The clinical and cost effectiveness of treatments for idiopathic pulmonary fibrosis: a systematic review and economic evaluation

Pulmonary rehabilitation

Individuals with ILD experience many of the same symptoms as those with COPD, commonly shortness of breath, fatigue and reduced exercise tolerance. It would, therefore, seem a priority to address these symptoms. Pulmonary rehabilitation is an established evidence-based intervention for individuals with COPD, with the precise aims of reducing the physical and emotional impact of the disease on the individual. 37 It is well documented that rehabilitation improves exercise capacity, dyspnoea and QoL. 37 The delivery of rehabilitation to individuals with COPD is advocated in national guidelines; however, until the 2013 NICE guidelines this had not been extended to IPF. 31 Many individuals are, therefore, currently not offered pulmonary rehabilitation. Further details about the intervention itself can be found in Description of technologies under assessment.

The referral to rehabilitation can be instigated by primary or secondary care physicians; in most cases, this would be after a detailed assessment and optimisation of treatment from a physician with a special interest in IPF. Prior to commencing a course of rehabilitation, medical management should be optimised; this may include the provision of supplemental oxygen. The most advantageous time to refer is unclear in the evidence, although it is anticipated that a prompt referral is more likely to be beneficial before an individual is too disabled by the disease.

Lung transplant

Lung transplantation is the only treatment that has been shown to improve survival in patients with IPF. The 5-year survival of IPF patients post transplant is in the region of 40–50%, which compares with an overall 52% 5-year survival rate. 38 As indicated in the NICE guidelines,31 treating clinicians should carefully consider whether or not their patient might be eligible and refer potentially suitable patients to a recognised transplant centre for a full evaluation. IPF is the most common ILD referred for lung transplantation throughout the world. Data from the International Society for Heart and Lung Transplantation (ISHLT) collected from 1 January 2011 to 30 June 2012 show IPF to be the third most common diagnosis leading to lung transplant throughout Europe, representing 17% of all lung transplants. 38 Data from the NHS Blood and Transplant service reveals that 17% (numbering between 23 and 33 patients per year) of all lung transplants in the UK between April 2009 and March 2013 were for fibrosing lung disease (NHS Blood and Transplant, 26 February 2013, personal communication). It is anticipated that the majority of patients in this category had IPF. This differs significantly from the current experience in the USA, where IPF is now the most common indication for lung transplantation (35% of transplants). 39 In the USA, the allocation of organs is governed by the United Network for Organ Sharing (UNOS) system which calculates a lung allocation score (LAS). This calculates a post-transplant survival score and a waitlist survival score to give an overall LAS. Patients with a higher score are prioritised on the basis of urgency.

As IPF is a disease of older adults, many patients may not be considered for lung transplantation on the basis of their age. Most centres use age > 65 years as a relative, but not an absolute, contraindication to lung transplant. Data from ISHLT from 2011 to the first half of 2012 reveals that, worldwide, 15.4% of patients who received a transplant were > 65 years of age. 38 However, again, there was a considerable difference between European and American practice, with 4.4% of European recipients being > 65 years, compared with 25.3% of North American recipients. Recent data from the USA suggest that post-transplant mortality is no higher in patients > 70 years when compared with those in the 60- to 69-year-old age group. 40

People considered for transplant referral should be free of malignancy for at least 2 years and have no evidence of significant coronary artery disease, heart failure or any significant disease in a vital organ. They should be not be infected with human immunodeficiency virus (HIV), chronic hepatitis B or C infection or be dependent on tobacco and other substances. Similarly, a lack of social support and significant psychological problems are contraindications to lung transplant. Poor nutritional status and being significantly overweight [body mass index (BMI) > 30 kg/m²] or underweight is another important contraindication. 41

Data from the UNOS show that overall mortality post lung transplant is 22% higher in patients with a BMI > 30 kg/m². 42 Low-dose prednisolone use (< 0.3 mg/kg/day) is not considered a contraindication.

In patients who have no contraindications to lung transplant, the following specific criteria have been proposed to guide a timely referral in IPF:41

1. radiological or pathological evidence of UIP

2. diffusion capacity/transfer factor for carbon monoxide < 40% predicted

3. a reduction of 10% in the FVC over a 6-month follow-up period

4. a reduction in oxygen saturation below 88% on 6MWT

5. the presence of honeycombing on a HRCT scan (fibrosis score > 2). 43

However, as more patients with IPF than any other condition die on lung transplant waiting lists, the clinician should have a low threshold to refer a deteriorating patient even if they do not yet meet the above criteria. 44 Similarly, many experts would advocate early referral for a deteriorating patient even if the radiological and/or pathological findings are not entirely consistent with a UIP pattern, as evidence of deterioration is a powerful marker of a poor prognosis.

Traditionally, single-lung transplant has been preferred over double-lung transplant in IPF, with the latter being preferred for patients with cystic fibrosis or conditions associated with chronic infections. Some centres have suggested that long-term survival may be better post double-lung transplant,38,45 while others report little difference. 46,47 This remains a point of considerable debate and given the lack of availability of organs for donation it is likely that single-lung transplant will continue to be widely used. Exceptions to this may be in younger patients or those with a nidus of infection in the native lung.

Palliative care

As stated, patients with IPF have a poor prognosis, although determining individual prognosis is more difficult than in, for instance, lung cancer, and treatment options are limited. 27,48 Lee and colleagues49 describe three ‘pillars’ of care: disease-centred management, symptom-centred management, and education and self-management.

American Thoracic Society guidelines suggest that palliative care principles should be initiated when a patient with respiratory disease becomes symptomatic. 48 This may be particularly helpful in IPF as diagnosis can be variable and there may be rapid and unexpected deterioration resulting in death. 27 The guideline of the British Thoracic Society recommends that BSC should be considered a specific and important treatment strategy in all patients with IPF. 50 The recent NICE guideline31 states that BSC should be initiated at the point of diagnosis. Best supportive, or palliative, care is a proactive approach to symptomatic treatment and may include oxygen therapy, pulmonary rehabilitation, opiates, antireflux therapy, the withdrawal of steroids and other immunosuppressants, early recognition of terminal decline and liaison with palliative care specialists. 50 This should also allow patients with IPF the opportunity to consider their expectations, goals and preferences, but there is a reluctance of professionals to broach these difficult topics. 27 This might be one explanation as to why the majority of IPF patients in a recent study had no palliative care input in their last year of life, despite these recommendations regarding the early integration of palliative care. 24

The NICE 2013 guideline suggests that if a patient with IPF has breathlessness on exertion, ambulatory oxygen should be considered. 31 Where there is breathlessness at rest, consideration should also be given to benzodiazepines and/or opioid therapy.

Disease-modifying treatments

Cyclophosphamide and azathioprine are immunosuppressant treatments which may also suppress inflammation and have been used in some patients with IPF. Both cyclophosphamide and azathioprine are used, in some cases in combination with the corticosteroid prednisolone. The ATS/ERS 2011 consensus guidance suggests that these are not used routinely in IPF. 10 The NICE guideline states that where the combination of azathioprine and prednisolone is already being used, there should be consideration of gradual withdrawal of these treatments. 31 In the UK, the use of cyclophosphamide and prednisolone is mostly restricted to the treatment of acute exacerbations, although not widely used.

N-acetylcysteine (NAC) is a precursor to the antioxidant glutathione, which may be reduced in the lungs of patients with IPF. It is an antioxidant therapy that does not have specific marketing authorisation for use in IPF, and it is readily available in health-food shops and over the internet. It can be used alone or, as has been done in the past, in combination with prednisolone and azathioprine (referred to as triple therapy). The ATS/ERS 2011 consensus guidance10 suggests that NAC (alone or in triple therapy) is not used in the majority of patients but it may be a reasonable choice in a minority. The 2013 NICE guideline31 suggests that it may be used alone after discussion of the risks of the treatment, and that if used in combination with azathioprine and prednisolone, these latter two treatments should be gradually withdrawn.

