Nocturnal temperature-controlled laminar airflow device for adults with severe allergic asthma: the LASER RCT

Epidemiology

Asthma affects > 5.4 million people in the UK, with nearly 500,000 experiencing severe symptoms and frequent exacerbations that are inadequately controlled with available treatments. 1,2 The burden of severe asthma on the NHS is enormous, accounting for 80% of the total asthma cost (£1B3), with frequent exacerbations and expensive medications generating much of this cost. 4 In 2009, there were 1131 deaths caused by asthma,5 with those whose asthma remains poorly controlled facing the greatest risk. 6,7 Patients with severe asthma bear the greatest burden of asthma morbidity. They experience more frequent and severe exacerbations,8 which reduce their quality of life, impair their ability to work and place an enormous burden of anxiety on them and their families. 9 There is also an increased risk of significant depression. 10 One in five asthmatics in the UK report serious concerns that their next asthma attack will be fatal. 1 As highlighted in the 2010 Asthma UK report Fighting for Breath,11 these patients also face discrimination from employers, health-care professionals and society as a whole as a result of their asthma.

The unmet need in severe asthma

Current treatments, including oral corticosteroids (OCSs), ‘steroid-sparing’ immunosuppressants and monoclonal antibody therapies, often have limited efficacy and potentially serious side effects (steroids, immunosuppressive agents) or are prohibitively expensive (monoclonal antibodies). The adverse effects of long-term oral steroids include adrenal suppression, decreased bone mineral density, diabetes mellitus and increased cardiovascular mortality. 12 The anti-immunoglobulin E (IgE) treatment omalizumab (Xolair^®; Novartis Pharmaceuticals UK Ltd, Frimley, UK) has been shown to reduce exacerbations by up to 50%13 and improve quality of life in severe allergic asthma, but costs up to £26,640 per year. 14 The National Institute for Health and Care Excellence (NICE)14 reappraised the use of omalizumab in 2012 and, although recognising the grave effects of severe uncontrolled asthma on quality of life for patients and their families, concluded that it is cost-effective within the NHS only when its use is limited to those with severe persistent confirmed allergic IgE-mediated asthma experiencing four or more severe exacerbations in the preceding 12 months. A large number of patients are therefore left with a significant unmet clinical need and a specific requirement for therapies that reduce systemic steroid exposure.

It is also important to acknowledge the often unrecognised role that carers play in looking after patients with poorly controlled severe asthma. The Fighting for Breath11 document highlighted the strain that this can place on the physical and mental health of this vital informal workforce. In addition, having to take time off work, having to work part-time or not being able to work at all as a result of a patient’s care needs places a significant financial burden on carers.

National/international strategies to improve asthma care

The 2011 Department of Health and Social Care report An Outcomes Strategy for COPD and Asthma in England15 recognises the huge burden that poorly controlled asthma places on people’s lives and the NHS. It also describes the political commitment to improve asthma control and reduce asthma-related emergency health-care needs and deaths. 15 The 2011 British Thoracic Society (BTS) and Scottish Intercollegiate Guidelines Network (SIGN) national asthma guideline16 and the 2010 World Health Organization (WHO) consultation on severe asthma8 have highlighted an urgent need for research in severe asthma, acknowledging the limitations of available treatments and the dearth of clinical trials on which to base management recommendations. In its research strategy for 2016–21, Asthma UK17 identified the development of new treatments as a priority for improving clinical outcomes and patient well-being and reducing the cost of treating severe asthma within the NHS. It also identified the need to gain a better understanding of the impact of exposure to substances known to trigger asthma and the impact of strategies that regulate and control this exposure as a key priority.

Rationale for temperature-controlled laminar airflow therapy

At night, airborne particles are carried by a persistent convection current established by the warm body, transporting allergens from the bedding area to the breathing zone. 30 Proof-of-concept studies have shown that the temperature-controlled laminar airflow (TLA) device reduces the total number of airborne particles of > 0.5 µm in the breathing zone by 3000-fold (p < 0.001) and cat allergen exposure by 30-fold (p = 0.043) and significantly reduces the increase in the number of particles generated when turning in bed for all particle sizes. 31 Compared with a best-in-class traditional air cleaner, TLA is able to reduce exposure to potential allergens by a further 99%. 32 We postulate that this highly significant reduction in nocturnal exposure, targeted to the breathing zone, explains why TLA may succeed when so many other measures, including traditional air cleaners, have failed.

Evidence of benefit with temperature-controlled laminar airflow therapy

Compared with placebo, the TLA device has proven efficacy for asthma-related quality of life and bronchial inflammation (measured by exhaled nitric oxide) in a pan-European, multicentre, 12-month Phase III study33 (n = 282, age range 7–70 years). The greatest benefit was seen in the more severe asthma patients requiring higher intensity treatment [Global INitiative for Asthma (GINA) step 4] and with poorly controlled asthma [7-point Asthma Control Questionnaire (ACQ) score of < 19]. GINA step 4 is consistent with American Thoracic Society (ATS)/European Respiratory Society (ERS) severe asthma guideline definitions34 and BTS/SIGN guideline treatment step 4 [inhaled corticosteroid (ICS) dose of ≥ 1000 µg/day beclomethasone dipropionate (BDP) equivalent plus an additional controller medication such as a long-acting β₂-agonist, leukotriene receptor antagonist or sustained-release theophylline]. Although not powered to ascertain an effect on exacerbations, a post hoc analysis of the Phase III study data showed a decreased exacerbation rate in more severe patients treated with TLA compared with placebo, with a trend towards significance (mean TLA 0.23 exacerbations/year, placebo 0.57 exacerbations/year; p = 0.07). 33 Although a cost-effectiveness analysis based on the results from this trial found no significant differences in emergency department (ED) visits, hospitalisation days, medication usage and, therefore, overall costs between the two study groups,35 this may have reflected the fact that the trial was powered to detect a difference in asthma-related quality of life and did not specifically include patients at risk of exacerbations (average annual rate was 0.2 exacerbations per year), a predictor of increased asthma health-care resource use and costs. 36 Despite the lack of a significant reduction in health-care resource use and associated costs, subsequent economic modelling showed that TLA would be cost-effective in Sweden at the current monthly rental price (SEK2000, ≈£167), mainly because of increases in quality of life. 34

Using the results from a very small, 10 + 10-week randomised controlled crossover trial in Sweden,37 a modelling study addressed the potential cost-effectiveness of TLA therapy over a projected 5-year period. 38 Assuming no impact of TLA on health-care resource use, TLA was cost-effective compared with placebo at a device cost of €8200 (≈£6890), at a willingness-to-pay threshold of €35,000 (≈£30,000) per quality-adjusted life-year (QALY) gained. We considered that the greatest potential for cost-effectiveness could occur in severe asthma through reducing exacerbations, which was not addressed in the Swedish trial because of the short observation period and less severe patient population included.

A further pragmatic, patient-centred randomised controlled trial (RCT) of this novel non-pharmacological treatment in severe allergic asthma was thus warranted.

Inclusion criteria

Potential participants had to meet all of the following inclusion criteria by randomisation visit 2 to be considered eligible for the study:

• Adults (aged 16–75 years inclusive).

• A clinical diagnosis of asthma for ≥ 6 months supported by evidence of any one of the following –

  • Airflow variability with a mean diurnal peak expiratory flow (PEF) variability {calculated by [PEF(highest) – PEF(lowest)]/[PEF(mean)]} of > 15% during the baseline 2-week period or a variability in forced expiratory volume in 1 second (FEV₁) of > 20% across clinic visits within the preceding 12 months, with concomitant evidence of airflow obstruction (FEV₁/FVC ratio < 70%, where FVC is forced vital capacity).

  • Airway reversibility with an improvement in FEV₁ of ≥ 12% or 200 ml after inhalation of 400 µg of salbutamol (Ventolin^® GlaxoSmithKline UK) via a metered dose inhaler and spacer at the first study visit or within the preceding 12 months.

  • Airway hyper-responsiveness demonstrated by methacholine challenge testing, with a provocative concentration of methacholine (Nova Laboratories, Leicester, UK) causing at least a 20% reduction in FEV₁ (PC₂₀) of ≤ 8 mg/ml, or an equivalent test (see Table 29).

• Severe asthma –

  • Requirement for high-dose ICSs (≥ 1000 µg/day of BDP or equivalent; see Table 30) plus a second controller (long-acting β₂-agonist or antimuscarinic, theophylline or leukotriene antagonist) and/or systemic corticosteroids.

  • If on maintenance corticosteroids, the maintenance dose must have been stable for 3 months; this excluded any interim need for short-term steroid bursts to treat exacerbations.

• Poorly controlled asthma demonstrated by both –

  • Two or more severe asthma exacerbations requiring systemic corticosteroids of ≥ 30 mg of prednisolone or equivalent daily (or ≥ 50% increase in dose if maintenance dose of ≥ 30 mg prednisolone) for ≥ 3 days during the previous 12 months, despite the use of high-dose ICSs and additional controller medication.

