This analysis is among the first to examine use of anti-fibrotic medications in the US outside of the clinical trial environment and provides new insights into treatment patterns. Overall, 61% (60.7%) of persons with IPF were using an anti-fibrotic; 58 % (57.7%) were using a single anti-fibrotic. Conversely, 39.3% of persons with IPF were not prescribed one of these medications. Among those using a single medication (703 patients), 312 (44.4%) were taking nintedanib, and 391 (55.6%) were taking pirfenidone. Anti-fibrotic use varied by registry site. Rates of discontinuation were approximately 11% for both medication; the most common reason given was side effects. Many persons with severe IPF who may not have qualified for INPULSIS or ASCEND clinical trials were being treated with an anti-fibrotic in clinical practice. Twenty-three percent (23%, 274) of persons with IPF were using or had recently used an immunomodulatory agent.
The PFF-PR population is similar to the populations of the ASCEND and INPULSIS trials with regards to proportion male, and means of age and severity of lung disease as indicated by FVC (percent-predicted) and DLCO (percent-predicted). The PFF-PR cohort includes patients using an anti-fibrotic medication with more severe and less severe lung disease than those enrolled in the clinical trials at time of enrollment. As the registry does not describe severity of lung function impairment at time of medication initiation, a direct comparison is not possible with the clinical trial populations.
Patient characteristics were also similar in the PFF-PR registry as compared to Australian, German and European registries with regards to FVC, age, and gender [3]. Although anti-fibrotic treatment rates may vary between international registries due to differences in enrollment periods and anti-fibrotic availability, limited comparisons can be made. For example, the German INSIGHTS-IPF registry that began enrolling in 2012 reported pirfenidone use by 44.2% of persons compared to 25.6% use of nintedanib and 32.1% use of pirfenidone in this study.
In general, more severe lung disease – as indicated by lower DLCO and supplemental oxygen use – was associated with anti-fibrotic use. This may be due to providers and patients deferring anti-fibrotic initiation in patients with less severe disease. Such a strategy is not supported by evidence that shows anti-fibrotic use prevents irreversible lung function loss at all levels of disease severity [18]. There was decreasing use of anti-fibrotics with increased age; this may suggest that personalized decision-making results in deferral of anti-fibrotic use in patients of more advanced age for whom loss of future lung function might be of relatively lower concern when balanced against potential side effects and cost. Increased use of anti-fibrotics in recent clinical trial participants may be due to the relative lack of comorbid conditions that allowed these patients to be eligible for clinical trials. Further, clinical trial participants may be different than non-clinical trial participants [19].
One surprising finding is that nearly 40% of persons with IPF were not prescribed one of these disease-altering medications. It is possible that provider unfamiliarity with these newer medications, concerns about side effects, or concerns about financial cost may be reasons for deferral of medication initiation.
Differences in insurance type and status were not associated with anti-fibrotic use or selection. Similarly, cost was rarely listed as a reason for medication discontinuation. It is unlikely that differences among payers across regions was responsible for the significant regional differences in anti-fibrotic use and selection observed.
It is possible that the increased time since diagnosis in anti-fibrotic users is attributable to a survival benefit suggested for anti-fibrotic use [20]. The observation of increased likelihood of pirfenidone use in persons with IPF with longer time since diagnosis also could be related to participation in the CAPACITY trials which were enrolling patients prior to nintedanib trials [2, 16]. Registry site participation in a specific nintedanib or pirfenidone clinical trial was not associated with differences in prescription of that medication.
Registry site was incorporated into a random effects model as a random intercept. Registry site may serve as a proxy for the physician(s) who are an integral component of the shared decision-making process involved in treatment decisions. The random effects model provides evidence of variation in anti-fibrotic use between sites after controlling for patient and registry site characteristics; this variation merits additional study.
There was an association between pirfenidone use in patients with cardiovascular disease (CAD and pulmonary hypertension). This may relate to the initial report of increased myocardial infarction among nintedanib users compared to placebo in the INPULSIS trials [2], the risk of which has been clarified in subsequent analysis [21]. Similarly, anticoagulant use was associated with pirfenidone use. This finding is likely due to differences in exclusion criteria between the INPULSIS and ASCEND trials.
In the final multivariable analysis for nintedanib vs. pirfenidone use, enrollment at a Midwest registry site was associated with higher likelihood of pirfenidone use. To explore this apparent variation by region, the analysis explored for association between the meteorological variable average monthly UV index at the registry site and anti-fibrotic selection, as photosensitivity is a documented adverse effect of pirfenidone [22]. It is possible that patients living in geographic areas with higher average or maximum UV index would be more prone to photosensitivity reactions [23] and therefore avoid use of pirfenidone. However, this association was not seen in the current analysis which could not account specifically for location of patient residence and therefore used registry site as a surrogate. It is possible that unexplained variations in anti-fibrotic selection between registry site exist and that a few proportionately higher-enrolling regional centers affect the overall prescription pattern for the region. Use of a random effects model, limited by inadequate statistical power, could help answer this question. Understanding apparent regional variations in anti-fibrotic selection merits further study.
Twenty-three percent (23%, 274) of persons with IPF were using or had recently used an immunomodulatory agent. Use of these medications for IPF would be discordant with clinical guidelines [24]. One possible reason for prescription of these medications could be treatment of extra-pulmonary disease unrelated to IPF. Another possible use for these medications could be for the proposed disease entity interstitial pneumonia with autoimmune features (IPAF); such a treatment strategy has not been verified by randomized clinical trials. It is also possible that patients were given non-IPF diagnoses and treated with immunomodulatory medications prior to evaluation at the registry site, where the diagnosis was changed to IPF and different treatment recommendations given.
Strengths of this study include analysis of detailed patient information for a cohort of well-characterized persons with IPF. Missing data was minimal and where present, such as for missing values of DLCO, is likely explained by the known difficulty in completing DLCO measurement for patients with the worst lung function [25]. This study benefits from observation of patients in a more real-world setting to evaluate relationships between patient- and registry site-level characteristics and anti-fibrotic use and selection. Use of the random effects model explores the influence of registry site, as a proxy for provider, on anti-fibrotic use. Additionally, the analysis of average UV index and effect on anti-fibrotic selection and the relationship between prior clinical trial participation by a center on subsequent prescription are unique. Inclusion of patients currently or recently participating in clinical trials does not invalidate the analysis as the period of PFF-PR enrollment and entry data gathering did not overlap with the time periods of the ASCEND, INPULSIS or CAPACITY trials.
Limitations include lack of provider-specific information, which required use of registry site as a proxy for physician in this study. However, a single influential provider may influence peer and trainee practice indirectly and directly through participation in multidisciplinary conferences, a core component in the care of persons with IPF. Limitation of analysis to registry sites, of which many were university settings, might limit generalizability. Ultimately, the accuracy of the clinical IPF diagnosis at the enrolling registry site cannot be verified. Confidence in the diagnosis would be increased by more frequent use of multidisciplinary discussion, reported in the registry for 43% of persons with IPF. This analysis only evaluated relationships between characteristics and medication use at time of enrollment to the registry; information regarding anti-fibrotic use prior to 12 months preceding registry enrollment was not collected and subsequent anti-fibrotic use was not analyzed. Statistical power limited the performance of the random effects model to the analysis of nintedanib vs. pirfenidone use which might have helped explain observed regional variations in anti-fibrotic utilization.