Pirfenidone is an orally bioavailable synthetic molecule with antifibrotic and anti-inflammatory properties. The mechanism of action at the molecular level is not fully understood; however, pirfenidone has marketing authorisation for use in mild to moderate IPF in Europe. Pirfenidone was not included in the scope of the NICE 2013 clinical guideline;31 however, it has been assessed by NICE under the technology appraisals programme. Final guidance to the NHS issued in 201323 states that pirfenidone is recommended as an option where an individual patient has a FVC between 50% and 80% predicted. This is the case only if the manufacturer of pirfenidone provides the treatment at a previously agreed (undisclosed) discount price.

Symptomatic treatments

Prednisolone is a corticosteroid which can be used to reduce inflammation. Most widely used in the past before the nature of IPF was fully understood and the diagnosis more specific, prednisolone is now used predominantly for symptom control, including for acute exacerbations. 10

Pulmonary rehabilitation is conventionally offered as a package of supervised exercise and education over a 6- to 8-week period by a multidisciplinary team. The ATS/ERS 2011 consensus guidance10 suggests that this is used in the majority of patients but this may not be a reasonable choice for some. In the UK, the clinical guideline from NICE31 recommends that individuals should be assessed for pulmonary rehabilitation where appropriate.

Most experience with pulmonary rehabilitation comes from treating COPD, although the adoption of pulmonary rehabilitation in the UK for IPF is becoming more widespread. The process of rehabilitation commences with a detailed assessment; it is most likely that a detailed study of lung function will have already characterised the patient. The rehabilitation assessment should include an evaluation of health-related quality of life (HRQoL), levels of anxiety and depression, and exercise capacity as a minimum. Additional measures that might be considered could be measures of peripheral muscle strength, self-efficacy and physical activity, but these are not widely reported for IPF. HRQoL measures are those commonly employed for individuals with COPD, for example the St George’s Hospital Respiratory Questionnaire (SGRQ)51 or the chronic respiratory disease questionnaire. 52 The exercise test is usually conducted outside the laboratory and in the UK is commonly the 6MWT53 or the incremental shuttle walking test. 54 During and immediately after an exercise test it is important to record levels of oxygen saturation, and, in discussion with the patient, make a decision about the use of supplemental oxygen while exercising. Repeat testing with oxygen might be expected to correct the desaturation to some degree and/or improve the level of perceived breathlessness.

The usual format of rehabilitation is one of educational and supervised exercise. The groups are not specifically designed for individuals with IPF and, therefore, the education sessions may not all be entirely appropriate. Generally, the educational topics include medicines management, disease pathology, chest clearance techniques, breathing control, healthy eating, travelling with a lung disease, inhaler techniques, relaxation, stress management and energy conservation. The exercise programme is a combination of strength and aerobic exercises, prescribed and progressed on an individual basis. Bouts of purposeful exercise should also be completed at home, to accumulate five sessions per week of aerobic activity. On completion of a course of rehabilitation, it is recommended that a repeat assessment be conducted to reinforce the benefits of rehabilitation.

Maintaining the benefit of rehabilitation is challenging; however, there are opportunities to continue with regular supervised exercise in the community with exercise specialists trained to manage participants with chronic respiratory disease.

Oxygen therapy can either be used as a long-term therapy or be taken as a supplement after exertion. The ATS/ERS 2011 consensus guideline10 suggests that long-term oxygen therapy is used in patients with IPF and clinically significant resting hypoxaemia, and the NICE 2013 guideline31 also refers to the use of oxygen for breathlessness on exertion as discussed above.

Palliative care/BSC should be considered from the outset as an adjunct to disease-focused care. This may include corticosteroids, oxygen therapy, treatments for other symptoms such as cough, advanced directives and end-of-life care issues including hospice care.

Treatments for specific symptoms

Sildenafil is a phosphodiesterase-5 inhibitor with marketing authorisation for use in pulmonary hypertension and it has been used as a treatment for IPF. There are currently no recommendations from the ATS/ERS consensus committee of the 2011 guidance10 and in the NICE guideline sildenafil was not recommended as a treatment for modifying IPF owing to uncertainty around both benefits and harms. 31

Bosentan is a dual endothelin receptor A and B antagonist that has marketing authorisation for pulmonary hypertension. Endothelin is a vasoconstrictor and growth factor that is involved in the pathogenesis of pulmonary hypertension. The 2011 ATS/ERS10 consensus guidance suggests that these are not used based on the potential risks and costs of therapy. The NICE clinical guideline concurs with this. 31

Thalidomide is a potent immunomodulatory, anti-inflammatory and antiangiogenic drug, and may be a potential therapy in IPF for cough. 55 Thalidomide does not have a UK marketing authorisation for IPF; however, the NICE clinical guideline31 states that thalidomide may be considered in a patient with intractable cough.

Treatments undergoing assessment in UK participants

BIBF 1120 or nintedanib is an intracellular inhibitor of tyrosine kinases and is currently under evaluation in phase III randomised controlled trials (RCTs) as a potential therapy for IPF. It has received orphan status from the European Medicines Agency in 2013 (orphan status provides some favourable marketing incentives to manufacturers). 56

Survival

At 12 months, four participants in each treatment group of the Raghu and colleagues70 trial had died. Survival estimates based on an additional observational period (presumed duration of 9 years) were compared between those treated with azathioprine and prednisolone, and those treated with placebo and prednisolone, and were reported in an unadjusted and age-adjusted data set (Table 7). The HR of survival in the unadjusted set was 0.48 (95% CI 0.17 to 1.38), a non-statistically significant effect (p = 0.16). In the age-adjusted analysis, the azathioprine group were seen to show a survival advantage over those treated with placebo (HR 0.26, 95% CI 0.08 to 0.88), suggesting a 64% reduction in the risk of death with azathioprine; however, it is unclear if this is statistically significant (two p-values are reported: p = 0.02 using a large sample approximation and p = 0.05 using the randomisation test). Moreover, the likely inclusion of participants with NSIP may explain this treatment effect, and the small sample size and potential risk of bias in this study should be considered when interpreting these data.

Forced vital capacity

There was no significant difference in FVC per cent predicted at 12 months (difference of 4.8%; p = 0.87) or in the proportion of participants categorised as ‘improved’ (Table 8 and Figure 2).

FIGURE 2.

Forest plot of mean change in per cent predicted FVC/VC from trials of pharmacological interventions.

Diffusing capacity of the lung for carbon monoxide

There was no significant difference in DL_CO per cent predicted at 12 months (difference 6.4%; p = 0.70), or in the proportion of participants categorised as ‘improved’ (Table 9 and Figure 3).

FIGURE 3.

Forest plot of mean change in per cent predicted DL_CO from trials of pharmacological interventions.

Other secondary outcomes

Raghu and colleagues70 also report the partial pressure of oxygen as an outcome measure, the results of which can be found in Appendix 5.

Adverse events

Adverse events reported in the Raghu and colleagues70 trial can be seen in Table 10. Overall, there were 28 adverse events in the azathioprine-treated group and 25 in the placebo-treated group. More participants treated with placebo reported a subjective adverse event (gastrointestinal or neuropsychiatric) than those treated with azathioprine, and more azathioprine-treated participants had vertebral fractures. All other adverse events were reasonably well matched between the two groups, although this is based on observation of the data only.

Summary

One small study with an unclear risk of bias provided data on the effectiveness of azathioprine and prednisolone compared with placebo and prednisolone. This trial may include participants who would be diagnosed with NSIP according to current criteria. Treatment with azathioprine and prednisolone led to an improvement in survival compared with placebo and prednisolone when this was age-adjusted, but did not lead to improvements in measures of lung function. Azathioprine did not appear to lead to significant adverse events when compared with placebo over the 12-month duration of this study. Caution is required in the interpretation of these data in the light of the small sample size, potential risk of bias and possible inclusion of NSIP.