  • ACQ (7-point scale) score of > 1 at screening visit 1 and randomisation visit 2.

• Atopic status –

  • Sensitisation to one or more perennial indoor aeroallergens (including house dust mites, domestic pets or fungi) to which participants are likely to be exposed during the study, demonstrated by a positive skin prick test against Dermatophagoides pteronyssinus (Der p 1) or Dermatophagoides farinae (Der f 1), Aspergillus fumigatus (Asp f 1), Alternaria alternata (Alt a 1) or Cladosporium herbarum (Cla h 1), cat – Felis domesticus (Fel d 1) – or dog – Canis familiaris (Can f 1) (wheal diameter of ≥ 3 mm more than a negative control test) or specific IgE of ≥ 0.35 IU/l determined by blood test.

• Recent medical stability –

  • Exacerbation free and taking stable maintenance asthma medications (not including short-acting bronchodilator or other reliever therapies) for at least 2 weeks prior to screening visit 1 and in the period between screening visit 1 and randomisation visit 2 (the screening period). Participants suffering a severe exacerbation during the screening period were rescreened 2 weeks after returning to their maintenance asthma medications.

• Adherence –

  • Able to use the TLA device during sleep on at least five nights per week (excluding holidays).

  • Able to understand and give written informed consent prior to participation in the trial and able to comply with the trial requirements.

Exclusion criteria

Potential participants who met any of the following exclusion criteria were excluded from participating in the study:

• Current smokers or ex-smokers abstinent for < 6 months.

• Ex-smokers with a ≥ 15 pack-year smoking history.

• Partner who is a current smoker and smokes within the bedroom where the TLA device is installed.

• TLA device cannot be safely installed within the bedroom.

• Intending to move out of study area within the follow-up period. (Participants moving out of the study area after randomisation were not automatically withdrawn. Every effort was made to continue treatment and trial follow-up.)

• Documented poor treatment adherence.

• Occupational asthma with continued exposure to known sensitising agents in the workplace.

• Previous bronchial thermoplasty within 12 months of randomisation.

• Treatment with omalizumab (Xolair^®, Novartis Pharmaceuticals UK Ltd, Grimley, UK) (anti-IgE) within 120 days of randomisation.

• Using long-term oxygen, continuous positive airway pressure or non-invasive ventilation routinely overnight, as it is known that this impairs the effect of the TLA device.

• Uncontrolled symptomatic gastro-oesophageal reflux that may act as a persistent asthma trigger.

• Presence of clinically significant lung disease other than asthma, including smoking-related chronic obstructive pulmonary disease (COPD), bronchiectasis associated with recurrent bacterial infection, allergic bronchopulmonary aspergillosis (mycosis), pulmonary fibrosis, sleep apnoea, pulmonary hypertension or lung cancer, that, in the opinion of the principal investigator (PI), is likely to be contributing significantly to a participant’s symptoms.

• Clinically significant comorbidity (including cardiovascular, endocrine, metabolic, gastrointestinal, hepatic, neurological, renal, haematological and malignant conditions) that remains uncontrolled with standard treatment.

• Currently taking part in other interventional respiratory clinical trials.

Excess treatment costs

The Department of Health and Social Care agreed to provide a subvention payment to cover the ETCs for those recruited to the active intervention arm, contributing £2088 per participant. The sponsor was able to access this funding on behalf of all sites only once the study had recruited sufficient participants in a year to reach a threshold of £50,000 costs. At this point, the Department of Health and Social Care would contribute £2088 for each additional participant recruited to the active intervention arm of the study. The first year began on the date that a centre recruited its first participant to the study and the second year began on the first anniversary of that date, the third year on the second anniversary, and so on. The maximum subvention available for this trial was £131,768. The NHS was required to meet all other treatment costs associated with the study through local commissioning arrangements. The sponsor was able to invoice the Department of Health and Social Care 6-monthly in arrears for the ETCs incurred above the £50,000-per-annum threshold on behalf of all of the recruitment sites. Centres in Wessex (Portsmouth and Southampton) had made arrangements for the Wessex Clinical Research Network to fund the ETCs up to the £50,000 threshold and the trial and recruitment were set up to maximise that support, which we gratefully acknowledge.

Active devices

The active TLA device (Airsonett^®; Airsonett AB, Ängelholm, Sweden) significantly reduces nocturnal allergen exposure by filtering ambient air through a high-efficiency particulate air filter, slightly cooling the air (0.5–0.8 °C) and ‘showering’ it over the participant during sleep. The reduced temperature allows the filtered air to descend in a laminar stream, displacing allergen-rich air from the breathing zone and thereby reducing allergen exposure without creating draught or dehydration. 39 The device is installed next to the participant’s bed, as shown in Figure 3, and is easy to use with no identified safety concerns in previous trials. The device is CE (Conformité Européenne) marked and licensed for use in the UK for allergic asthma. The device uses the same amount of electricity as a 60-W light bulb and has an anticipated lifespan of 5 years, with filter changes required every 6 months.

FIGURE 3.

The TLA device manufactured by Airsonett. Reproduced with permission from Airsonett AB (Ängelholm, Sweden).

Reproduced with permission from Airsonett

Placebo devices

The placebo devices were adjusted to deliver isothermal air instead of slightly cooled air and holes in the filter effectively allowed the air to bypass it while still maintaining an equivalent sound and airflow level as with the active device. This allowed the placebo device to deliver a laminar flow of non-filtered, non-descending, isothermal air that, when mixed with the warm body convection current, ascended towards the ceiling and thus had no effect on the normal air flow pattern around the breathing zone. There was no difference in the air delivery rate, perceived air movements or sound level between the active and the placebo device. The human body is not able to detect an absolute temperature difference of 0.75°C and, therefore, there is no perceptible temperature difference sleeping beneath an active or a placebo device. Electricity usage was the same as for the active device and the filter was changed at 6-monthly intervals.

Treatments when stable

All participants were evaluated during the study follow-up visits (study visits 3–6) by clinicians with expertise in severe asthma. These experts were able to identify and exclude alternative or comorbid pathologies contributing to poor asthma control and confirm treatment adherence.

No adjustment or reduction in asthma medications (excluding antihistamines and nasal corticosteroids) was allowed during the trial (unless required for patient safety reasons) because of the significant risk of precipitating severe asthma exacerbations. Any variation in non-asthma medication usage was recorded at each follow-up visit [including the use of over-the-counter (OTC) medications].

Those participants using variable maintenance and adjustable reliever therapy (MART), which combines ICS and bronchodilator therapy in a single inhaler, were converted to a fixed-dose regimen (preferably without changing inhalers) and an alternative short-acting bronchodilator [e.g. salbutamol, terbutaline (Bricany^®, AstraZeneca UK Ltd)] by the site team for the duration of the trial. A LASER ‘BDP equivalent’ dose calculator was developed to allow centres to easily calculate the BDP equivalent of their ICS based on the dose and frequency of use and on known pharmacokinetics of all available inhaled therapy.

Participants using a self-management plan (SMP) prior to the trial were allowed to continue and were asked not to change this during the trial treatment period.

Asthma exacerbations

Asthma exacerbations were managed following best clinical practice in the appropriate setting following the national BTS/SIGN guidelines. 16

If participants required urgent medical attention at any time during the follow-up period, they were instructed to call 999 and/or to attend the ED. Participants who did not require urgent medical attention were instructed to follow their normal process for seeking medical attention from their general practitioner (GP), practice nurse or asthma specialist within working hours and to contact their local primary care out-of-hours service at other times.

Participants who self-managed their OCSs were instructed to contact 999 if they required urgent medical attention or to self-manage in the community as directed by their agreed SMP if they did not require urgent medical attention.

Participants reported severe exacerbations to their local site trial team as soon as possible after exacerbation onset.

Clinicians prescribed the process for reducing and ultimately stopping corticosteroid treatment and returning to a normal maintenance dose after each exacerbation, determined by individual patient need.

Measurement of primary outcome

Participants were asked to start a participant exacerbation diary (PED) when exceeding the ‘exacerbation dose’ threshold of systemic corticosteroids, individually defined for each participant during randomisation visit 2. The PED included PEF measurements (using the trial electronic PEF device), OCS dose, reliever medication use and nocturnal asthma symptoms. Participants were asked to report severe exacerbations to their local site trial team as soon as possible after onset via a dedicated telephone line or a secure NHS e-mail account. Whenever possible, participants were asked to attend an exacerbation review with their local trial team within 72 hours to corroborate the exacerbation, at which the local trial team completed an exacerbation review form.

If participants were not able to attend an exacerbation review, the PED was collected at the next follow-up visit.

