.4)	11/99 (11.1)	NS at p < 0.05
	Unscheduled visits, n/N (%)	5/88 (5.7)	5/99 (5.1)	NS at p < 0.05
	FEV1 decline > 10%, n/N (%)	7/88 (8.0)	13/99 (13.1)	NS at p < 0.05
Pijnenburg 2005106	Change in mean symptom severity scores between visit 1 and visit 5	0.1	0.6	Mean daily scores change p = 0.40
Szefler 2008104	Maximum days with symptoms	1.93 (2.60)	1.89 (2.69)	Mean difference 0.04 (95% CI −0.22 to 0.29), p = 0.78
	Days of wheeze	1.71 (2.52)	1.69 (2.64)	Mean difference 0.03 (95% CI −0.21 to 0.26), p = 0.83
	Days of interference with activities	0.87 (1.79)	0.95 (1.98)	Mean difference −0.08 (95% CI −0.26 to 0.10), p = 0.38
	Nights of sleep disruption	0.52 (1.30)	0.50 (1.25)	Mean difference 0.03 (95% CI −0.11 to 0.16), p = 0.71
	Days of school missed	0.19 (0.79)	0.23 (0.84)	Mean difference −0.04 (95% CI −0.12 to 0.05), p = 0.38
	ACT score in the last month	21.89 (2.83)	21.83 (2.88)	Mean difference 0.06 (95% CI −0.28 to 0.40), p = 0.72
Verini 2010105	Symptom score (mean for ordinal data: intermittent asthma = 1; mild/moderate persistent asthma = 2; severe persistent asthma = 3)	T1: 1.09 (0.81); T2: 0.56 (0.75); T3: 0.75 (0.95)	T1: 1.09 (0.77); T2: 0.93 (0.61); T3: 0.92 (0.82)	NR
HRQoL
Petsky 2010107	HRQoL (metric not specified), mean (95% CI)	Baseline: 84.38 (77.27 to 91.48); 12 months: 86 (74.84 to 97.1)	Baseline: 86 (81.49 to 90.51); 12 months: 83.75 (78.6 to 88.9)	NR
Other medication use
Fritsch 2006103	Median (IQR) beta-agonist puffs/day	1 (0 to 7)	0 (0 to 2)	NR
	Montelukast (LTRA) (unclear if median or mean)	1.26 mg/day	0 mg/day	p < 0.01
Peirsman 2013109	Median (IQR) percentage of symptom-free days	83.7 (27.1 to 91.9)	79.6 (51.7 to 94)	Mann–Whitney test: p = 0.58
	Children who missed school, n/N, (%)	10/46 (21.7)	12/46 (26.1)	Chi-square test: p = 0.63
	Children whose caregivers had to take time off, n/N, (%)	6/45 (13.3)	8/46 (17.4)	Chi-square test: p = 0.59
	Nights of sleep disruption	0.52 (1.30)	0.50 (1.25)	0.03 (95% CI −0.11 to 0.16), p = 0.71
	Days of school missed	0.19 (0.79)	0.23 (0.84)	−0.04 (95% CI −0.12 to 0.05), p = 0.38
	ACT score in the last month	21.89 (2.83)	21.83 (2.88)	0.06 (95% CI −0.28 to 0.40), p = 0.72
Petsky 2010107	–	–	–	–
Pijnenburg 2005106	Beta-agonist use	NR	NR	p = 0.28
Pike 2013108	Change in FeNO (ppb), mean (95% CI)	+3.1 (–5.5 to 11.6)	+ 3.3 (–8.5 to 15.1)	NS at p < 0.05
Szefler 2008104	Salmeterol (mean difference from baseline), µg/day	–6.5	–12	NR
Verini 2010105	Mean of ordinal data, in which antihistamines, ketotifen and chromones = 1; specific immunotherapy, LABAs and LTRAs = 2; ICS = 3	T1: 1.5 (0.7); T2: 1.43 (0.7); T3: 1.53 (0.6)	T1: 1.03 (0.9); T2: 1.62 (0.6); T3: 1.4 (0.7)	NR
	Number of patients using LTRAs	T1: 8; T2: 8; T3: 11	T1: 3; T2: 8; T3: 7	NR
	Number of patients using no anti-inflammatory drugs	T1: 4; T2: 5; T3: 2	T1: 14; T2: 2; T3: 6	NR