Forced vital capacity

The annual rate of decline in FVC was the primary outcome in this study. After 54 weeks this was lower (0.06 l) in the BIBF 1120 300 mg/day-treated group than in the placebo-treated group (0.19 l). The difference-between-groups (0.13 l) comparison was not statistically significant (p = 0.06) using the prespecified closed testing procedure for multiplicity. With hierarchical testing this was statistically significant (p = 0.01). A statistically significant difference was seen between the absolute change in FVC at 54 weeks between BIBF 1120 300 mg/day and placebo (p = 0.001), and the absolute change in per cent predicted FVC at 54 weeks between the two groups (p < 0.001), as seen in Table 11. Figures 2 and 4 show, graphically, the mean change in per cent predicted FVC and in FVC litres, respectively. The proportion of participants with a reduction in mean FVC of > 10% or 200 ml was lower in the BIBF 1120 300 mg/day group than that seen in the placebo group (23.8% vs. 44%, respectively). This was shown to be statistically significantly different between groups (p = 0.004) and is displayed in Figure 5. It is unclear whether this was calculated as an absolute or relative change in FVC.

FIGURE 4.

Forest plot of mean change in FVC/VC from trials of pharmacological interventions.

FIGURE 5.

Forest plot of FVC/VC decline of ≥ 10% from trials of pharmacological interventions. M–H, Mantel–Haenszel.

Diffusing capacity of the lung for carbon monoxide

The Richeldi and colleagues71 publication stated that it measured DL_CO as an outcome measure; however, no data were reported at follow-up for this outcome.

Six-minute walk test

The Richeldi and colleagues71 publication stated that it measured the 6MWT as an outcome measure; however, no data were reported at follow-up for this outcome.

Other secondary outcomes

Richeldi and colleagues71 also reported the oxygen saturation and the TLC as outcome measures, the results of which can be found in Appendix 5.

Acute exacerbations

The incidence of acute exacerbations was statistically significantly lower in the BIBF 1120 300 mg/day-treated participants than in the placebo-treated participants in the Richeldi and colleagues trial (Table 12). 71 The rates were 15.7 per 1000 patient-years in the placebo group and 2.4 per 1000 patient-years in the BIBF 1120 300 mg/day-treated group (p = 0.02). This equated to a RR of 0.16 (95% CI 0.03 to 0.70) in favour of treatment.

St George’s Hospital Respiratory Questionnaire

The total score, symptoms score, activity score, and impacts score on the SGRQ were reported as change from baseline scores at follow-up (54 weeks) in the Richeldi and colleagues71 trial. As can be seen in Table 13, the change scores were statistically significantly different between the BIBF 1120 300 mg/day group and placebo group for the change in total score, symptom score and activity score. No statistically significant effect of BIBF 1120 was seen on the impacts score of the SGRQ.

The authors stated in the publication that the minimally clinical important difference (MCID) in total score on the SGRQ is 4 points, and they analysed the proportion of participants improving by at least 4 points. The proportions improving by at least four points on the SGRQ total score was 29.1% in the BIBF 1120 300 mg/day-treated group compared with 16.1% in the placebo-treated group (p = 0.03). It should be noted that elsewhere in the publication it states that the MCID is 5–8 points. For a fuller description of the SGRQ, see Chapter 4, Systematic review of health-related quality of life studies.

Deaths

The number of deaths from respiratory causes was reported in the Richeldi and colleagues71 trial. These were two (2.4%) in the BIBF 1120 300 mg/day arm and eight (9.2%) in the placebo arm (Table 14). The paper reports that the difference between groups was not statistically significant (p = 0.06).

Deaths from any cause were more similar between the two groups, with seven (8.2%) participants dying from any cause in the BIBF 1120 300 mg/day treatment arm and nine (10.3%) in the placebo-treated arm (p-value not reported, paper stated not significant).

Adverse events

The study reports on a range of adverse events (see Table 14). The proportion with any adverse event was similar across both groups, with 94.1% of those treated with BIBF 1120 300 mg/day and 90.6% of the placebo-treated group having an adverse event recorded. On observation of the data, more participants treated with BIBF 1120 300 mg/day had diarrhoea, nausea, vomiting, abdominal pain, headache or decreased appetite (see also specific gastrointestinal events in Table 14), and more participants treated with placebo had upper respiratory tract infection (URTI) and nasopharyngitis. The rate of severe adverse events and serious adverse events appears to be similar between the two groups, as is the rate of hospitalisation due to adverse events (see Table 14). More fatal adverse events occurred in the placebo group than the BIBF 1120 300 mg/day-treated group. Adverse events categorised as drug-related occurred in 64.7% of BIBF 1120 300 mg/day-treated participants, compared with 29.4% of placebo-treated participants.

Summary

BIBF 1120 at 300 mg/day led to more favourable outcomes than placebo on some measures of the FVC, rate of acute exacerbations and the number of deaths. The primary outcome in this study was the annual rate of decline in FVC, and the observed benefit did not reach statistical significance. QoL as reported on the SGRQ was generally favourable in those treated with BIBF 1120 300 mg/day. Serious adverse event rates and hospitalisation rates appear to be similar between the two groups. BIBF 1120 300 mg/day led to more gastrointestinal events than placebo. Treatment with placebo led to more respiratory events. Risk of bias was low in this study. This study also randomised participants to three other treatment groups according to the dose of BIBF 1120 and results for these can be seen in Appendix 5.

Progression-free survival

In the Raghu and colleagues IPFCRN 2012 trial,68 the outcome of time to death or disease progression, defined as a relative drop in FVC of > 10%, was reported. Although not labelled as such, this has been considered as PFS as this definition has been applied in other included studies (see Assessment of effectiveness of pharmacological interventions for mild to moderate idiopathic pulmonary fibrosis, Pirfenidone) to mean PFS. The HR for the estimate of PFS between the triple-therapy group and the placebo group was not reported in the trial. The publication states that there was no statistically significant difference in the risk of death or disease progression between groups; however, no p-value was reported to support this. The Kaplan–Meier plot was published in the study article.

Progression-free survival was not reported as an outcome in the Demedts and colleagues67 trial comparing triple therapy with double therapy, or the trial by Homma and colleagues69 comparing inhaled NAC with control.

Time to death/time to death or hospitalisation

The HR for time to death or the HR for time to death or disease progression was not reported in the Raghu and colleagues IPFCRN 2012 trial. 68 The Kaplan–Meier plots were published in the study article.

In the same study,68 the HR for time to death or hospitalisation was reported to be 12.11 (95% CI 2.83 to 51.85, p < 0.0001), indicating a shorter time to death or hospitalisation in the triple-therapy-treated population.

Outcomes relating to time to death or hospitalisation were not reported in the Demedts and colleagues67 trial comparing triple therapy with double therapy, or the trial by Homma and colleagues69 comparing inhaled NAC with control.

Discussion of all-cause mortality, respiratory mortality and any-cause mortality or hospitalisation is provided below.

Forced vital capacity

Demedts and colleagues67 compared triple therapy with double therapy and assessed the VC in litres and the VC per cent predicted. At 12 months it can be observed that the change in VC was lower in the triple-therapy group than the double-therapy group; however, no statistical analysis of the difference between groups was undertaken (Table 15). Figure 4 presents results graphically where it can be seen that treatment with triple therapy is approaching statistical significance on this outcome (95% CI touches the line of no effect). A statistically significant difference in the mean VC litres between groups in favour of triple therapy at 12 months was observed (difference 0.18; p = 0.02) when using the least squares (LS) mean method. The final value per cent predicted VC was also shown to be statistically significantly different between groups at 12 months when using the LS mean method (with the LS mean the difference was 4.79%; p = 0.02), but not when calculating the mean difference using the unadjusted data (see Table 15 and Figure 2, calculated by reviewer). The mean change from baseline in VC per cent predicted was not reported. All of the analyses were undertaken using the last observation carried forward (LOCF) approach.