Information about exacerbations was also collected from the participant-completed daily diary in which participants recorded their daily corticosteroid dose (TLA diary) and from follow-up visit forms completed by the clinician delivering each follow-up visit.

Details about how these various sources of exacerbation data were combined to make a useable primary outcome are detailed in Data collection.

Quantitative

The three quantitative secondary outcomes of the LASER trial were:

1. Asthma control – to assess the impact of nocturnal TLA treatment on asthma control.

2. Quality of life – to ascertain the effect of TLA treatment on quality of life in participants with poorly controlled severe allergic asthma and their adult carers.

3. Impact and cost-effectiveness – to evaluate the impact of TLA treatment on health-care utilisation and related costs to fully assess the cost-effectiveness, both at 1 year and over the lifetime of the participant, of nocturnal TLA treatment using a cost–utility analysis to determine the incremental cost per QALY gained, and to evaluate the impact of TLA treatment on education/work days lost.

The data from which these outcomes could be determined were collected as detailed in the following sections. All data collected at the screening or randomisation visit were recorded by the attending clinician on CRFs. All data collected at any of the 3-, 6-, 9- or 12-month follow-up visits were recorded by the attending clinician on follow-up visit forms.

Qualitative

In addition to the quantitative secondary outcomes described (see Quantitative), there was one qualitative secondary outcome:

• acceptability – to qualitatively evaluate participant and carer perceptions, values and opinions of the trial process and TLA device to identify potential modifications to improve participant acceptance and to inform future implementation of the device within the NHS setting.

Supplementary variable

National charities

First, we partnered with national charities with an interest in the new treatment, Asthma UK and Allergy UK, which publicised the trial on their websites and posted trial information on both Twitter and Facebook that signposted patients to the LASER trial website; see Figure 7 for example content from Asthma UK, which has 29,000 followers on Twitter and has had > 40,000 likes on Facebook. Allergy UK has 5000 followers and has had > 7000 likes on Facebook.

FIGURE 7.

Asthma UK website, Facebook and Twitter content. Reproduced with permission from Asthma UK (London, UK).

Reproduced with permission from Asthma UK

Using Google Analytics, a spike in activity on the website was seen when posts/tweets were published.

Facebook and Twitter

Second, we partnered with Tillison Consulting (see https://tillison.co.uk; accessed 29 November 2018) to develop a LASER trial Facebook page and set up business accounts that facilitated trial advertisements on Facebook and Twitter. This business account permits targeted advertising, enabling us to set demographic parameters based on trial inclusion criteria (aged between 18 and 65 years, both men and women, location up to 50 km from a recruiting centre) and also to target people who have a search or following history that indicates an interest in asthma or asthma awareness. The tweets and posts directed people towards the study website and through the screening questions. Techniques such as ‘Google remarketing’ were also used, in which people who had visited the website would be shown ‘branded adverts’ in the following 30-day period, although these were less successful than the social media campaigns.

On Facebook, > 100,000 individuals were reached, there were 1361 clicks through to the website and a large number of patients engaged with the trial via ‘likes’, ‘shares’ and ‘comments’, which would have further increased the trial’s visibility if it led to patients sharing the information with friends or family with the condition. Similarly, on Twitter, > 150,000 people were reached.

Trialbee

Finally, in March 2015 we partnered with the Swedish company Trialbee (Malmö, Sweden), which specialises in social media solutions for clinical trials. On arriving at the Trialbee LASER trial ‘landing page’ (Figure 8), where there is information about the trial, interested patients answered a series of initial screening questions concerning their eligibility and those that passed screening were passed on to the trial team.

FIGURE 8.

Text from the Trialbee LASER trial landing page.

The Trialbee screening questions were answered 14,059 times, with 910 people passing questions. Of these, and following more in-depth screening by our trial team, 57 were eligible for the trial and 27 of these were consented and randomised to participate in the trial. This represented 16% of people recruited during this time period, demonstrating the power of social media for recruiting patients to clinical trials. In addition to being a cost-effective way to recruit patients, another benefit of social media engagement is that it empowers people to approach the research team of their own volition, thus possibly selecting more motivated, engaged research participants.

Delays in opening the first four and subsequent sites

There were a number of delays in opening the initial four recruiting sites because of delays in research and development (R&D) department approvals. We therefore identified a number of potential additional recruiting centres.

Change in National Institute for Health and Care Excellence recommendations for omalizumab

After the trial began, NICE reappraised its technology appraisal guidance for omalizumab,14 a drug manufactured by Novartis that is a costly injectable anti-IgE therapy, delivered from centres with asthma specialists, and a treatment option that competes directly with the TLA device tested in this trial. In the 2007 guidance,43 the NICE recommendation included confirmation of IgE-mediated allergy to a perennial allergen by clinical history and allergy skin testing and with either two or more severe exacerbations of asthma requiring hospital admission within the previous year or three or more severe exacerbations of asthma within the previous year, at least one of which required admission to hospital and a further two that required treatment or monitoring in excess of the patient’s usual regimen in an accident and emergency (A&E) unit.

The revised guidance broadened the inclusion criteria to include any patient who needed continuous or frequent treatment with OCSs (defined as at least four courses in the previous year).

This expanded the population of patients who might be eligible for omalizumab (no requirement to attend hospital or EDs and everyone on OCSs for asthma, irrespective of their prior exacerbation history). This had a negative impact on LASER trial recruitment as it dramatically shrunk the anticipated ‘prevalent’ population of severe asthma patients previously identified by recruiting centres when performing the extensive feasibility assessments and recruitment projections.

Omalizumab treatment pathway

This change in the licensed indication for omalizumab led to considerable variation among centres regarding the treatment option offered to patients. Several centres offered omalizumab therapy to patients prior to them being considered for the LASER trial, even though they potentially met the eligibility criteria for both treatments; some specialists were concerned that half of the patients may be randomised to placebo in this trial and therefore could still remain potentially uncontrolled.

The eligibility criteria allowed the recruitment of participants with severe allergic asthma who had failed or responded only partially to omalizumab therapy after a washout period of 120 days. Even then, a ‘trial’ of omalizumab treatment involves 2-weekly injections for 16 weeks before a patient is deemed to have responded or failed treatment. Thus, any potential participant who was first offered omalizumab in preference to the LASER trial device would have received omalizumab for 16 weeks (4 months) and then could not be screened for another 3 months, delaying recruitment to the LASER trial even further.

Competition with other commercial research trials

There has been a plethora of monoclonal antibody treatments for severe asthma in the last 3 years, sponsored by the pharmaceutical industry and facilitated by all local Clinical Research Networks through their commercial portfolio to encourage centres to participate in more commercial trials. The 30-day metrics of the Clinical Research Networks ensures that centres are incentivised and performance managed to recruit patients into all portfolio trials, including those that were competing with the LASER trial. Naturally, nearly all of our centres were approached to participate in such trials, with significant variations between centres in terms of participation; in some cases participants were recruited to other trials in preference to the LASER trial.

NHS England service specifications for severe asthma centres

In 2013, NHS England controversially approved the service specification for a limited number of centres to be designated as specialist asthma centres (SACs) (Birmingham, Brompton, Leicester and Manchester being approved at the time of commencement of the LASER trial). The designation and numbers of SACs were finally confirmed in March 2019. One criterion for specialist status had been the introduction of ‘bronchial thermoplasty’ as a new treatment option in severe asthma, with a requirement that each SAC perform 10 procedures per year to maintain status; those aspiring to become a SAC are required to perform a similar number of procedures over a 12-month period. Given the political incentives, potential participants eligible for the LASER trial were, in some instances, considered for bronchial thermoplasty in preference to the LASER trial in order to meet the politico-clinical targets set by the service specification.

Screen failures prior to randomisation

The number of patients requiring active screening to the point of consenting was not known; consequently, centres had been unaware of the number of potential participants that needed to be screened over any given period of time to ensure a sufficient volume of randomisation. Based on estimates, it was anticipated that two of every three patients screened would progress on to randomisation. Although relatively minor, this guidance allowed centres to manage their recruitment performance.

Resource problems

Despite the trial being on the Clinical Research Network portfolio, there were several instances of Clinical Research Networks not being able to support sites with clinical staff because of lack of funding, which required the transfer of potential participants to other centres. Other centres similarly completed a feasibility assessment but did not open because of staff shortages.

Solutions to boost recruitment

The actions taken to overcome the challenges to recruitment were wide-ranging:

• increase the number of centres

• greater engagement and performance management of existing centres

• website optimisation

• search engine optimisation

• use of Google Display

• Google remarketing

• use of social media

• media broadcasts and outputs to facilitate recruitment

• broaden eligibility criteria (i.e. reduction in ICS dose required, reduction in lower age limit for inclusion and reduction of screening period from 4 to 2 weeks)

• modification of patient pathway at existing centres to ensure trial visibility (Chief Investigator visited some of the centres)

• detailed interrogation of primary care for suitable participants.

