Discussion
This study provides real-world evidence of improved PFS at 6 months and reduced decline in FVC over time in patients with IPF treated with antifibrotic therapies. Although motality of 6, 12 and 18 months appeared lower in the antifibrotic group, this was not significant. However, exclusion of patients who discontinued treatment within 6 months from the analysis resulted in significantly improved mortality in the antifibrotic group. Patients appeared to tolerate nintedanib better than pirfenidone for the first 3–6 months of treatment. Continued treatment in the first 3 to 6 months of antifibrotic therapy was associated with significantly improved 12 month mortality. Multivariate analysis of the entire cohort identified BMI of ≤25 and DLco of ≤35% predicted as risk factors for mortality. Baseline BMI of ≤25, age of ≥75 years and female sex were found to be risk factors for antifibrotic discontinuation secondary to both adverse drug reactions and disease progression.
The promising effects of antifibrotics on PFS and FVC decline, but absence of significant effect on mortality, are consistent with the results of the clinical trials.3–5 However, the mortality rates in this study are higher than those seen in RCTs, which we speculate are secondary to more lenient inclusion criteria for antifibrotic therapies in everyday clinical practice. For example, the antifibrotic group appeared to be of older age and to have worse DLco than the pirfenidone arm of the CAPACITY trial.3 Patients in real-life clinical practice are likely to have more comorbidities as trials such as the ASCEND trial excluded patients likely to die within 2 years.4 The intolerance of antifibrotics is high when compared with the clinical trials. This highlights the importance of real-world data to inform clinical decision making as the availability of counselling and resources may be different in these settings when compared with trial conditions.
The discontinuation rates at 3 and 6 months were significantly better with nintedanib, which we have shown could have beneficial effects on survival. These findings may favour commencement of nintedanib as first-line antifibrotic therapy. However, this appears to differ from previous work that showed discontinuation rates were comparable between nintedanib and pirfenidone,18 19 and so further studies are required to investigate this. Discontinuation rates here were also higher than a recent US pulmonary fibrosis registry-based study where it was only around 11%, which may be related to different healthcare settings.18
There are cases where pirfenidone would be the preferred therapy over nintedanib, such as due to comorbidities or pharmacological interactions. This study shows that pirfenidone is less well tolerated at all time points, but more so in the first 6 months as following this, the Kaplan-Meier curve flattens out. Of the patients who stopped pirfenidone by 12 months, 86% had stopped within 6 months. This suggests that patients who can tolerate the early phase of therapy are more likely to continue long term with discontinuation rates at 12 months not significantly different from nintedanib. Nausea was the most frequent side effect in patients on pirfenidone, which can be managed through concomitant prescription of an antiemetic. Loperamide was often coprescribed with nintedanib and found to be effective. However, the same effectiveness may not have been seen with pirfenidone as often it is a combination of side effects with general malaise that leads to discontinuation.
The side-effect profile seen here is like those in previous trials. In a pooled analysis9 of the TOMORROW and INPULSIS trials of nintedanib, 61.5% and 24.3% of patients experienced diarrhoea and nausea, respectively, similar to here. For pirfenidone, rash and diarrhoea were less frequently reported than in a pooled analysis of the CAPACITY and ASCEND trials,20 although a very similar proportion of patients suffered from nausea. The much lower reporting of photosensitive rash may be attributed to the effectiveness of counselling patients regarding the use of sunblock. For both medications, appetite loss was more common in our study, which is most likely due to the older, multimorbid patient group.
DLco has previously also been shown to be a predictor of mortality.21–23 This is also true for baseline BMI in both the preantifibrotic and postantifibrotic eras.24 25 Decline in BMI has also been shown to be a risk factor for increased mortality.26 27 Our results are surprising as it appears patients with a normal BMI are at an increased risk of dying. However, this finding, alongside DLco as a predictor of mortality, is consistent with Fang et al’s real-world pirfenidone study.28 We also noted that baseline BMI was a predictor of antifibrotic discontinuation, which has been reported previously,29 with Kato et al30 finding that low BMI increases the risk of experiencing nintedanib-associated gastrointestinal side effects.
Age31 32 and female sex29 31 have also been shown to be risk factors for discontinuation previously. The results of the multivariate analysis suggest that there are demographic factors that influence discontinuation that need further study.
Limitations
The limitations of the study are related largely to those encountered in all real-world retrospective analyses. The infrequency of lung function being performed in everyday clinical practice led to missing FVC data. FVC data imputation was not carried out for patients who died. This was addressed through the use of PFS as an outcome when comparing the two groups. There are also limitations with the control group used. This control group was formed retrospectively. Some of these patients were not started on antifibrotics as they had stable disease, potentially reducing the positive effect of the antifibrotics on PFS. There were some differences in the baseline characteristics of the two groups, but these variables were not found to independently influence mortality. The overall number of patients in this study is relatively small and is particularly limited by the size of the control group (n=64). We also acknowledge that this is a single centre study, which limits the generalisation of the data to the diversity of NHS ILD centres in the UK.