After a mean follow-up of 32 weeks of treatment, no difference in mean change in FVC litres between triple therapy and placebo in the Raghu and colleagues IPFCRN 201268 trial was observed (p = 0.85) (see Table 15 and Figure 4).

Inhaled NAC therapy did not confer a statistically significant benefit over no-treatment control on FVC litres in the Homma and colleagues69 trial (difference 60 ml; p = 0.27) (see Table 15 and Figure 4). In this study, data were dichotomised into two categories: those with a decline of < 10% FVC litres, and those with a decline of > 10% FVC litres. It is uncertain if these were a priori groups or post hoc. Data were not presented for the proportions within the treated and placebo groups; however, the differences between groups in these two categories were presented (see Table 15), but these did not reach the threshold for statistical significance. The results of this study need to be interpreted cautiously given the uncertain risk of bias identified when assessing the study quality.

Diffusing capacity of the lung for carbon monoxide

All three included trials using NAC therapy alone or in combination recorded DL_CO (Table 16, see Figures 3 and 6). This was reported as the mean change in mmol/minute/mmHg and per cent predicted DL_CO in the Demedts and colleagues67 trial of triple therapy versus double therapy and the Homma and colleagues69 trial of inhaled NAC versus control, and DL_CO mmol/minute/mmHg only in the Raghu and colleagues IPFCRN 201268 trial of triple therapy versus placebo. Triple therapy led to a smaller decline in DL_CO mmol/minute/mmHg after 12 months than double therapy in the Demedts and colleagues67 study; however, statistical analysis was not undertaken on the difference between groups on these change scores. The mean value of DL_CO (see Table 16) was statistically compared between groups and showed a statistically significant difference in favour of triple therapy (difference 0.75, 95% CI 0.27 to 1.23; p = 0.003) when using the LS mean method. On the per cent predicted DL_CO, a similar pattern was observed in the mean values, where there was a difference between groups of 5.08% (95% CI 1.17% to 8.99%, p = 0.01).

In the Raghu and colleagues IPFCRN 201268 study, the difference in DL_CO change from baseline (corrected for haemoglobin) was not statistically significant (difference 0.06; p = 0.93) between triple therapy and placebo. Homma and colleagues69 did not report data on these outcomes; they reported only that there were no statistically significant differences between groups at 48 weeks.

Six-minute walk test

The 6MWT was reported in two of the RCTs evaluating NAC68,69 (Table 17). The difference between groups in the change in the distance walked on the 6MWT favoured placebo in the Raghu and colleagues IPFCRN 201268 trial, although the difference was not statistically significant between the two groups. The Homma and colleagues69 study comparing inhaled NAC with placebo reported that the distance walked was not statistically significantly different between groups, although the data were not provided in the study publication.

Lowest SpO₂ during the 6MWT was also stated to be an outcome in the Homma and colleagues69 trial; however, data were not reported. The publication states there were no statistically significant differences between those treated with inhaled NAC and those treated as controls (p-value not reported).

Other secondary outcomes

Other outcomes reported in these studies include TLC; clinical, radiological, physiological (CRP) scoring system; exercise performance; and HRCT outcome. The results can be found in Appendix 5.

Dyspnoea

Dyspnoea was reported as an outcome in the three included RCTs investigating NAC therapy alone or in combination (Table 18). Different measures of dyspnoea were used. In the Demedts and colleagues trial67 of triple therapy compared with dual therapy, a dyspnoea score which measured breathlessness on a scale from 0 to 20 was used. On this measure, higher scores indicate worse dyspnoea. In the Raghu and colleagues IPFCRN 2012 trial68 of triple therapy compared with placebo, the USCDSBQ scale was used. This scale rates shortness of breath on a scale from 0 to 120, with higher scores indicating greater breathlessness. The Homma and colleagues69 study reported the proportion of participants with improved or stable dyspnoea (definition not provided). Dyspnoea scores were not statistically significantly different between triple therapy and double therapy in the Demedts and colleagues trial (p = 0.65)67 or between triple therapy and placebo in the Raghu and colleagues IPFCRN 2012 trial68 (see Table 18). The proportion of participants with improved or stable dyspnoea were similar in the inhaled NAC therapy arm and the control arm in the Homma and colleagues trial69 (p = 1.00).

Acute exacerbations

The rate of acute exacerbations were reported in the Raghu and colleagues IPFCRN 201268 trial comparing triple therapy with placebo (Table 19). There were five (6%) acute exacerbation incidents in the triple-therapy group and no incidents in the placebo group. Analysis of the difference between groups was not reported. Acute exacerbations were not reported as an outcome in the Homma and colleagues69 study; however, the number of participants excluded from the analysis because of acute exacerbations was recorded. These were four in the placebo group and one in the inhaled NAC group.

Quality of life

Demedts and colleagues67 and the Raghu and colleagues IPFCRN 201268 trial assessed QoL using the SGRQ. However, results were not reported by Demedts and colleagues;67 therefore, no conclusions can be made regarding QoL comparing triple therapy with double therapy. As noted previously, the SGRQ has a total score and three subscales; each scale ranges from 0 to 100, with lower scores indicating better QoL. The Raghu and colleagues IPFCRN 201268 group reported all scales of the SGRQ and also assessed QoL by the Short Form questionnaire-36 items (SF-36) and the European Quality of Life-5 Dimensions (EQ-5D). The SF-36 measures functional health and well-being scores on eight scales that correlate with two aggregate scores (physical and mental health). Each score ranges from 0 to 100, with a higher score indicating better function. Scores are normalised to a mean (SD) of 50 (10). The EQ-5D measures general QoL on a self-report questionnaire on a scale of 0 to 1.00 (higher score indicates better QoL) and on a VAS (described in the Raghu and colleagues IPFCRN 2012 trial as a thermometer score) with a range of 0 to 100 (higher indicates better QoL). For a fuller description of the SF-36 and EQ-5D, see Chapter 4, Systematic review of health-related quality of life studies. Homma and colleagues did not report QoL as an outcome.

Full results can be seen in Table 20. In the Raghu and colleagues IPFCRN 201268 trial, the change in SGRQ scores in those treated with triple therapy was lower (better) than in those treated with placebo; however, the differences were not statistically significant on total score or two of the subscales (only the symptoms score was statistically significantly different; p = 0.014). On the SF-36 in this study, the difference between groups on the physical score was not statistically significant (p = 0.58). On the SF-36 mental score, the change from baseline score was smaller (better) in the triple-therapy group than the placebo group, and the difference was statistically significant (p = 0.027). On the two EQ-5D indices, the difference in changes in scores between the two groups was not statistically significant (see Table 20).

Adverse events and mortality

A range of adverse events were reported by Demedts and colleagues67 and these are summarised in Table 21. The study report that none of the differences between the study groups were significant, except for those relating to bone marrow toxicity, which occurred in 4% of participants receiving triple therapy and in 13% receiving double therapy; p = 0.03. Detailed descriptions of the types of events included in these categories were reported but have not been data extracted. There were seven deaths in the triple-therapy arm and eight in the dual-therapy arm (9% vs. 11%, respectively) of the Demedts and colleagues67 study and this was not statistically significant between the two groups (p = 0.69).