Asthma control and other medication use

Table 42 provides details of outcomes relating to asthma control and medication use. Four studies provided some data on asthma control, none of which demonstrated any significant effects favouring either the intervention or the control, although in the study by Verini et al. 105 significance was not reported. Furthermore, there was lack of uniformity in how asthma control was measured. Fritsch et al. 103 recorded the absolute number of participants per group whose symptom severity score increased to 2 (i.e. severe symptoms). Ten participants in the intervention group (11.4%), and 11 in the control group (11.1%) fulfilled this criterion (difference not significant). These researchers also reported the absolute number of participants per group who experienced a decline in FEV₁ of > 10%: seven in the intervention group (8%) and 13 in the control group (13.1%; p = not significant). Szefler et al. 104 presented between-group differences for a number of symptomatic indicators of control, with higher numbers favouring the intervention. The measure comprised days of wheeze (risk ratio 0.04, 95% CI −0.22 to 0.29; p = 0.78), days of interference with activities (risk ratio 0.03, 95% CI −0.21 to 0.26; p = 0.83), nights of sleep disruption (risk ratio −0.08, 95% CI −0.26 to 0.10; p = 0.38), days of school missed (risk ratio 0.03, 95% CI −0.11 to 0.16; p = 0.71) and ACT score in the last month (risk ratio −0.04, 955 CI −0.12 to 0.05; p = 0.38). Pijnenburg et al. 106 calculated symptom scores based on diary card data for dyspnoea, wheezing and cough. Day and night were scored separately and each symptom was scored between 0 and 3, giving a maximum possible total score of 18. The change in mean daily scores between baseline and visit 5 was 0.1 in the intervention group and 0.6 in the control group (between-group difference p = 0.4). Finally, Verini et al. 105 measured symptom control using the GINA scale,136 which classified participants as having remission asthma (GINA score 0), intermittent asthma (GINA score 1) or persistent asthma (GINA score 2 or 3). The means (SDs) for these categorical data were presented for both groups at all three time points: at visit 1 the values were 1.09 (0.81) in the intervention group compared with 1.09 (0.77) in the control group; at visit 2 they were 0.56 (0.75) and 0.93 (0.61) respectively; and at visit 3 they were 0.75 (0.96) and 0.92 (0.82) respectively. No intergroup comparisons were conducted, although means in the intervention arm are numerically lower than those in the control arm.

With respect to additional medication use, three studies provided data using different metrics and mostly without formal comparison statistics. Szefler et al. 104 reported the mean difference from baseline in salmeterol usage; This was –6.5 µg/day in the intervention group and –12 µg/day in the control group (p-value not reported). Verini et al. 105 created a categorical measure of medication use in which antihistamines (e.g. ketotifen) and chromones = 1; specific immunotherapy, LABAs and LTRAs = 2; and ICSs = 3. The means (SDs) of these data were presented for both groups. In the intervention group the values were 1.5 (0.7) (visit 1), 1.43 (0.7) (visit 2) and 1.53 (0.6) (visit 3). For comparison, the values for the control group at the same time points were 1.03 (0.9), 1.62 (0.6) and 1.4 (0.7) respectively. Verini et al. 105 also provided absolute numbers using LTRA. For the intervention group these were 8, 8 and 11 at visits 1, 2 and 3 respectively and for the control group these were three, eight and seven respectively. Fritsch et al. 103 reported the median (IQR) number of beta-agonist puffs/day. In the intervention group this was one (0 to 7) whereas in the control group the number was zero (0 to 2). Overall, two studies reported numerically higher additional medication use in the intervention arm104,105

Studies found during the search update108,109 agreed with previous observations as neither reported a statistically significant difference in metrics of asthma control.

Adverse events, mortality, compliance and test failure rates

Data on adverse events, mortality and compliance were reported only in Szefler et al. 104 No statistically significant differences between the groups were reported for any adverse events or for mortality. For the intervention and control groups, respectively, adverse events were reported for eyes, ears, nose and throat (8.3% vs. 8.1%; p = 0.87), gastrointestinal disorders (13.4% vs. 14.1%; p = 0.78), haematology disorders (27.2% vs. 28.9%; p = 0.44), infections (55.8% vs. 52.2%; p = 0.46), musculoskeletal symptoms (15.9% vs. 18.5%; p = 0.44), nervous system disorders (34.4% vs. 33.7%; p = 0.20), respiratory signs and symptoms (33.7% vs. 34.1; p = 0.92) and skin symptoms (15.6% vs. 17.8%; p = 0.18). Medication compliance was shown to be 86% in the intervention group and 92% in the control group. No mortality was observed in either group.