In this retrospective study of a randomised trial of simtuzumab in idiopathic pulmonary fibrosis (IPF), prodromal decline in forced vital capacity (FVC) was significantly associated with increased risk of mortality, respiratory and all-cause hospitalisations, and categorical disease progression. Predictive modelling of progression-free survival event risk was used to assess the effect of population enrichment for patients at risk of rapid progression of IPF; C-index values were 0.64 (death), 0.69 (disease progression), and 0.72 (adjudicated respiratory hospitalisation) and 0.76 (all-cause hospitalisation). Predictive modelling may be a useful tool for improving efficiency of clinical trials with categorical end points.
- idiopathic pulmonary fibrosis
Statistics from Altmetric.com
Assessment of new therapies for idiopathic pulmonary fibrosis (IPF) requires long clinical trials with large populations due to marked heterogeneity of IPF progression. Improved prediction of IPF progression could streamline trial design, allowing for enrichment of vulnerable patients at risk for rapid disease progression.1–3
This post hoc analysis of data from a large phase II trial of the lysyl oxidase-like 2 (LOXL2) inhibitor simtuzumab (RAINIER)—which terminated for lack of efficacy4—assessed the relationship between variables, including change in forced vital capacity in weeks 1–14 (ΔFVC 0–14) and either categorical risk of disease progression or subsequent change in forced vital capacity (ΔFVC) over 12 months. Based on these results, a predictive model for progression-free survival (PFS) events was developed to approximate the influence of enrichment for risk on the RAINIER population.
Materials and methods
Details of the RAINIER study of simtuzumab in IPF (NCT01769196) conducted from March 2011 until January 2016, and terminated due to lack of efficacy, were previously reported4 and summarised in online supplementary file 1. This analysis included simtuzumab-treated and placebo-treated patients. Progression of IPF was defined as ≥10% relative decline in FVC% predicted from baseline with ≥5% absolute change, and PFS as time to progression or death.
Multivariate analysis of ΔFVC from weeks 14–66 was developed using age, body mass index, smoking status, geographical region, duration of IPF diagnosis, FVC, residual volume, diffusing capacity of the lung for carbon monoxide (DLCO)% predicted, 6 min walk distance (6MWD), St George’s Respiratory Questionnaire (SGRQ), serum LOXL2 level and quantitative high-resolution CT (HRCT) fibrosis score. Additional variables included change in longitudinal variables (FVC, DLCO% predicted, SGRQ, and 6MWD) from weeks 0–14. Multivariate survival analysis of death, categorical disease progression, and respiratory and all-cause hospitalisation was developed with the baseline variables and ΔFVC 0–14. Model development and performance details are provided in online supplementary file 1. The prodromal period included week 14 postbaseline measurements and provided ≥52 weeks of subsequent follow-up for the majority of patients.
At 14 weeks, 501/544 RAINIER patients remained on study. Online supplementary table 1 summarises patient baseline and clinical characteristics. The majority of patients were male, former smokers, aged 68.1±7.3 years and from North America.
Early rapid decline in lung function (FVC 0–14 of ≥100 mL; 210 patients) was associated with increased mortality risk (HR 1.31, 95% CI 1.05 to 1.64, p=0.017), adjudicated respiratory hospitalisation (HR 1.22, 95% CI 1.08 to 1.38, p=0.001) and categorical disease progression (HR 1.95, 95% CI 1.75 to 2.17, p<0.0001). Over a 52-week period spanning study weeks 14 and beyond, FVC at week 14 and change in FVC and SGRQ for the prodromal period of weeks 0–14 (ΔFVC 0–14) were negatively associated with ΔFVC modelled as a continuous variable (online supplementary table 2).
Key results for modelling the outcomes of categorical disease progression, death, all-cause hospitalisation and adjudicated respiratory hospitalisation are provided in table 1. Larger decline in ΔFVC 0–14 was associated with increased risk of categorical events; baseline FVC was not a significant predictor of risk of first all-cause hospitalisation.
Early rapid decline in lung function was associated with slower subsequent decline in FVC (R=−20.50 mL, p<0.001); ΔFVC beyond week 14 was negatively associated with ΔFVC 0–14 (estimate=−0.205, p<0.001) (table 1).