The Raghu and colleagues IPFCRN 2012 group report data on the rate of serious adverse events in the triple-therapy and placebo arms of their trial. 68 These can be seen in Table 22, where it can be seen that the event rate of any serious adverse event was 31% in the triple-therapy arm, compared with 10% in the placebo arm (p = 0.001). Respiratory system serious adverse events contributed the most to the difference observed, and were also seen to be statistically significantly higher in the participants in the triple-therapy arm, compared with the placebo arm (p = 0.03). All other serious events recorded in the study were not statistically significantly different between the two groups.

Specific adverse events with a significant between-group difference from the Raghu and colleagues IPFCRN 201268 trial can be seen in Table 23. Overall, there was a higher proportion of individuals with an adverse event in the triple-therapy arm than in the placebo arm (p = 0.09), with rates of general disorder, skin disorder and renal and urinary system disorders also being recorded more frequently in the triple-therapy arm than in the placebo arm. All other adverse events, which were labelled as non-serious and had non-significant between-group differences, were reported by the study publication and can be seen in Appendix 5. The hospitalisation rates (all-cause) were also seen to be statistically significantly different between the two groups, with 23 (30%) of the triple-therapy participants and seven (9%) of the placebo participants being hospitalised during the 60 weeks of the study; p < 0.001.

The Raghu and colleagues IPFCRN 2012 trial68 reported mortality outcomes under a number of different categories (Table 24). All-cause mortality was statistically significantly greater in the triple-therapy-treated participants (10%) than the placebo-treated participants (1%); p = 0.01. A similar pattern was observed in the rates of respiratory mortality (9% triple therapy, 1% placebo; p = 0.02). HRs for estimated any-cause mortality and any-cause mortality or hospitalisation were also seen to favour placebo (see Table 24). Estimated any-cause mortality or at least a 10% decline in FVC was similar across the two groups.

Frequently recorded adverse events in the study by Homma and colleagues69 can be seen in Table 25. The paper states that there were no significant differences in the number of adverse events reported for the two groups. The severity of the events was less than grade 2. No adverse events were recorded in participants in the control group. There were some grade 1 and grade 2 adverse events (bacterial pneumonia, cough, sore throat and hypercholesterolaemia) in the inhaled NAC group.

Summary

Triple therapy did not confer any improvement on the FVC compared with placebo in one study but did confer an improvement on the VC when compared with double therapy in one other study. Inhaled single-therapy NAC did not confer a statistically significant benefit over a control intervention in a third included study. Secondary outcomes including DL_CO and 6MWT were reported with mixed results. There were no differences in rates of dyspnoea in the three studies. QoL did not appear to differ between participants treated with triple therapy when compared with placebo, and acute exacerbations were greater with triple therapy but not significantly. Triple therapy led to more deaths, serious adverse events and adverse events than placebo in one trial. In the other two studies, adverse events did not differ significantly between randomised groups. Of the three studies, two had a low risk of bias; however, the third study by Homma and colleagues69 had an unclear risk of bias and therefore results should be interpreted with caution.

Time to progression

Time to progression (TTP) was calculated in the two Capacity trials62 (Table 26). TTP was defined as the time to ‘worsening’ of IPF, with no further details provided. The HR for the estimated TTP was not statistically significant in the Capacity study 006 (HR 0.73, 95% CI 0.43 to 1.24; p = 0.248) or in the Capacity study 004 (HR 0.84, 95% CI 0.50 to 1.42; p = 0.515), with no benefit of pirfenidone seen. 62

Progression-free survival

Progression-free survival was defined in the two Capacity trials62 as time to confirmed ≥ 10% decline in per cent predicted FVC, ≥ 15% decline in per cent predicted DL_CO, or death. The HR for estimated PFS was not statistically significant in the Capacity study 006 (HR 0.84, 95% CI 0.58 to 1.22; p = 0.355) but was statistically significant in the Capacity study 004 [HR 0.64 (95% CI 0.44 to 0.95); p = 0.023], indicating a 36% reduction in the risk of death or disease progression with pirfenidone treatment (Table 27). 62

In the Taniguchi and colleagues65 trial, PFS was defined as death or a ≥ 10% decline in VC from baseline. In this study the HRs were not reported. The publication states that the HR for the high-dose pirfenidone group versus placebo was statistically significantly (p = 0.03) in favour of treatment (the HR for the low-dose pirfenidone group vs. placebo was not). There is an uncertain risk of bias in this study, which should be taken into account when interpreting these data.

Forced vital capacity

The two Capacity trials62 report the mean change in per cent predicted FVC at 72 weeks (see Table 27). In the Capacity 006 trial, the difference between groups (0.6%) was not statistically significant (p = 0.501). 62 In the Capacity 004 trial, the difference of 4.4% between groups was shown to be statistically significantly different (p = 0.001). 62 Combining these two trials in a random-effects meta-analysis (see Figure 2) demonstrated no statistical difference overall (mean difference 2.61, 95% CI –1.11 to 6.33; p = 0.17) and moderate statistical heterogeneity between the trials (I² = 45%). Taniguchi and colleagues65 reported the mean change in VC litres at 52 weeks. In this study the difference in change in VC between the high-dose pirfenidone group (1800 mg/day) and placebo was seen to be statistically significant (difference 0.07 litres; p = 0.04). In the RCT by Azuma and colleagues66 at 6 months the mean change in VC litres was not statistically significantly different between those treated with pirfenidone 1800 mg/day and those treated with placebo (difference 0.07; p = 0.10) (see Table 27 for full details). At 9 months a statistically significant difference between groups was seen, with the mean change in VC between groups being 0.10 litres (p = 0.04). Combining these two trials in a fixed-effects meta-analysis (see Figure 4) showed a statistically significant benefit of treatment with pirfenidone compared with placebo (mean difference 0.08, 95% CI 0.03 to 0.12; p = 0.0006) and no statistical heterogeneity between the trials (I² = 0%).

The FVC/VC outcomes (both per cent predicted FVC/VC and mean change in FVC/VC) from the four included pirfenidone studies were combined in a random-effects meta-analysis using the standardised mean difference (Figure 6). A statistically significant benefit of treatment with pirfenidone (standardised mean difference 0.24, 95% CI 0.06 to 0.41; p = 0.008) was found. Moderate statistical heterogeneity was present, as indicated by the I² test (45%). The uncertain risk of bias in the Azuma and colleagues66 and Taniguchi and colleagues65 studies should be noted when interpreting the effects shown.

FIGURE 6.

Meta-analyses of pirfenidone trials: FVC/VC mean change and per cent predicted mean change outcomes at end of study.

The Capacity trials62 also reported the proportion of participants with a decline in FVC of at least 10% (Table 28 and Figure 5). It is unclear if this was calculated as an absolute or relative change. In the Capacity 006 study, 23% of participants treated with pirfenidone and 27% of participants treated with placebo showed a decline of at least 10% in the per cent predicted FVC. Statistical analysis was undertaken on five categories of change in FVC per cent predicted in this study (severe decline ≥ 20%; moderate decline < 20% but ≥ 10%; mild decline < 10% but ≥ 0; mild improvement > 0 but < 10%; moderate improvement ≥ 10%) and this showed no statistically significant differences between groups in the proportion of participants within these categories. In Capacity 004, 20% of participants in the pirfenidone group and 35% of participants in the placebo group showed a decline on the per cent predicted FVC of at least 10%. In the analysis of all five categories in this study, a statistically significant effect of pirfenidone was seen (p = 0.01). It is unclear whether or not these subgroup analyses were adequately powered to detect a difference and, therefore, caution is required in their interpretation.