Extension of the screening period

To continue to be deemed potentially eligible for the trial, individuals needed to demonstrate acceptable compliance with the electronic PEF recordings and ACD during the 2-week screening period. However, in the event of electronic PEF device malfunction or if, in the investigator’s opinion, there were significant extenuating circumstances, the screening period was extended by up to a further 2 weeks. Participants experiencing a severe exacerbation during the screening period were no longer eligible but could be rescreened 2 weeks after returning to their maintenance asthma medications.

Demographics, asthma history and asthma reviews

The following data were recorded about each participant on the CRF at the randomisation visit only:

• demographics –

  • age (P1, P4)

  • sex (P4)

  • socioeconomic class (P4)

  • ethnicity (P4)

• asthma history –

  • date of asthma diagnosis (P1, P4)

  • history of life-threatening and near-fatal asthma exacerbations [intensive treatment unit (ITU) admissions] (P6)

  • number of severe asthma exacerbations in previous 12 months (P1, P5)

  • history of previous asthma treatment (P1)

  • history of atopy (P1)

  • family history of asthma/atopy (P6)

  • asthma triggers (P1)

  • medical or surgical comorbidities (P1)

  • occupational history (P1)

  • smoking history (P1)

  • height (cm)/weight (kg) for measuring predicted lung function (P3).

The following data were collected at each follow-up visit as well as at the randomisation visit and recorded by the attending clinician on the follow-up visit form/CRF:

• asthma review

  • current asthma symptoms and treatment (P1, P3, P5)

  • current medications (P1)

  • history of severe asthma exacerbations since the previous trial visit and current participant-reported clinical status (still in exacerbation or recovered) (P2)

  • unscheduled asthma-related health-care use (P3)

  • work/study days lost as a result of asthma symptoms (P3).

Lung function measures

The following indicators of lung function were collected at the study visits:

• Pre-bronchodilator FEV₁ (P1, P3, P5) – spirometry was conducted at the screening visit, the randomisation visit and the 3-, 6-, 9- and 12-month follow-up visits to collect the following variables:

  • FEV₁ (l)

  • FVC (l)

  • FEV₁/FVC ratio

  • forced expiratory flow at 25–75% of the pulmonary volume (FEF_25–75) (%).

  FEV₁ and FVC were documented both as absolute values and as a percentage of the predicted value. A spirometer conforming to ATS/ERS standards44 was used as specified by the manufacturer’s instructions.

• Reversibility testing (P1, P3, P5) – post-bronchodilator FEV₁ (both percentage change and volume change) was measured at the screening visit and 12-month follow-up visit only. Following ATS/ERS standards,44 post-bronchodilator FEV₁ was defined as FEV₁ recorded 15 minutes after administration of 400 µg of salbutamol via a metered dose inhaler and spacer device. An improvement in FEV₁ post-bronchodilator use of ≥ 12% or 200 ml was considered significant.

• FeNO (P1, P3, P5) – FeNO was measured before spirometry at the randomisation visit and the 3-, 6-, 9- and 12-month follow-up visits. The measurements were made using a NIOX MINO^® device (Aerocrine AB^®, Solna, Sweden) as specified by the manufacturer’s instructions and outlined in the ATS/ERS standards. 45

Allergy testing

The following allergy tests were made during the screening visit to determine whether the allergy-related trial inclusion criteria were met:

• Skin prick testing (P1) – a standard skin prick test procedure using common indoor aeroallergen (Der p 1, Der f 1, Asp f 1, Alt a 1, Cla h 1, Fel d 1 and Can f 1) extracts along with negative (saline) and positive (histamine) controls was performed on all subjects. This occurred during the randomisation visit instead of at the screening visit if antihistamine hold was required (see Appendix 1). Skin prick testing was performed in accordance with the practice parameter released by the American Academy of Allergy, Asthma and Immunology. 46 A positive skin prick test reaction was measured as a wheal of at least 3 mm diameter greater than the negative control.

• Serum-specific IgE testing (P1) – if skin prick testing was not available, a blood sample was taken to measure serum-specific IgE to common indoor aeroallergens (Der p 1, Der f 1, Asp f 1, Alt a 1, Cla h 1, Fel d 1 and Can f 1). A specific-serum IgE concentration of > 0.35 IU/l was considered to represent allergen sensitisation. Serum-specific IgE testing could also have been used if there was uncertainty about a skin prick test result or if there was a negative skin prick test result in the context of a patient on long-term maintenance systemic corticosteroids.

• Measurement of serum total IgE and peripheral blood eosinophil count (P1) – a blood sample was collected to measure serum total IgE and peripheral blood eosinophil levels.

Participant questionnaires

• Asthma Control Questionnaire (P3): the well-validated 7-item ACQ47 was used to assess asthma control over the 7 days leading up to the screening visit, the randomisation visit and the 3-, 6-, 9- and 12-month follow-up visits. It was administered at the same time during each visit, with the participant blind to the results of other tests. The ACQ includes five symptom scores, the amount of daily rescue bronchodilator usage and a measure of airway calibre (FEV₁% predicted). Responses are given on a 6-point scale and the overall score is the mean of the responses (0 = totally controlled, 6 = severely uncontrolled). The ACQ has strong evaluative and discriminative properties and has been shown to be very responsive to within-patient changes in asthma control over time. It has a validated minimally important difference of 0.5 to demonstrate clinical significance.

• Standardised Asthma Quality of Life Questionnaire (P3): asthma-specific quality of life was measured using the AQLQ(S)48 at the randomisation visit and the 3-, 6-, 9- and 12-month follow-up visits. The AQLQ(S) consists of 32 questions within four domains – symptoms, activity limitation, emotional function and environmental stimuli – and it has strong measurement properties and a validated minimally important difference of 0.5. 49 Patients are asked to think about how they have been during the previous 2 weeks and to respond to each of the 32 questions on a 7-point scale (7 = not impaired at all, 1 = severely impaired). The overall AQLQ(S) score is the mean of all 32 responses and the individual domain scores are the means of the items in those domains.

• EuroQol-5 Dimensions, five-level version (P3): generic health-related quality of life (HRQoL) was measured using the EQ-5D-5L questionnaire50 at the randomisation visit and the 3-, 6-, 9- and 12-month follow-up visits. The EQ-5D-5L is a standardised measure of health providing a simple generic measure of health for clinical and economic appraisal. Patients are asked to think about their health on the day that they are completing the questionnaire and report on any problems (none, slight, moderate, severe and unable/extreme) relating to five attributes (mobility, self-care, usual activities, pain/discomfort and anxiety/depression). Patients are then required to rate their health using a 100-point visual analogue scale (VAS) (0 = worst health you can imagine, 100 = best health you can imagine). The EQ-5D-5L is the most widely used HRQoL measure in adults in the UK and has been shown to be a reliable and valid means of measuring quality of life in asthma patients. 51 In addition to the EQ-5D-5L score, the EuroQol Visual Analogue Scale (EQ-VAS) was also ascertained at each of these visits. The EQ-VAS was determined by asking the participants to indicate their health status at each visit on a 20-cm vertical scale with end points of 0 and 100.

• 22-item Sino-Nasal Outcome Test (P3): the SNOT-22 score is a well-validated and sensitive measure of rhinosinusitis health status52 that was recorded at the randomisation visit and the 3-, 6-, 9- and 12-month visits. The SNOT-22 consists of 22 questions related to symptoms and the social/emotional impact of those symptoms (symptoms are rated on a scale from 0 = no problem to 5 = problem as bad as it can be). Participants were asked to rate the problems according to how they had been over the previous 2 weeks.

• Indoor Air Quality Questionnaire (P6): the Indoor Air Quality Questionnaire was completed by participants at the randomisation visit to identify key factors affecting air quality within the home environment. This questionnaire is a bespoke domestic indoor air quality assessment tool derived from a combination of the Stockholm53 and Southampton54 Indoor Environment Questionnaires.

• Global Evaluation of Treatment Effect questionnaire (P3): the GETE questionnaire55 is a simple measure of perceived treatment effectiveness that was completed by participants and clinicians at the 12-month visit. This questionnaire has been used in the evaluation of other treatments in patients with severe allergic (IgE-mediated) asthma. For the purposes of the LASER trial, it required participants and physicians to rate the global treatment effectiveness of the TLA device as excellent (complete control of asthma), good (marked improvement of asthma), moderate (discernible but limited improvement in asthma), poor (no appreciable change in asthma) or worsening (deterioration in asthma).

