In a cohort of patients diagnosed with IPF between 2010 and 2019, we observed a positive association between prescription of ICS in the year prior to diagnosis and hospitalisation for pneumonia, which remained significant after adjustment for key confounders. We found a similar, slightly weaker relationship, for PPI. We also observed a small increased risk for all-cause mortality in the “regular ICS user” group (compared with the “non-user” control group); however, we found no evidence of an association between the prescription of PPIs and all-cause mortality.

While other studies have reported increased pneumonia risks among longer-term users of ICS and PPIs, this is one of the largest studies to explicitly address this question in an IPF population. Our findings suggest that the same risks are present in people with IPF, which – given their age and high prevalence of conditions which are indications for ICS and PPIs – should be viewed as a potential cause for concern. At the very least the evidence for increased pneumonia risk should prompt clinicians to carefully consider the risk–benefit balance of continuing (or initiating) ICS/PPI use in people newly-diagnosed with IPF. This may be especially pertinent in people whose presenting respiratory symptoms may have been misdiagnosed as COPD or late-onset asthma prior to their IPF diagnosis, and in whom continuation of ICS is unlikely to confer any benefit. Several studies have established that misdiagnosis of non-specific respiratory symptoms is not only common in IPF but also leads to diagnostic delays, of up 5 years in some cases, as well as poorer outcomes [29, 30].

In our study cohort, around three quarters of patients who were prescribed ICS prior to their IPF diagnosis continued to be prescribed ICS post-diagnosis. Both prior- and post-ICS prescription was highly correlated with a prior diagnosis of either COPD or asthma, but we were not able to ascertain what proportion of our study cohort had comorbid COPD or asthma and what proportion had been misdiagnosed and therefore potentially inappropriately treated. Nevertheless, given estimates of the proportion of IPF patients with comorbid COPD/emphysema tend to centre around 30% [3], we suspect that a sizeable number of IPF patients may be continuing to be prescribed ICS-containing inhalers post diagnosis even though they do not have confirmed obstructive disease.

In contrast, we found little evidence of over-prescribing of PPIs in our study cohort. The IPF management guidelines that were in place during the period of this study [31, 32] include a conditional recommendation to consider treating people with IPF with antacid therapies, irrespective of GORD symptoms. This was based on early retrospective data demonstrating improved survival [33] and a slowing of disease progression [34] in people with IPF on antacids. Subsequent studies, which found no longer-term benefit of antacids and in one case evidence of increased risk of respiratory infections in patients with advanced IPF [35], have brought this recommendation into question [36,37,38] and the latest (2022) guidelines instead advise clinicians to only prescribe antacid medication to treat symptoms of GORD when present [39]. We found that only 20.6% of our cohort were prescribed PPIs on at least four occasions in the year following their diagnosis, suggesting that even during the period that they were active, the older guidelines were not necessarily being followed.

While in our study cohort both ICS and PPI prescribing was associated with an increased risk for pneumonia, only those prescribed ICS on a regular basis appeared to be at increased risk for all-cause mortality. We observed a statistically significant 25% increase in risk for all-cause mortality in regular ICS users, after adjustment for age, smoking status and comorbid conditions including COPD and lung cancer. The reasons behind this apparent association are likely complex, but without further investigation, we can only speculate as to the possible drivers. Part of the increased mortality risk might be attributed to higher likelihood of diagnostic delays in those prescribed inhaled therapies prior to their IPF diagnosis, denying patients the opportunity to initiate anti-fibrotic therapies at an earlier disease stage. Support for this hypothesis comes from a study conducted by Hoyer et al. involving 264 patients which found that longer diagnostic delays (over 1 year) were associated with worse progression-free survival times than shorter delays (< 1 year) (HR = 1.70, 95% CI: 1.18–2.46, p = 0.004). Interestingly, diagnostic delay was not significantly associated with all-cause survival (HR = 1.54, 95% CI: 0.95–2.51, p = 0.08) but the small sample size may mean that this study was underpowered with respect to mortality [30].

The main strength of this study derives from the use of CPRD’s Aurum dataset to define a large population-based cohort of people with a diagnosis of IPF. Validation work has confirmed the high positive predictive value (PPV) of GP-recorded diagnoses of IPF in this database [28]; this primary-care dataset has the added advantage of containing accurate prescribing information. Nevertheless, we have had to rely on prescribing history as a proxy for medication use as Aurum does not provide data on whether prescriptions were filled; this may result in overestimation of actual medication use. Conversely, over-the-counter antacid therapies have not been captured, and it is therefore possible that we have underestimated PPI use. However, for a high proportion of our study cohort, prescriptions are free of charge (i.e. in the over 60s) and thus use of over-the-counter antacids may be less significant.

Our study has a number of other limitations. Identification of our primary outcome events, hospitalisation for pneumonia, relies on diagnosis and ICD-10 coding by hospital consultants and staff, and we cannot be certain that diagnoses have been confirmed by microbiological testing. It is also likely that there is some misclassification of pneumonia outcomes, with some events in truth being acute exacerbations of IPF. Between 10% and 20% of people with IPF experience an acute exacerbation each year and such events are associated with high mortality [40]. However, misclassification is most likely non-differential and as such would have minimal impact on our Cox-regression-based conclusions.

This being an observational study, establishing cause-and-effect can be challenging and we acknowledge that we may not have fully adjusted for all potential sources of confounding. In the context of this study, confounding by indication is likely to be a significant issue. Adjustment for COPD and asthma did indeed noticeably reduce the magnitude of our HRs, but even after adjustment for these conditions, ICS use – even irregular use – was strongly associated with an increased risk of severe pneumonia, from which we inferred that while clearly present, confounding by indication alone cannot account for our findings. Furthermore, our sensitivity analyses, in which we explored the possibility of effect modification by COPD and/or asthma, found little evidence to suggest that the presence of COPD/asthma was driving the observed association: our stratified analyses indicate increased risks for both outcomes irrespective of the presence or absence of COPD and/or asthma (see Supplementary Material 1). The same is true of the PPI analyses: we found no evidence of effect modification of the association between PPI use and pneumonia by the presence of GORD or hiatus hernia. That said there may well be other risk factors for pneumonia which are also associated with the ICS and PPI use that we have not been able to account for (e.g. previous infections, more severe disease).

While we acknowledge that use of a time-varying exposure may have been a more robust modelling strategy, we considered that in our cohort this approach was not warranted and unlikely to materially affect our conclusions. As part of our descriptive analyses, we ascertained that most people who were prescribed ICS before their IPF diagnosis continued to be prescribed ICS after. A sensitivity analysis in which we excluded those individuals who only initiated ICS therapy after IPF diagnosis from the control group generated broadly similar results. In addition, we determined that most people who were prescribed PPI in the baseline exposure period were also prescribed PPI in the 3 months prior to their first pneumonia hospitalisation.

Finally, we were unable to explore what may be driving the observed increased risk for all cause-mortality in regular ICS users. However, this question merits further study, especially as this finding appears to be peculiar to the IPF patient population. In people with COPD, ICS use is not generally associated with increased mortality and is likely to be protective in some subgroups, such as frequent exacerbators and those with concomitant asthma. Replication of our study in other IPF cohorts would confirm whether an increased mortality risk is unique to our study cohort, which with a survival time of 1.7 years from IPF diagnosis, is likely to be dominated by individuals with more advanced IPF.