Azuma and colleagues66 also report VC litres in a categorical analysis, using three categories: ‘improved’, ‘stable‘ and ‘deteriorated’ (see Figure 5). Improvement was defined as at least a 10% improvement in VC, deteriorated was defined as at least a 10% decrease in VC, and any other value was categorised as stable. It is unclear whether this was calculated as an absolute or relative change. At 9 months a statistically significant difference between groups was observed on this categorical analysis (p = 0.003). The proportion of participants with an improvement in their VC was 9% in the pirfenidone group and 0% in the placebo group. The proportion with a decline in their VC was 13% in the pirfenidone group and 36% in the placebo group, and those classified as stable were 78% in the pirfenidone group and 64% in the placebo group. It is unclear whether or not the analysis of these data was adequately powered, and in addition, the analysis was not undertaken on the ITT population (n = 67 out of 72 pirfenidone and n = 33 out of 35 placebo) and caution is, therefore, required in the interpretation of these data.

Combining the three trials in a random-effects meta-analysis demonstrated a significant effect in favour of pirfenidone on the proportion of people with a decline in FVC of ≥ 10% (RR 0.62, 95% CI 0.41 to 0.93; p = 0.02 see Figure 5). Moderate statistical heterogeneity was present between the trials (I² = 57%).

Diffusing capacity of the lung for carbon monoxide

All four included pirfenidone trials report data on the DL_CO at end point (Table 29). In the Capacity 006 study62 the mean change from baseline in the per cent predicted DL_CO at 72 weeks was not statistically significant between the two groups (–9.8% vs. –9.2%, pirfenidone vs. placebo, respectively; p = 0.996). This was also the case in the Capacity 004 study62 (mean change from baseline –7.9% vs. –9.9%, pirfenidone vs. placebo, respectively; p = 0.145), and when combining these two trials in a meta-analysis (mean difference 0.68, 95% CI –0.87 to 3.22; p = 0.6; see Figure 3). The mean change in DL_CO ml/minute/mmHg in the Taniguchi and colleagues65 trial was also not statistically significantly different between the high-dose pirfenidone and placebo arms (p = 0.23) after 52 weeks. Finally, mean change in DL_CO ml/minute/mmHg was not statistically significant between the pirfenidone-treated participants and the placebo-treated participants at either 6 months (p = 0.49) or 9 months (p = 0.21) in the Azuma and colleagues trial. 66 However, combining these last two trials in a meta-analysis produces a beneficial effect in favour of pirfenidone that approaches statistical significance (mean difference 0.53, 95% CI –0.00 to 1.05; p = 0.05; Figure 7).

FIGURE 7.

Forest plot of mean change in DL_CO from trials of pharmacological interventions.

A categorical analysis of DL_CO was also reported by Azuma and colleagues. 66 Three categories, ‘improved’, ‘stable‘, and ‘deteriorated’, were used, where a change of 15% was used to categorise either improvement (+15%) or deterioration (–15%), and all other values were categorised as stable. At 9 months no statistically significant difference between groups was observed on the categorical analysis (p = 0.158). The proportion of participants with an improvement in their DL_CO was 16% in the pirfenidone group and 6% in the placebo group. The proportion with a decline in their DL_CO was 36% in the pirfenidone group and 47% in the placebo group, and those classified as stable were 50% in the pirfenidone group and 47% in the placebo group. Caution is required in the interpretation of these data.

Six-minute walk test

The mean change on the distance walked on the 6MWT was reported in the two Capacity RCTs (Table 30). In the Capacity study 006, a statistically significant treatment effect for pirfenidone was seen at 72 weeks, with a 31.8 m difference in the distance walked between the two groups seen (p = 0.0009). In the Capacity study 004, there was no statistically significant effect of treatment with pirfenidone (difference in the distance walked 16.4 m; p = 0.171).

All four pirfenidone studies report data on the lowest SpO₂ during the 6MWT (Table 31 and Figure 8). In neither of the Capacity studies was a statistically significant difference between groups observed at 72 weeks (Capacity 006: difference –0.5, p = 0.893; Capacity 004: difference 0.8, p = 0.087). In the Taniguchi and colleagues65 study, after 52 weeks no statistically significant differences between the high-dose pirfenidone group and placebo (p = 0.74) was observed. The change in the lowest SpO₂ on the 6MWT was the primary outcome in the Azuma and colleagues66 trial. No statistically significant differences were observed when analysed at 6 months’ (p = 0.148) or at 9 months’ (p = 0.722) follow-up.

FIGURE 8.

Meta-analysis of pirfenidone trials: mean change in worst SpO₂ during 6MWT.

Combining the four trials in a random-effects meta-analysis found no significant difference in SpO₂ during the 6MWT (mean difference 0.49, 95% CI –0.32 to 1.30; p = 0.24). Moderate statistical heterogeneity was evident (I² = 45%) between the trials (see Figure 8).

Azuma and colleagues66 also report a categorical analysis (improved, stable, deteriorated) of the SpO₂ during the 6MWT at 9 months. A > 4% increase was classified as ‘improved’, a ≤ 4% decrease was classified as ‘deteriorated’ and any values in between these two was classified as ‘stable’. Analysis of these data showed a statistically significant treatment effect in favour of pirfenidone (p = 0.02), although caution is required as this was unlikely to be a powered analysis, and the population was not the ITT population. Treatment with pirfenidone led to an improvement in 24% of cases, compared with 6% of cases in the placebo group. More participants deteriorated in the placebo group (18% pirfenidone, 33% placebo); however, the rates of participants classified as having a stable SpO₂ during the 6MWT were similar (58% vs. 61%, pirfenidone vs. placebo group, respectively).

Other secondary outcomes

Other outcomes reported in these studies include TLC, partial pressure of oxygen and HRCT outcome, the results of which can be found in Appendix 5.

Dyspnoea

Dyspnoea was measured by the University of California, San Diego, Shortness-of-Breath Questionnaire (UCSDSBQ) in the two Capacity trials, the results of which can be seen in Table 32. 62 The UCSDSBQ indicates severity of dyspnoea on a scale from 0 to 5 on 21 activities of daily living, along with three ratings on limitations caused by dyspnoea or fear of dyspnoea, for a total score ranging from 0 to 120, with a higher score indicating more dyspnoea. There were no statistically significant differences between the intervention and placebo groups on this outcome in either study (p = 0.604 Capacity 006; p = 0.509 Capacity 004). Dyspnoea was also reported by Azuma and colleagues;66 however, the measure used was not described and no data were reported except the p-value which indicated no statistical significant difference between groups (p = 0.64).

Acute exacerbations

The rate of acute exacerbations was reported by two of the pirfenidone studies (Table 33). At 52 weeks Azuma and colleagues66 demonstrated a statistically significant difference in the rate of acute exacerbations in those treated with pirfenidone and those treated with placebo, in favour of pirfenidone (0% vs. 14%, respectively; p = 0.003). Taniguchi and colleagues65 also reported the incidence of acute exacerbation but found no difference between the high-dose pirfenidone and placebo groups at 52 weeks (see Table 33).

Quality of life

Only the study by Azuma and colleagues66 reported having measured QoL; however, no data for this outcome were reported in the trial publication. The paper stated that there was no statistically significant difference between groups (p = 0.8720).

Adverse events

Serious adverse events were recorded in the Capacity RCTs62 and can be seen in Table 34, which shows IPF-related serious adverse events occurring in at least two participants in any treatment group. The rate of any serious adverse event was observably similar between those treated with pirfenidone and those treated with placebo in the Capacity study 00662 (pirfenidone 7.6% vs. 9.8% placebo), although no statistical testing of these rates were reported. All other serious adverse events reported in the study appear to be comparable between the treated and untreated study populations (see Table 34). A similar pattern was seen in the treatment arms of the Capacity 004 study (see Table 34).

All four included studies report details of any adverse events in their respective trials. In both Capacity trials,62 adverse events occurring in 10% or more of participants receiving pirfenidone and an incidence of 1.5 times or greater in placebo were reported (Table 35). The authors state that events were generally mild or moderate, without clinical consequences. Rates of any adverse events were seen to be similar between treatment group and placebo in both studies. Nausea, fatigue, diarrhoea and rash were the most frequently reported adverse events in the pirfenidone-treated participants (incidence in the region of 25–38% in both Capacity studies). Rates of these adverse events were in the region of 13% to 21% in the placebo-treated participants (for full details see Table 35). No statistical analyses of the rates between the groups were reported. All other adverse events appeared to be similar across groups with the exception of dyspepsia, which was seen to be more widely recorded in the pirfenidone groups.