Applying the estimated parameters from the PFS model (online supplementary file 1) to patients enrolled in RAINIER, we obtained an average probability of a PFS event at 1 year of 47.8%; 517 patients meeting RAINIER eligibility criteria are needed to reach 247 events, a threshold selected based on power calculations of RAINIER.4 Based on this model, 93.2% of RAINIER-eligible patients have a >10% probability of experiencing a PFS event on placebo; for 17.6% of RAINIER-eligible patients, the probability of a PFS event is >80% (table 2). With increased enrichment, the number of patients required to achieve the target event rate decreases; however, screen failure rate increases (table 2). Comparing the observed PFS event rate in RAINIER to model predictions, incremental enrichment from a 60% probability threshold to an 80% probability threshold increased the observed proportion of patients with PFS events by 4.7%; however, the number of patients required at screening increased by 81%.
In this retrospective analysis, patients with IPF with accelerated decline over a period of 14 weeks are at increased risk of categorical adverse events (10% decline in FVC, all-cause and respiratory hospitalisations or death) over the subsequent 52 weeks, relative to patients with a slower rate of decline (table 1). In future trials using composite end points of categorical events, rate of decline over this short prodromal period could be used to enrich or stratify a study population with patients at increased risk of events indicative of disease progression. Table 1 suggests baseline HRCT score may further enrich prediction. However, if linear decline in FVC as a continuous variable is the primary end point, the prodromal rate of decline over 14 weeks is not apparently useful for stratification or enrichment. Here, more rapid decline in FVC is associated with increased risk of categorical adverse outcomes (but not linear decline), consistent with analysis of the pirfenidone registration trials, despite significant differences in inclusion criteria between those trials and RAINIER.5–7
Predictive modelling-based enrichment of RAINIER-eligible patients for rapid progression decreases the number of patients required to observe a target number of PFS events but increases screen failure rate. High screen failure rates may increase enrolment duration and number of sites required, making excessive enrichment counterproductive for overall trial efficiency. Enrichment for high risk of progression could also better represent patients with severe IPF who may be excluded from clinical trials. Further work is needed to define optimal inclusion criteria for IPF clinical trials. Study limitations include the following: this was a retrospective analysis, patients with mild lung function impairment were excluded, and placebo-treated and simtuzumab-treated patients were combined.
Rapid prodromal decline in FVC appears less useful in predictive modelling of subsequent decline in FVC as a continuous variable in patients with IPF. Predictive models based on clinical trial data incorporating recent decline in FVC may improve trial design using categorical efficacy end points.
The authors thank Jenny J. Zhang, PhD, for biostatistics support. Editorial support was provided by Terri Schochet, PhD, of AlphaBioCom, LLC, and funded by Gilead Sciences, Inc.
Correction notice This article has been corrected since it published Online First. The second to last sentence has been corrected.
Contributors TOR provided the text for the first draft and collated comments for all drafts. QG performed the statistical analysis. FJM, GR and BL provided the text for the first draft. GR, FJM, KKB, DJL and PWN served on the advisory committee that oversaw the primary study. VC, RJK and KFG adjudicated clinical end points. SDP and VS contributed to the LOXL2 biomarker analysis. DAL and SMH analysed the high-resolution CT data. All authors reviewed and provided comments on the final draft of the manuscript. All authors except BL reviewed the study design. Some of the placebo data were presented as an abstract at ATS in 2017.
Funding The study was funded by Gilead Sciences, Inc.