• Work Productivity and Activity Impairment (P3): the WPAI(A) version 2 is a validated questionnaire tool for assessing work productivity and activity impairment56 that was completed by participants at the randomisation visit and at the 3-, 6-, 9- and 12-month visits. This questionnaire consists of six questions addressing absenteeism, presenteeism (reduced effectiveness while working), overall work productivity loss (absenteeism and presenteeism) and activity impairment. Participants were asked to count the number of hours that they had missed from work because of problems associated with their asthma and the number of hours missed for other reasons (such as holidays) over the past 7 days. In addition, participants were asked to report the impact of asthma on their work and on the ability to do regular daily activities (on a scale from 1 = asthma had no effect to 7 = asthma completely prevented me). WPAI(A) outcomes are measured as percentages; a higher percentage relates to greater impairment and reduced productivity. A modified WPAI(A) was used for student participants.

Adult carer questionnaires

• Adult Carer Quality of Life (P3): the AC-QoL questionnaire57 measures the overall quality of life of adult, unpaid carers and was completed by participant carers at the randomisation visit and again at the 12-month visit. The eight domains of this 40-item questionnaire assess support for caring, carer choice, carer stress, money matters, personal growth, sense of value, ability to care and carer satisfaction. Carers are asked to recall their experiences over the previous 2-week period. Scores on the questionnaire have a possible range of 0–120, with higher scores indicating better carer quality of life.

• Work Productivity and Activity Impairment (P3): a modified WPAI for caregivers [WPAI(CG)] was completed at the randomisation visit and 12-month follow-up visit.

Asthma Control Diary

Participants were issued with a validated ACD58 at the screening visit, the randomisation visit and the 3-, 6- and 9-month visits to record data for the 2 weeks leading up to each subsequent visit. Towards the end of the trial, a text reminder system was set up to remind participants when to start their ACD and to remind them of the date of their follow-up visit.

Participants recorded the following data on a daily basis for 2 weeks:

• PEF rate – during the trial, participants performed three morning PEF measurements using a handheld device supplied by the trial team and recorded these measurements in the ACD. During the screening period, participants also performed three evening PEF measurements to assess variability as part of the eligibility assessment. These additional PEF data were stored on the PEF device and downloaded at the randomisation visit.

• Symptom and reliever medication use – the ACD measures a morning score (two items; 0–6 point scale), a bedtime score including bronchodilator requirement (four items; 0–6 point scale) and a best morning PEF rate measured as percentage of predicted best (0–6 point scale). The overall daily score is the mean of the responses (0 = perfectly controlled, 6 = severely uncontrolled).

• Device usage – device usage data (displayed on the device screen) documenting the number of hours that the device had been active were collected by:

  • participants at 3, 6, 9 and 12 months

  • the engineering team at 6 and 12 months to coincide with planned filter changes.

  These data were recorded in the ACD, for review at the 3-, 6-, 9- and 12-month follow-up visits. Device usage was also one item discussed during the 1-month telephone review, revealing where mitigating steps needed to be put in place to improve adherence early on in the trial.

Temperature-controlled laminar airflow diary

Participants were issued with a TLA diary at the randomisation visit to record the following data on a daily basis:

• use of the TLA device (hours), that is, participant-reported treatment adherence

• reliever used (number of times)

• dose of corticosteroids (mg)

• work/study days lost because of asthma symptoms (hours).

Adherence to completing the TLA diary was another item discussed during the 1-month telephone review, revealing where mitigating steps needed to be put in place to improve adherence early on in the trial.

Resource use log

Participants were issued with a resource use log at the randomisation visit and the 3-, 6- and 9-month follow-up visits to record health-care resource use between study visits. The purpose of these logs was to help participants recollect events as they completed the participant questionnaires administered during the follow-up visits. To keep the questionnaires as simple as possible, they did not distinguish whether the contacts with the health-care service were through the NHS or through private providers, meaning that an underlying assumption of the resource use analyses is that all contacts were financed by the NHS.

The date stamps of the hospitalisation records were correlated with those recorded by health-care providers to avoid double counting.

Participant exacerbation diary and exacerbation review form

Participants were asked to start a PED when exceeding the ‘exacerbation dose’ threshold of systemic corticosteroids individually defined for each participant during randomisation visit 2. The PED recorded the following during exacerbations for a period of 7 days: morning PEF ( l/minute) (recorded using the trial-supplied electronic PEF device), steroid dose (mg), reliever use (number of times) and nocturnal wakening (yes/no).

In addition to completing the PED, participants were asked to report severe exacerbations to their local site trial team as soon as possible after onset via a dedicated telephone line or a secure NHS e-mail account. Whenever possible, participants were asked to attend an exacerbation review session with their local trial team within 72 hours to corroborate the exacerbation, at which the local trial team completed the exacerbation review form. The following information was captured on the exacerbation review form: onset of symptoms, dose of steroid used (mg), corroboration of use of oral steroids of at least 30 mg for 3 days, and change in symptoms and reliever use.

If participants were not able to attend an exacerbation review, a review of the exacerbation occurred by telephone and a exacerbation review form was still completed. If a telephone review was not possible either, an exacerbation review form was completed at the next follow-up visit. The nature of the review (face-to-face or telephone) was recorded on the exacerbation review form and the participant’s PED was collected.

Follow-up visit forms

A regular 3-monthly follow-up visit form was completed by the attending clinician, with participant input at each follow-up visit (at 3, 6, 9 and 12 months), prompting discussion with participants about key issues related to the preceding 3 months. This included number of exacerbations, a review of the TLA diary, medication use, health resource use log, work and study days lost, reported device use, completion of trial questionnaires, FeNO and spirometry.

Sampling and recruitment

Informed consent for participation in the qualitative study was sought at the screening visit. All participants taking part in the LASER trial were contacted towards the end of their 12-month follow-up period with an invitation to attend local focus group sessions.

Not all participants who consented to take part in the focus group sessions were selected: 5–10 participants were selected for each of the focus group sessions on the basis that they best reflected multiple variation (including balance of sex, age and ethnicity).

Focus groups

A focus group is an interview technique in which a group of individuals are encouraged to debate and discuss specific topic areas. The rationale for choosing focus group discussions during the LASER trial was to stimulate alternative views and experiences, evaluate the process and assess the feasibility of the intervention.

In total, 28 participants were invited to attend focus group sessions based on the fact that they had completed at least 9 months of the trial and had consented to attend local focus group sessions when asked at the start of the trial; 20 of these 28 participants agreed to participate and were sent invitation letters to attend one of three focus group sessions hosted by the Portsmouth site during January 2016. The three sessions were held on separate dates at different times of the day, including an evening focus group, in order to provide choice to those invited to attend and maximise attendance. The venue chosen was a non-NHS, non-trial centre, within an easily accessible location with on-site parking, namely the Premier Inn at Port Solent.

Acceptance of the invitations covered consent for the use of verbatim quotations and assurance that the information that a participant provided would be kept strictly in confidence and anonymised.

The focus groups were led by a senior qualitative researcher from the University of Portsmouth, with a senior medical research fellow attending as facilitator and note keeper.

Focus group 1 was held on 21 January 2016 at 18.00. Five participants agreed to attend but only one participant attended on the day. It was agreed to conduct a one-to-one interview with this participant, following the focus group topic guide.

Focus group 2 was held on 26 January 2016 at 10.00. Seven participants agreed to attend but only three participants attended on the day.

Focus group 3 was held on 28 January 2016 at 14.00. Eight participants agreed to attend but only six attended on the day.

In total, 10 participants took part in the focus group sessions, seven females and three males. These 10 participants comprised both satisfied and non-satisfied participants (determined from a combination of device-reported usage and participant-reported device usage).

At the beginning of each focus group discussion, the facilitator checked that participants were aware of the purpose of the discussion, likely topics to be discussed and their right to leave, at any time, without giving a reason for doing so. All participants were asked to respect the views of others and to take turns in speaking to aid the recording of the discussions.

The topics were based on key themes identified during the qualitative telephone interviews conducted during the pilot phase. Free discussion of experiences and ideas was encouraged throughout.

The focus group discussions were audio-recorded and lasted between 60 and 90 minutes. Afterwards, audio-recordings were transcribed verbatim by the Way With Words (London, UK) secure online transcription service. Participants were offered tea or coffee and biscuits, with water available throughout.

Primary statistical analysis

The primary statistical analysis was based on the intention-to-treat (ITT) population, which is defined as all participants included and analysed according to their allocated treatment group irrespective of the treatment received. Note that, in the primary analysis, the five participants who were randomised and withdrew consent to allow data usage were still included, with data set to missing and an assumption that no severe exacerbations were experienced. All other participant data deemed to be ineligible post randomisation or associated with protocol violations were also included in these primary outcome data. Two blinded analyses (without information on treatment allocation added to the clinical database) of the data were undertaken to define the per-protocol population used for the sensitivity analyses (see Defining populations for data analysis) and the minimum data population to identify participants with a threshold minimum number of data:

• Per-protocol population (as the ITT population but excluding) –

  • participants who did not receive a device

  • participants who withdrew consent for data use.

• Minimum data population – all participants who have at least 90 days of steroid dose information reported (in the TLA diary or PED) or who have reported at least one PED. This requirement included the assumed doses in the TLA diary and PEDs when the dosage was left blank (see Procedure for dealing with missing, unused and spurious data for explanation).