Adverse events recorded in ≥ 5% participants in either the high-dose pirfenidone or the placebo group in the Taniguchi and colleagues trial65 can be seen in Table 36. The rate of any adverse events was similar across the pirfenidone treatment arm and the placebo arm (high dose vs. placebo p = 0.50). However, there were some specific adverse events that were seen to be statistically significantly different between treatment and placebo groups. Photosensitivity was seen in 51.4% of high-dose pirfenidone participants, compared with 22.4% of placebo participants (p < 0.01). Anorexia, dizziness and raised glutamyl-transpeptidase were all higher in those treated with high-dose pirfenidone than in those treated with placebo (see Table 36 for details). Two adverse events were more common in the placebo group than the high-dose pirfenidone group: nasopharyngitis and URTIs (Table 36). The report also provides details of statistical analyses of the rates between the two pirfenidone groups, which can be seen in Appendix 5. Specific adverse events leading to discontinuations for each treatment group were also reported but not data extracted. The numbers of dropouts for any adverse events were discussed earlier in this chapter (see Pharmacological interventions for mild to moderate idiopathic pulmonary fibrosis, Pirfenidone). No one particular adverse event appears to have been a more common reason for discontinuation.

Azuma and colleagues66 reported adverse events occurring in ≥ 10% of participants at 6 months. Rates of any adverse events were statistically significantly higher in the pirfenidone group than in the placebo group (p = 0.04) (Table 37). Photosensitivity, stomach discomfort, anorexia, nausea, fatigue and raised glutamyl-transpeptidase were all statistically significantly more common with pirfenidone than they were with placebo. All other adverse events were similar between the comparison groups. Reasons for treatment discontinuations were also reported by Azuma and colleagues66 and can be seen in Appendix 5. Although adverse events were more frequently reported to be a reason for discontinuation, none of the specific adverse events or adverse events in general was statistically significantly more likely to lead to discontinuation.

Summary

On measures of FVC, pirfenidone appears to demonstrate an effect when compared with placebo treatment. One trial (Capacity 00662) appears to be an outlier with results being non-significant and the reason for this is unclear. Results on other outcomes such as DL_CO, 6MWT distance and the oxygen saturation during the 6MWT generally showed no effect of treatment with pirfenidone, with the exception of the 6MWT in the Capacity 006 trial. 62 Acute exacerbations were assessed in two studies, with results favouring pirfenidone in one but not the other study. QoL outcomes were not reported and, therefore, the effect of pirfenidone on QoL cannot be estimated. Adverse event rates were generally similar between those treated with pirfenidone and those treated with placebo, although some specific adverse events (gastrointestinal events and photosensitive skin rashes) appear more likely with pirfenidone. Two studies had low risk of bias, but two had an unclear risk of bias which should be considered when interpreting results.

Cough and quality of life

Cough measured by VAS and cough-related QoL measured by the CQLQ were statistically significantly improved after 12 weeks of thalidomide compared with placebo (Table 38). The authors state that the MCID of the CQLQ in IPF is unknown, but the MCID for the Leicester Cough Questionnaire, which has been shown to be similar to the CQLQ, is 1.3. A mean difference in score of 11.4 on the CQLQ is, therefore, likely to be a clinically important effect. Sensitivity analyses undertaken to explore the effect of missing data found that thalidomide use resulted in a statistically significant improvement in CQLQ scores in each of the analyses. A statistically significant improvement in the total score, symptom domain and impact domain, but not the activity domain, of the SGRQ was also found with thalidomide (see Table 38).

Adverse events

Significantly more people experienced at least one adverse event with thalidomide than with placebo (77% vs. 22%; p = 0.001) (Table 39). The most common adverse events with thalidomide were constipation (36%), dizziness (27%), viral upper respiratory tract infection (URTI) (23%) and malaise (14%). Adverse events requiring dose reduction occurred in three people [constipation in two (9%) and bradycardia in one (5%)].

Summary

One small randomised crossover trial with a low risk of bias provided data on the effectiveness of thalidomide for supressing cough in IPF. Cough, cough-related QoL and respiratory-related QoL were significantly improved with thalidomide compared with placebo. Adverse events were experienced with thalidomide. Caution is required in the interpretation of these data given the small sample size.

Forced vital capacity

The results for per cent predicted FVC from the trial by the IPFCRN 2010 group (Zisman and colleagues)72 can be seen in Table 40. After 12 weeks of treatment there was no statistically significant effect for sildenafil, compared with placebo (mean difference 0.32, 95% CI –1.12 to 1.76; p = 0.66). Figure 2 presents the results of FVC graphically, where it can be seen that the CIs cross the line of no effect.

Diffusing capacity of the lung for carbon monoxide

Treatment with sildenafil led to a statistically significantly lower rate of decline of per cent predicted DL_CO from baseline than treatment with placebo: –0.33 versus –1.87%, respectively; p = 0.04 (Table 41 and see Figure 3).

Six-minute walk test

The presence of an improvement of at least 20% in the 6MWT distance at 12 weeks was the primary outcome measure in the Zisman and colleagues IPFCRN 201072 trial of sildenafil. As seen in Table 42, the proportions with an improvement of at least 20% were not statistically significantly different between groups (10% vs. 7% in the sildenafil and placebo groups, respectively; p = 0.39).

Other secondary outcomes

Other outcomes reported in the Zisman and colleagues IPFCRN 2010 trial72 include the partial pressure of oxygen, carbon dioxide and the oxygen saturation, the results of which can be found in Appendix 5.

Acute exacerbations

The incidence of acute exacerbation was not statistically significantly different between groups in the Zisman and colleagues IPFCRN 201072 trial, with 2% of the sildenafil-treated participants and 4% of the placebo-treated participants being recorded as having had an acute exacerbation (Table 43).

Quality of life

Quality of life of participants in the trial by Zisman and colleagues (IPFCRN)72 was measured using the SGRQ, the SF-36 and the EQ-5D. The scoring for the SGRQ was previously described [see Assessment of effectiveness of pharmacological interventions for mild to moderate idiopathic pulmonary fibrosis, BIBF 1120 (nintedanib) and Pirfenidone] and fuller details can be found in Chapter 4 (see Systematic review of health-related quality of life studies). The trial authors report that the MCID for the SGRQ is 5–8 points. The SF-36 measures aggregate scores on separate subscales for physical and mental health (see Chapter 4, Systematic review of health-related quality of life studies for more detail), with higher scores indicating better function. The study authors state that the SF-36 has a MCID of 2–4 points in IPF. The EQ-5D measures general QoL on a self-report questionnaire, with higher scores indicating better QoL (see Chapter 4, Systematic review of health-related quality of life studies) and on a VAS with a range of 0 to 100 (higher indicates better QoL). The study reports that the MCID is approximately 0.08 for the self-report questionnaire and 7 points for the VAS.

The changes from baseline scores of the SGRQ total score, symptoms score, activity score and impacts score were statistically significantly improved in those treated with sildenafil, compared with those treated with placebo (Table 45). On the SF-36 physical and mental health scores, and the EQ-5D self-report index and the VAS, there were no statistically significant differences between the two groups after 12 weeks of therapy. The study reported the baseline and end-point values on each of the SF-36 components, and full details of these can be seen in Appendix 5. Only the change from baseline score on the general health component was statistically significant between groups [sildenafil –1.04 (95% CI –2.52 to 0.44), placebo –3.89 (95% CI –5.37 to –2.42), difference 2.86 (95% CI 0.76 to 4.95); p = 0.008].