Competing interests GR has provided consultation for Gilead Sciences, Inc., for this work; for Bellerophon, Biogen, Boehringer Ingelheim, BMS, FibroGen, Gilead Sciences, Inc., Nitto, Promedior, Roche-Genentech, Sanofi-Aventis and Veracyte outside the submitted work; and has received research grants from the National Institutes of Health (NIH) for idiopathic pulmonary fibrosis studies outside the submitted work. BL reports speaker fees from Genentech outside the submitted work. KKB reports grants from the National Heart, Lung, and Blood Institute
(NHLBI); grants and consulting fees from Actelion; Amgen; and Gilead Sciences, Inc.; and personal fees from Aeolus, Almirall, Altitude Pharma, AstraZeneca, Bayer, Biogen/Stromedix, Boehringer Ingelheim, Celgene, Centocor, FibroGen, Galecto, GlaxoSmithKline, MedImmune, Novartis, Pfizer, Promedior, ProMetic, Roche/Genentech, Sanofi/Genzyme and Veracyte. VC reports personal fees from Actelion, Bayer, Biogen Idec, Boehringer Ingelheim, Gilead Sciences, Inc., GlaxoSmithKline, Merck Sharp and Dohme, Novartis, Pfizer, Roche/InterMune and Sanofi; grants from Actelion, Boehringer Ingelheim, GlaxoSmithKline, Pfizer and Roche; and personal fees from Boehringer Ingelheim outside the submitted work. KFG, PWN and JWS declare no competing interests. RJK reports personal fees from Boehringer Ingelheim, Genentech, MedImmune and Gilead Sciences, Inc.; and other fees from Afferent and Bristol-Myers Squibb outside the submitted work. DJL reports grants from Gilead Sciences, Inc., during the conduct of the study; grants and personal fees from Boehringer Ingelheim; personal fees from Degge Group; France Foundation; Genentech; Gilead Sciences, Inc., and Veracyte; grants from Bayer, FibroGen and Global Blood Therapeutics; and grants, consulting fees and institutional support from Pulmonary Fibrosis Foundation outside the submitted work. AUW reports personal fees from Actelion, Bayer, Boehringer Ingelheim, InterMune and Roche during the conduct of the study. TPW reports grants from Gilead Sciences, Inc., during the conduct of the study; personal fees from Boehringer Ingelheim; Genentech and Gilead Sciences, Inc.; and grants from Boehringer Ingelheim, Celgene, Genentech, Global Blood, Kadmon, Pulmonary Fibrosis Foundation Therapeutics and Sanofi outside the submitted work. DAL reports grants from NHLBI and grants and consulting fees from Boehringer Ingelheim, Parexel, Roche/Genentech, Siemens and Veracyte outside the submitted work. SH reports consulting for Parexel during the conduct of the study and for Boehringer Ingelheim outside this study; and has a patent system and method for automatic detection and quantification of pathology using dynamic feature classification pending to National Jewish Health. EM is an employee of bioMérieux. KFG, SDP, VS, QG and JSS are employees and may hold stock in Gilead Sciences, Inc. At the time of writing, TGOR was an employee of and a stockholder in Gilead Sciences, Inc. FJM reports personal fees from the Adept, Academic Continuing Medical Education, American Thoracic Society, Afferent, Axon Communication, Boehringer Ingelheim, Clarion, Falco, Genentech, Ikaria/Bellerophon, Kadmon, National Association for Continuing Education, Nycomed/Takeda, Pfizer, Potomac and Veracyte; grants from the NIH; nonfinancial support from Biogen/Stromedix, Boehringer Ingelheim, Centocor and Gilead Sciences, Inc.; and other support from Bayer, Boehringer Ingelheim, Genentech and Veracyte during the conduct of the study; grants, personal fees and nonfinancial support from AstraZeneca, GlaxoSmithKline and Nycomed/Takeda; grants and personal fees from Forest; grants from BioScale; personal fees from Amgen, Annenberg, Axon, Boehringer Ingelheim, California Society for Allergy and Immunology, CME Incite, ConCert, Genentech, HayMarket, Ikaria/Bellerophon, Informa, Integritas, InThought, Janssen, Lucid, Methodist Hospital, Miller Medical, National Association for Continuing Education, Novartis, Paradigm, Pearl, Peer Voice, Pfizer, Prime, Roche, Sunovion, Theravance, Unity Biotechnology, UptoDate, WebMD and Western Society of Allergy and Immunology; nonfinancial support from Annenberg, California Society for Allergy and Immunology, CME Incite, ConCert, Genentech, Janssen, Miller Medical, National Association for Continuing Education, Novartis, Pearl, Pfizer, Roche, Sunovion, Theravance, WebMD and Western Society of Allergy and Immunology; and other support from GlaxoSmithKline and Merio outside the submitted work.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.