Secondary statistical analysis

Analysis of secondary outcomes was carried out on a variable-by-variable complete case basis, that is, participants were included in each individual secondary outcome analysis only if all data relevant to that specific analysis were available, with no requirement for their remaining (non-relevant) data. Secondary outcomes with insufficient data were not analysed.

Definitions

The primary end point definition was rate of severe exacerbations in a 12-month period.

Severe exacerbations were defined in accordance with ATS/ERS guidelines40 as a worsening of asthma requiring systemic corticosteroids, that is, ≥ 30 mg of prednisolone or equivalent daily (or a ≥ 50% increase in dose if maintenance 30 mg of prednisolone or above) for ≥ 3 days.

Courses of corticosteroids separated by ≥ 7 days were treated as separate severe exacerbations.

A post hoc analysis included the worsening of asthma requiring systemic corticosteroids, that is, ≥ 10 mg of prednisolone or equivalent daily for ≥ 3 days, with exacerbations separated by ≥ 7 days treated as separate exacerbations.

Data collection

In an attempt to not miss any exacerbations, the primary outcome was collected from several different sources, which are listed in Table 4, along with a description of the source of these data (patient or trial site). A description of the nature of the data available for each source is also given as some sources provided detailed information including dates and doses of corticosteroids, and others provided only the number of severe exacerbations with no corroboration regarding the severity and duration of the attack in relation to the definition described above (see Temperature-controlled laminar airflow diary, Participant exacerbation diary and exacerbation review form and Follow-up visit forms for full descriptions of the data collected by each of these sources).

After discussion with the trial team, investigators, TSC and DSMC, it was decided that the most appropriate information to use for the primary outcome was the dated information as this was the most accurate and would enable duplicate reporting to be removed and the definition to be applied. The primary outcome for this study was therefore determined from the dated primary outcome data only. The undated primary outcome data were used for the sensitivity analyses only, which are described in Sensitivity analyses.

Statistical model

The primary efficacy end point, which was the rate of severe exacerbations over the 12-month period, was analysed using the negative binomial model. The only adjustment made to the model was the minimisation factors used in the random assignment. The minimisation factors used in the model were prevalent compared with incident cases (new to the centre), exacerbation frequency in the previous 12 months (two, three, more than three), use of OCSs (yes/no) and pre-bronchodilator FEV₁ (> 50% predicted yes/no). The recruiting centre was included in the model.

The negative binomial model was chosen over the originally planned Poisson model because it is more flexible and gave a better fit to the overall blinded data.

In the current data, there are approximately 60 cases in which the participant reported a severe exacerbation in the PED but there was no report or no data in the daily diary. There are also approximately 50 severe exacerbations reported by the research team without corresponding daily diary entries or a PED report. Thus, there are approximately 110 severe exacerbations without documentation of how long the participants were being observed. Therefore, as it could not be ascertained which days participants were actively reporting severe exacerbations if they did not complete the daily diary, it was decided that, for dated severe exacerbations, the observation period would extend from randomisation to 365 days after randomisation for the primary analysis. By taking a full year as the denominator for all participants, we essentially assumed that all days were at risk as we did not have exact days at risk for the model. This method most likely leads to an underestimation of the rate of events; however, it was deemed the best approach to handling the problem of a large number of missing diary data. See the methods section on the project web page (www.journalslibrary.nihr.ac.uk/programmes/hta/123328/#/; accessed 29 November 2018) for more details.

When using undated data from the follow-up visit form, all data up to and including the 12-month follow-up visit were used.

Sensitivity analyses

The following individual sensitivity analyses were carried out to test the robustness of the primary outcome to the missing data.

Data processing

Several sources provided primary end point data. Reports of exacerbations close in date were inspected and adjusted to prevent duplicate counting of events; for example, if the PED said that an event started 1 day later than the diary data suggested, then only one exacerbation was counted.

Thereafter, and following blinded review of the data (blinded to treatment allocation) and discussion with the clinical team, including the Chief Investigator, TSC and DSMC, data processing decisions as given in the statistical analysis plan (SAP) supplement [see the methods section on the project web page (www.journalslibrary.nihr.ac.uk/programmes/hta/123328/#/)] were made.

Placebo group

For the placebo group, data for 29,147 (66.1%) of the 44,091 possible study days of daily maintenance steroid use were collected; 6419 (22.0%) of the 29,147 days of maintenance dose information analysed were based on the assumption of a zero dose because this field of the daily diary was left blank but other fields (i.e. any of device use, reliever use or time off work) were completed.

In addition to missing maintenance steroid dose data, exacerbation steroid dose data were also missing. A 30-mg dose was assumed for the 24 days for placebo group cases for which the exacerbation dose was left blank in the PED.

Active group

For the active group, data for 28,753 (66.2%) of the 43,398 possible study days of maintenance steroid dose use were collected; 4819 (16.8%) of the 28,753 days of maintenance dose information analysed were based on the assumption of a zero dose because this field of the daily diary was left blank but other fields were completed.

In addition to missing maintenance steroid dose data, exacerbation steroid dose data were also missing. A 30-mg dose was assumed for the 25 days for active group cases for which the exacerbation dose was left blank in the PED.

The rate at which dose data were available for analysis (either because they were reported or because a zero dose was assumed) over the 1-year duration of the trial is shown in Figure 10. The assumed data (zero dose or PED data) are plotted on a participant-by-participant basis in Figure 11. The voids indicate missing data, which are plotted in Figure 12. Table 6 summarises the missing data assumptions.

FIGURE 10.

Percentage of the 240 randomised participants who had a dose of oral corticosteroid reported (or an assumed zero dose) on the nth study day. Day of randomisation = 0, 1-year post randomisation = 365 days. LOWESS (locally weighted scatterplot smoothing) smoothed average applied with bandwidth 0.02.

FIGURE 11.

Assumed data: zero dose (in green) and 30-mg dose (in blue) by participant and days since randomisation.

FIGURE 12.

Missing dose data by participant and days since randomisation. Long tails are the result of follow-up visits that do not have related TLA diary data. For each participant, at the beginning of the trial, the diary was often not completed prior to delivery of the device. After the device was available, the completion rate was around 70% for the rest of the trial.

Sensitivity analyses

The sensitivity analyses designed to test the robustness of the results to the missing data, participant population definition and alternative definitions of the primary outcome (as described in Chapter 2, Sensitivity analyses) show that the primary outcome results in Table 10 are very sensitive to the assumptions made; by varying these assumptions the primary outcome treatment effects could either favour or not favour the TLA device. This conclusion was drawn by investigating the best- and worst-case scenarios for those participants who did not meet the minimum data requirement (see Chapter 2, Asthma Control Diary for definition). The following best- and worst-case scenarios describe the best- and worst-case numbers of severe exacerbations that might have occurred in this population of 45 participants during the 12-month trial period but which were not captured in the primary outcome data set:

• modified best and worst case –

  • best case = 0

  • worst case = 4

• plausible best and worst case –

  • best case = 0

  • worst case = 2.

The results of the sensitivity analyses made using the best- and worst-case scenarios are as follows:

• The point estimates of the risk ratio ranged from 1.4 (worst for active) to 0.6 (best for active) in the modified best- and worst-case analysis for the dated exacerbations, highlighting the wide range of possible outcomes under different assumptions about the outcomes for those with less than the minimum number of data reported. The plausible best- and worst-case analysis gave point estimates of the risk ratio ranging from 1.1 to 0.7 (sensitivity analysis 2; see Chapter 2, Study assessments, Participant exacerbation diary and exacerbation review form).

• The best- and worst-case analysis carried out using the maximum number of exacerbations as outcomes resulted in point estimates of 1.5 to 0.6, again highlighting the wide range of possible outcomes under different assumptions about the outcomes for those with less than the minimum number of data reported (sensitivity analysis 3a; see Chapter 2, Study assessments, Participant exacerbation diary and exacerbation review form).

• The best- and worst-case analysis carried out using the dichotomised dated exacerbations (participants with one or more severe exacerbations vs. no severe exacerbations) resulted in point estimates ranging from 2.6 to 0.5 (sensitivity analysis 3c; see Chapter 2, Study assessments, Participant exacerbation diary and exacerbation review form). The corresponding analysis made using the maximum number of severe exacerbations definition resulted in point estimates ranging from 3.5 to 0.5 (sensitivity analysis 3d; see Chapter 2, Study assessments, Participant exacerbation diary and exacerbation review form).

Table 10 presents the risk ratio point estimates for the different primary outcome sensitivity analyses. None of these risk ratios was statistically significant. There is therefore no evidence that the null hypothesis of no difference between the active and the placebo treatment can be rejected. The number of missing data could mask a difference between the treatments, but this could be in either direction.