Adverse events

Serious adverse events were reported where at least one event was recorded during follow-up in the Zisman and colleagues IPFCRN 201072 trial. These are shown in Table 46 within each of the major organ classes presented in the publication. Appendix 5 has the complete details of each condition within each major class. The number of participants with at least one serious adverse event was similar between groups, with 15% and 16% for the sildenafil and placebo groups, respectively (see Table 46). In terms of number of events, the study publication reports that among all participants, 14 serious adverse events occurred in the sildenafil group and 23 in the placebo group. The rates of serious adverse events by each major organ class were also similar between the two groups.

A range of adverse events were reported for those randomised to treatment with sildenafil and those randomised to treatment with placebo and can be seen in Table 47, which reports these for each of the major organ classes presented in the report. As can be seen, there are no statistically significant differences in the proportion of adverse events recorded between the two groups. The rate of occurrence of any adverse events was 90% in the sildenafil-treated participants, compared with 87% in the placebo-treated participants. The majority of adverse events in both groups were either respiratory or infections.

Deaths from any cause were reported in the Zisman and colleagues IPFCRN 2010 trial,72 with two (4%) participant deaths in the sildenafil-treated arm and four in the placebo-treated arm. There was no statistically significant difference between the proportions of deaths (p = 0.43).

Summary

The primary outcome of a 20% improvement on the 6MWT was not significant in the study comparing sildenafil with placebo. Lung function as measured by the FVC was also not statistically significant, but the DL_CO results favoured treatment with sildenafil and this may, therefore, represent a pulmonary vascular effect. Dyspnoea may be improved (depending on the measure used and test conditions) and acute exacerbations showed similar rates between groups. QoL was better in those treated with sildenafil when measured using the SGRQ, but not when using the SF-36 or the EQ-5D. Serious adverse events and adverse events showed similar rates between groups in this 12-week study. This study had an unclear risk of bias, which should be considered when interpreting the results.

Dyspnoea

Participants’ dyspnoea was measured in the Lindell and colleagues74 trial using the UCSDSBQ. This measures shortness of breath on a scale of 0–120, with higher scores indicating more breathlessness. After the 6-week disease management programme or the usual care control, the scores on the UCSDSBQ were not statistically significantly different between the two groups (p = 0.972) (Table 48). No change scores of the UCSDSBQ were provided.

Quality of life

Quality of life of participants in the Lindell and colleagues74 trial was measured using the SF-36 and results can be seen in Table 49. As discussed previously, the SF-36 measures QoL within two domains: physical and mental health. Each scale is scored from 0 to 100, with a higher score indicating better function (see also Chapter 4, Systematic review of health-related quality of life studies). After 6 weeks the SF-36 physical health score was lower (worse) in the participants assigned to the disease management programme than in those assigned to the usual care control. The difference between the groups was statistically significant (p = 0.038). No change scores on the SF-36 were provided. The scores on the SF-36 mental health scale were not statistically significantly different between the two interventions (p = 0.772).

Summary

One small pilot study with an unclear risk of bias reported limited evidence on the effects of a disease management programme in IPF. The disease management programme did not lead to any significant differences in dyspnoea. QoL appears to be adversely affected on measures of physical health but not on measures of mental health. The study was designed to obtain pilot data for a future study and is unlikely to be sufficiently powered.

Forced vital capacity

After the 10-week pulmonary rehabilitation programme or usual care, the FVC was very similar between the two groups in the Nishiyama and colleagues73 trial (p-value not stated; Table 50). The FVC, litres, at end point was not changed from baseline measurements in either group. This study had an uncertain risk of bias.

In the Jastrzebski and colleagues CCT61 it is reported that there were no statistically significant changes in the results of lung function tests (assumed to include VC as this was an outcome) after 12 weeks, but no data were reported.

Six-minute walk test

The distance walked on the 6MWT was statistically significantly lower in the pulmonary rehabilitation group than in the usual care group at 10 weeks (p < 0.01) in the Nishiyama and colleagues73 trial (Table 51). This does not take into account the differences at baseline between the groups on this measure where the distance was also lower in the pulmonary rehabilitation group. When the mean change is estimated (by reviewers) it can be observed that there is a positive change in the distance walked in the pulmonary rehabilitation group and a negative change in the distance walked in the control group, but these data cannot be compared statistically to establish if there are any differences between the two groups.

The distance walked on the 6MWT at 12 weeks was 600.8 m in the intervention group and 544.5 m in the control group of the Jastrzebski and colleagues study61 (see Table 51). The study reports only within-group comparisons, however, so it is unclear whether or not this is a statistically significant difference. The 6MWT distance at 6 weeks is also reported in this study and data can be found in Appendix 5.

Dyspnoea

Dyspnoea was measured using the BDI in the Nishiyama and colleagues73 RCT of pulmonary rehabilitation versus control. The BDI comprises three categories: functional impairment, magnitude of task and magnitude of effort. The total BDI score ranges from 0 to 12, with higher scores denoting milder dyspnoea. Results can be found in Table 52. The difference between groups after the 10-week programme was small and not statistically significant (p-value not reported). The change in BDI was not presented but has been calculated here, where it can be seen that the BDI did not change in the pulmonary rehabilitation group and improved only slightly in the control group.

In the Jastrzebski and colleagues study,61 outcomes using both the Borg Dyspnoea Index and the BDI were reported. These included the Borg Dyspnoea Index before and after the 6MWT, the BDI functional impairment, magnitude of task, magnitude of effort, and the summary score. Results for the Borg Dyspnoea Index and summary score of the BDI at 12 weeks can be seen in Table 52, where it appears that these did not differ significantly between groups; however, no between-group analysis was presented in the study to confirm this. Results of the BDI functional impairment, magnitude of task and magnitude of effort and dyspnoea outcomes at 6 weeks can be found in Appendix 5.

Quality of life

Quality of life of participants in the RCT of pulmonary rehabilitation was assessed using the SGRQ. 73 As discussed previously, the SGRQ has a total score and three subscales, and each scale ranges from a score of 0 to 100 (lower scores indicating better QoL). Table 53 summarises the results from the Nishiyama and colleagues73 trial where the SGRQ outcomes were statistically compared using the end-point scores. This showed that the difference between the groups was statistically significant only on the total score (p < 0.05), suggesting that those in the control group had better QoL than those participating in the pulmonary rehabilitation programme. This does not take into account differences on these scales that were observed between the two groups at baseline where the pulmonary rehabilitation group had worse QoL. When the change from baseline scores are estimated, it would appear that QoL improved in the pulmonary rehabilitation group whereas it declined in the control group, although this is based on observation of the data only.

In the Jastrzebski and colleagues study,61 QoL was assessed using the SF-36. This measure is described in more detail in Chapter 4 (see Systematic review of health-related quality of life studies). Data from the 12-week assessment of the individual SF-36 components and the two summary scores can be seen in Table 53. No statistical analysis was presented of the comparison between treatment groups; however, it can be observed that scores were similar between groups.

Other secondary outcomes

Other outcomes reported in the Nishiyama and colleagues73 trial include FEV₁, TLC, and the partial pressure of oxygen and carbon dioxide, the results of which can be found in Appendix 5. Other outcomes reported in the Jastrzebski and colleagues study61 include inspiratory capacity and oxygen cost diagram, the results of which can be found in Appendix 5.

Summary

One study with an unclear risk of bias and one with a high risk of bias provided limited data on the effectiveness of a pulmonary rehabilitation programme or the addition of inspiratory muscle training, on IPF. Results are uncertain as to the effects of these types of intervention, and there were baseline differences between groups on many key outcomes (QoL, 6MWT), which should be considered when interpreting the results of one of the studies. On measures of lung function (FVC) and dyspnoea, there does not appear to be a beneficial effect of the programme.