Post hoc analyses

The TSC requested two post hoc analyses of the primary outcome: (1) analysis of an alternative definition of a severe exacerbation and (2) subgroup analysis of patients who do and do not use dust mite covers.

Pilot phase telephone interviews

In total, 12 telephone interviews were conducted with participants and partners during September and October 2014 (see Chapter 2, The internal pilot study, for a description of these interviews). A full report of the findings from these interviews was prepared and submitted to the funding body (see Report Supplementary Material 1). A summary of the categories of themes emerging from these interviews is provided in the following sections, according to the responses from the 10 participants and those from the two partners. The method by which the content of these interviews was distilled down to eight categories is described in Chapter 2, Qualitative analysis.

Focus groups

A total of 10 participants (without partners) attended the three focus groups conducted by the Portsmouth site during January 2016 (see Chapter 2, Focus groups, for a description of these focus group sessions). Some of the topics discussed at these focus group sessions were prompted by the attending researcher, based on the categories that emerged from the pilot phase telephone interviews (see Pilot phase telephone interviews). The remaining topics of discussion were generated from the attending participants without prompts. All of the topics were consolidated into categories according to the methodology described in Chapter 2 (see Qualitative analysis) and these categories are described in detail in the following sections.

The learning points applicable to this clinical trial and the TLA product specifically and to future clinical trials generally are highlighted at the end of each section. Selected anonymised verbatim quotes are included to best illustrate the theme being described.

It is worth noting that these qualitative findings are based on a small purposive sample of trial participants and it is unlikely that the comments made during the focus group interviews can be generalised to include all participants in the trial and their experiences of using the TLA device. Nevertheless, the responses received do represent a range of participants and describe their experiences of the treatment, the device and the trial.

NHS

Unit costs for consultations with GPs and nurses were taken from the Personal Social Services Research Unit’s Unit Costs of Health and Social Care publications for 201562 and 2016. 63 For all other NHS contacts, unit costs were derived from the NHS Reference Costs 2015 to 2016. 64 For outpatient visits, we used the weighted average of all consultant-led, non-admitted face-to-face attendances provided by the NHS during a year (including weighting by first or follow-up appointment). The same method was used for visits to physiotherapists, occupational therapists and psychologists. In the absence of specific unit costs for counsellors, we assumed that these would be the same as for a psychologist. For visits to A&E, we used the weighted average of all emergency medicine contacts, excluding dental care and participant dead on arrival. For ambulance transport to A&E, the unit cost of a call to the emergency services was included, as well as that for ambulance transport.

Using the reasons for hospitalisation reported by participants in the resource use questionnaires, we obtained diagnosis and procedure codes. These were then translated into a Healthcare Resource Group (HRG) using the HRG4+ 2015/16 Reference Costs Grouper (https://digital.nhs.uk/services/national-casemix-office/downloads-groupers-and-tools/grouper-and-tools-archive/; accessed 2 April 2019). Each HRG was then linked to a series of elective, non-elective and day case reference costs obtained from the NHS Reference Costs 2015 to 2016. 64

Non-NHS, including productivity

Costs of prescribed and OTC medications were obtained from the British National Formulary (BNF). 65 For complementary medicines or supplements not listed in the BNF, unit costs were derived from commercial suppliers such as Holland & Barrett (Nuneaton, UK), which is one of the largest suppliers of health foods in the UK.

To cost the hours of missed work as a result of asthma, we used the mean hourly earnings for men and women, which are £17.01 and £14.06, respectively. 66

Standardised Asthma Quality of Life Questionnaire: disease-specific quality of life

The overall AQLQ(S) score for each participant at each follow-up visit was calculated as the mean of all 32 responses to the questionnaire. We also evaluated each participant’s score for each domain individually (symptoms, activity limitations, emotional function and environmental stimuli) as the mean of all responses in that domain. The mean differences in the AQLQ(S) scores of the two treatment groups (both across all questions and for each domain individually) were calculated for each follow-up visit with 95% CIs. The statistical significance of these differences was assessed using two-sided t-tests. The differences in the means of the 12-month and randomisation scores were also compared using two-sided t-tests.

EuroQol-5 Dimensions: generic quality of life

As recommended by NICE,69 and for those participants with complete EQ-5D-5L and EQ-VAS data only, the responses to each of the five questions in the EQ-5D-5L were converted into utilities using the validated mapping function to derive utility values for the EQ-5D-5L from those existing for the EQ-5D-3L (EuroQol-5 Dimensions, three-level version). 70 The mean differences in the utility values and EQ-VAS scores between the two treatment groups at each follow-up visit were calculated with 95% CIs. The statistical significance of these differences was assessed using two-sided t-tests. Changes between randomisation and 12-month scores were also compared using two-sided t-tests.

Carer quality of life

An insufficient number of AC-QoL questionnaires were completed by adult carers of the participants for these results to be meaningful.

NHS resource use

Except for number of days in hospital, statistical differences in NHS resource use between the two participant groups were evaluated assuming a Poisson distribution. Differences in number of days in hospital between groups were assessed using a Student’s two-sided t-test. Resource use was calculated for each 3-month follow-up visit period using the available data, as well as for the full 12-month period using complete data only, that is, from those participants with 3-, 6-, 9- and 12-month resource data recorded.

Non-NHS resource use

The mean differences in the expense borne by participants in the two participant groups over the 12-month trial period for OTC medicines and supplements were calculated with 95% CIs. The statistical significance of these differences was assessed using two-sided t-tests.

Productivity losses

The responses to the six WPAI(A) questions were used to assess productivity losses. For the question ‘Currently employed?’, the percentages for the two participant groups were calculated and the difference in these rates was assessed using the chi-squared test. The differences between the groups in terms of the responses to the remaining questions, which all have numerical responses (‘Hours missed due to asthma’, ‘Hours missed due to other reasons’, ‘Hours actually worked’, ‘Impact of asthma on work productivity’ and ‘Impact of asthma on regular daily activities’), were assessed using a Student’s two-sided t-test.

For participants with missing follow-up visit WPAI(A) responses, we used, when available, data from other questionnaires collected at that follow-up visit that also asked about the number of days of work and/or study missed as a result of asthma. In addition, when WPAI(A) information was missing, we assumed that those participants who reported being retired at baseline would continue to be so at follow-up. The WPAI(A) determined the number of daily work hours missed over a 7-day interval; we multiplied this by 91.25 days to calculate the productivity losses (i.e. work hours missed) over a 3-month period representing the interval between study visits.

An insufficient number of caregiver WPAI questionnaires were completed by adult carers of the participants to power a statistical analysis of these responses.

Costs of the device

The costs of the TLA device were provided directly by the manufacturer (Airsonett). 72 After consultation with Airsonett, we assumed that the lifetime of the device would be 10 years. We estimated costs, excluding value-added tax (VAT), under three different scenarios:

1. Device rental. Under this scenario, rather than purchase the device, the NHS would rent the device from the manufacturer. Under this rental agreement, the costs of device reconditioning, replacement of spare parts, servicing and filter changes would be included. The costs for the initial month of rental were £690 and for subsequent months were £174. Hence, the annual cost would be £2604 over the first year of rental and £2088 for subsequent years. Averaging the cost of rental over 10 years, the mean annual cost would be £2140.

2. Purchase with no full service agreement. Under this scenario, the NHS would purchase the device for £8780. It would then have to cover the costs of reconditioning, filters (two every year, excluding the first one, which is provided with the device) and any device accessories that would need replacing. Table 17 provides the unit costs of reconditioning, filters and accessories and the number of times that these would be required over the lifetime of the TLA device. We assumed that each accessory would require replacing once because of breakage.

   To estimate a mean annual cost, we first estimated the present value cost of device acquisition, reconditioning, filter changes and accessories. For this, we discounted all future costs using an annual rate of 3.5%, assumed that reconditioning and device accessory changes would occur at 5 years and added these costs to the costs of device acquisition and one filter change (i.e. costs incurred during the first year). In present value terms, the cost of acquisition, reconditioning, filter changes and device accessories would be £17,797. To obtain an equivalent annual cost,73 we depreciated these costs over the 10-year lifetime of the device using an annual rate of 3.5%. Therefore, the equivalent annual cost of this scenario to the NHS was £2140.

3. Purchase with full service agreement. Under this scenario, the NHS would purchase the device for £8780. With a full service agreement the NHS would then, after the first year, pay an annual maintenance fee of £1125.60, which would cover the costs of device reconditioning, replacement of spare parts, servicing and filter changes. To estimate a mean annual cost, we first estimated the present value cost of acquisition and service of the device. For this, we discounted the future costs of servicing using an annual rate of 3.5% and added these costs to the acquisition cost. In present value terms, the cost of acquisition and servicing of the device would be £17,643. To obtain an equivalent annual cost, we depreciated these costs over the 10-year lifetime of the device using an annual rate of 3.5%. 74 Therefore, the equivalent annual cost of this scenario to the NHS was £2121.

Given that there was no considerable difference in the annual cost under these three scenarios, we used the costs under scenario 1 (i.e. rental agreement) in the analysis.

Total societal costs

The total societal cost is the sum of the NHS resource use and non-NHS resource use, including costs arising from the productivity losses. The difference between groups in the mean total societal cost at each 3-month follow-up visit for available cases, as well as for the full 12 months where complete data existed, was assessed using Student’s two-sided t-test.

Complete-case analysis

To estimate the cost-effectiveness of the TLA device compared with placebo, we undertook a complete-case analysis including only participants who had complete cost and EQ-5D-5L utility information. We carried out an incremental analysis, with the mean cost difference between the TLA device and placebo divided by the mean QALY difference to give the incremental cost-effectiveness ratio (ICER). The non-parametric percentile method75 used to calculate the CI around this ratio used 1000 bootstrap estimates of the mean cost and QALY differences. We used the cost-effectiveness acceptability curve to show the probability that the TLA device is cost-effective at 1 year for different values of the NHS’s willingness to pay for an additional QALY. 74 Cost-effectiveness was estimated both from a NHS and from a societal perspective.

Reference case analysis, including imputation of missing data

Descriptive analyses revealed that 79% (380/484) of observations in the placebo group had complete cost and EQ-5D-5L data compared with 76% (360/476) of observations in the treatment arm. Logistic regression models comparing missing indicator baseline variables with experiencing a severe exacerbation showed that having missing cost data was significantly negatively associated with age (p < 0.031) and experiencing a severe exacerbation (p < 0.001). Missing utility data were also negatively associated with age (p < 0.079) and experiencing a severe exacerbation (p < 0.001). This suggests that the data were not missing at random, as is assumed in the complete-case analysis. This may lead to bias in analyses, as participants with complete data may vary from those with missing data, and to a loss of statistical power, as missing data reduce the available sample size. 76

As the analysis of the LASER trial data was undertaken on an ITT basis, multiple imputation was used to impute missing cost and utility values for the reference case analysis (i.e. analysis of all participants regardless of whether complete data were available or not). 67,68 As is recommended best practice, imputation was implemented separately by randomised treatment allocation. 77 Costs were imputed at the most disaggregated level at which the model would converge. However, it was not possible to impute individual costs for ambulance use and hospitalisation costs. As a result, we imputed values for the combined total hospital care costs category (including outpatient, A&E, ambulance and hospitalisation costs). The only exception to this was for the five participants who withdrew from the study following randomisation (and, in addition, who withdrew consent for their data to be used for analysis). As we were unable to use any of their baseline characteristics to impute missing outcome data, we imputed missing cost and utility data for these participants using the average estimates for cost and utility, stratified by treatment arm.

Rather than imputing missing responses for each of the five domains in the EQ-5D-5L, we imputed the overall EQ-5D-5L utility score. 78 The imputation of costs and utilities was conducted using predictive mean matching (i.e. imputation of data from similar participants with complete data) to account for the skewed nature of both cost and utility data. Imputation was conducted using the trial minimisation criteria (i.e. clinical site, age, sex, baseline exacerbation rate, use of OCSs and an indicator of pre-bronchodilator FEV₁ of < 50%) as well as number of exacerbations at follow-up. Following the rule of thumb that the number of imputations should be at least equal to the percentage of incomplete cases (originally 21% imputations and 24% incomplete cases),79 we generated 30 replacement values for each missing case, generating 30 imputed data sets.

Using the Stata (version 14) ‘mi estimate’ command, we obtained mean and standard deviation (SD) estimates of costs and utilities across participant groups using the ‘mi estimate: reg’ ordinary least squares regression command.

Markov model

The Markov model was developed in Microsoft Excel^® version 16.0 (Microsoft Corporation, Redmond, WA, USA) to follow a hypothetical, homogeneous cohort over 40 years and consists of four distinct health states – (1) daily symptoms, (2) severe exacerbations, (3) severe exacerbation-related death and (4) death by other causes – as shown in Figure 13. Participants begin in the model suffering from daily symptoms with a probability of either experiencing an exacerbation or dying from all other causes. Participants who experience an exacerbation can recover and move back to the daily symptoms state or suffer another exacerbation. They also have a small probability of an exacerbation being fatal or dying from other causes. The model has a cycle length of 3 months, mimicking the follow-up periods in the LASER trial. In the model, the cohort starts at an age of 47 years, which is the mean age of participants in the LASER trial. QALYs and costs incurred after the first year are discounted at an annual rate of 3.5%.

FIGURE 13.

Markov model.

Model parameters

The parameters in the Markov model are described in the following sections.

NHS perspective

Information on 3-month NHS resource use and unit costs, inpatient health-care unit costs and 3-month NHS costs collected at each follow-up visit is presented in Tables 40–42, respectively. Table 35 presents the mean NHS resource use estimates over the 12 months after randomisation. The only statistically significant reductions in NHS resource use patterns in the treatment group compared with the placebo group over the 12 months after randomisation were in the numbers of general practice consultations (both surgery visits and telephone consultations), occupational therapist visits and psychologist visits; over the course of the year, participants in the TLA device group had 0.94 less general practice surgery consultations (95% CI –1.68 to –0.19; p = 0.014), 0.64 less general practice telephone consultations (95% CI –1.10 to –0.19; p = 0.005), 0.37 less occupational therapist visits (95% CI –0.95 to –0.22; p = 0.002) and 0.43 less psychologist visits (95% CI –0.76 to –0.23; p < 0.001) than participants in the placebo group. Conversely, the TLA device group had statistically significantly more physiotherapy visits on average than the placebo group (0.42, 95% CI 0.04 to 0.80; p = 0.032).

The cost of all this NHS resource use is shown in Table 20. Excluding the costs of the TLA device, over the 12 months after randomisation, participants in the placebo and TLA device groups incurred mean annual NHS costs of £4188 (SD £5004) and £4182 (SD £6091), respectively (p = 0.995). In both groups, the main cost drivers were inpatient stays (£1575 in the placebo group and £1385 in the TLA device group), which accounted for around one-third of all NHS costs in both groups.

Costs of asthma prescription medications for both groups were also considerable: £1212 (SD £1255) in the placebo group (accounting for 29% of total NHS costs) and £1513 (SD £3047) in the TLA device group (accounting for 36% of total NHS costs). A total of seven participants were prescribed omalizumab (four in the placebo group and three in the TLA device group; p = 0.718).

After taking into account the costs of the TLA device, there was a statistically significant difference in total NHS costs at 12 months. Participants in the TLA device group incurred mean costs of £6296 (SD £6101) whereas those in the placebo group incurred mean costs of £4188 (SD £5004) (mean difference £2108, 95% CI £368 to £3849; p = 0.018).

NHS perspective

A total of 78 participants in the placebo group and 77 participants in the TLA device group had complete follow-up data for both quality of life (QALYs) and NHS costs. As shown in Table 22, the mean total NHS costs for these complete quality of life and NHS cost cases, including the cost of the TLA device, were £4171 (SD £5051) for the placebo group and £6360 (SD £6167) for the TLA device group, a mean difference of £2189 (95% CI £401 to £3977; p = 0.017). The mean QALY gain was 0.68 in the placebo group and 0.72 in the TLA device group, a mean difference of 0.04 (95% CI –0.02 to 0.10). The incremental cost per QALY gained when the TLA device was compared with placebo was therefore £49,685.

At the conventional £20,000-per-QALY-gained threshold, the TLA device would have a probability of being cost-effective of 0.15 from a NHS perspective (Figure 14). Only at a willingness-to-pay threshold of £50,000 per QALY gained would the probability of the TLA device being cost-effective be > 0.5 from a NHS perspective.

FIGURE 14.

Cost-effectiveness acceptability curve from a NHS perspective.

Societal perspective

A total of 74 participants in each participant group had complete follow-up visit data for both quality of life (QALYs) and societal costs (NHS and non-NHS resource use). As shown in Table 23, the difference in mean total societal costs between the groups was not significant: £5706 (SD £7207) in the placebo group and £7278 (SD £6986) in the TLA device group, a mean difference of £1572 (95% CI –£820 to £3966; p = 0.196). The mean QALY gain was 0.67 in the placebo group and 0.72 in the TLA device group, a mean difference of 0.05 (95% CI –0.01 to 0.11).

From a societal perspective, the incremental cost per QALY gained when the TLA device was compared with placebo was £29,831. At the conventional £20,000-per-QALY-gained threshold, the TLA device would have a probability of 0.36 of being cost-effective (Figure 15). Only at a willingness-to-pay threshold of > £30,000 per QALY gained would the probability of the TLA device being cost-effective be > 0.5 from a societal perspective.

FIGURE 15.

Cost-effectiveness acceptability curve from a societal perspective.