Clinical Effectiveness of the Anti-Fibrotic Medications for Idiopathic Pulmonary Fibrosis

Timothy M. Dempsey, MD1, Lindsey R. Sangaralingham, MPH2,3, Xiaoxi Yao, PhD4, Darshak Sanghavi,

MD3, Nilay D. Shah, PhD2,4, Andrew H. Limper, MD1,2,5

1 Department of Pulmonary and Critical Care Medicine, Mayo Clinic, Rochester, Minnesota

2 Mayo Clinic Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery,

Mayo Clinic, Rochester, Minnesota

3 OptumLabs®, Cambridge, Massachusetts

4 Department of Health Sciences Research, Mayo Clinic, Rochester, Minnesota

5 Correspondence: Dr. Andrew Limper, Gonda 18-South, Mayo Clinic, Rochester, MN 55905; Tel.: (507) 284-8484; Fax: (507) 284-5421; E-mail address: limper.andrew@mayo.edu

Author Contributions: TMD, AHL, LRS: developed concept and study design. LRS, XY, DS,

NDS acquired and evaluated the claims data. TMD: drafted initial manuscript. All authors

reviewed and approved final manuscript.

Sources of Support: This study was funded by the Mayo Clinic Robert D. and Patricia E. Kern

Center for the Science of Health Care Delivery, which receives no industry funding.

Running Head: Clinical Effectiveness of Anti-Fibrotics

Descriptor: 9.23 Interstitial Lung Disease

Disclosures/Disclaimers:

In the past 36 months, Dr. Shah has received research support through Mayo Clinic from the

Food and Drug Administration to establish Yale-Mayo Clinic Center for Excellence in

Regulatory Science and Innovation (CERSI) program (U01FD005938), from the Centers of

Medicare and Medicaid Innovation under the Transforming Clinical Practice Initiative (TCPI),

from the Agency for Healthcare Research and Quality (R01HS025164; R01HS025402;

R03HS025517; 1U19HS024075), from the National Heart, Lung and Blood Institute of the

National Institutes of Health (NIH) (R56HL130496; R01HL131535), National Science

Foundation, and from the Patient Centered Outcomes Research Institute (PCORI) to develop a

Clinical Data Research Network (LHSNet). ). In the past 36 months, Dr. Limper has also

received research support through Mayo Clinic for the study of interstitial lung diseases from the

Hurvis Foundation, The Annenberg Foundation, and the Rosemary Clooney Pulmonary Research

Fund. All others have no disclosures in regard to this manuscript.

FOOTNOTES (first page): This study was funded by the Mayo Clinic Robert D. and Patricia E.

Kern Center for the Science of Health Care Delivery, which receives no industry funding.

Word Count: 3179

Clinical impact: This is the first real-world analysis in the United States of the clinical

effectiveness of the anti-fibrotic medications pirfenidone and nintedanib. Using a large

insurance database to perform a retrospective cohort analysis, we observed that the medications

had an association with a reduced risk of all-cause mortality for up to two years of follow-up in

patients with idiopathic pulmonary fibrosis compared to an untreated matched cohort. Due to the

high mortality associated with idiopathic pulmonary fibrosis, this is a significant finding for

patients with the disease and the clinicians who treat it.

“At a Glance Commentary”

Scientific Knowledge on the Subject: Idiopathic pulmonary fibrosis is a progressive lung disease with high mortality that had no effective medical treatment options until the approval of the anti-fibrotic medications pirfenidone and nintedanib in 2014. While the randomized trials leading to their approval demonstrated the medications slow the decline in lung function and may have a trend toward decreased mortality, to date there has been no analysis of their effect on clinically important outcomes such as mortality and hospitalizations in every day clinical practice.

What This Study Adds to the Field: This is the first real-world analysis of the clinical effectiveness of pirfenidone and nintedanib. Using a large insurance database to perform a retrospective cohort analysis, it was observed that the medications had an association with a reduced risk of all-cause mortality and hospitalizations for up to two years of follow-up compared to an untreated matched cohort. There were no significant differences between the two drugs in all-cause mortality. The findings in this study support the randomized trial data suggesting these medications have an impact on clinical outcomes in patients with idiopathic pulmonary fibrosis.

1

ABSTRACT

Rationale: Since their approval, there has been no real-world or randomized trial evidence

evaluating the effect of the anti-fibrotic medications pirfenidone and nintedanib on clinically

important outcomes like mortality and hospitalizations.

Objectives: To evaluate the clinical effectiveness of the anti-fibrotic medications in patients with

idiopathic pulmonary fibrosis.

Methods: Using a large United States insurance database, we identified 8098 patients with

idiopathic pulmonary fibrosis between October 1, 2014 and March 1, 2018. A one-to-one

propensity score matched cohort was created to compare those treated with anti-fibrotic

medications (n=1255) to those not on treatment (n=1255). The primary outcome was all-cause

mortality. The secondary outcome was acute hospitalizations. Subgroup analysis was performed

to evaluate mortality differences by drug.

Measurements and Main Results: The use of anti-fibrotic medications was associated with a

decreased risk of all-cause mortality (Hazard Ratio 0.77; 95% confidence interval, 0.62 to 0.98; p

value=0.034). However, this association was present only through the first two years of

treatment. There was also a decrease in acute hospitalizations in the treated cohort (Hazard Ratio

0.70; 95% confidence interval, 0.61 to 0.80; p value<0.001). There was no significant difference

in all-cause mortality between patients receiving pirfenidone and those on nintedanib (Hazard

Ratio 1.14; 95% CI, 0.79 to 1.65; p=0.471).

Conclusions: Among patients with idiopathic pulmonary fibrosis, anti-fibrotic agents may be

associated with a lower risk of all-cause mortality and hospitalizations compared to no treatment.

2

Future research should test the hypothesis that these treatments reduce early but not long-term

mortality as demonstrated in our study.

Abstract word count: 250

3 to 5 key words: pirfenidone; nintedanib; idiopathic pulmonary fibrosis; mortality; acute

hospitalizations

Abbreviations: CI- confidence interval; FDA- Food and Drug Administration; ICD-9-

International Classification of Diseases, 9th edition; ICD-10- International Classification of

Diseases, 10th edition ; IPF- idiopathic pulmonary fibrosis; HR- Hazard Ratio; RCT- randomized

controlled trial; SD- standard deviation; US-United States

3

INTRODUCTION: Idiopathic pulmonary fibrosis (IPF) is a chronic, progressive fibrosing lung disease with high

mortality that is associated with the usual interstitial pneumonia pattern on computed

tomography scan and/or biopsy.1 Over the years, several medical therapies have been tested for

the treatment of patients with IPF, though they have largely failed to show benefit and some have

even demonstrated harm.1-3 In October 2014, the anti-fibrotic medications pirfenidone and

nintedanib were concurrently approved by the Food and Drug Administration (FDA) following

the results of two phase III, multicenter, placebo-controlled randomized trials (ASCEND and

INPULSIS 1 and 2) that demonstrated the medications slowed the decline in lung function in

patients with IPF.4,5 There was also a non-statistically significant trend toward improved survival

for both medications, though there was no change in respiratory symptoms, patient-reported

outcomes, or quality of life with either treatment. This prompted a “conditional

recommendation” for the use of the two medications in the 2015 American Thoracic Society IPF

clinical practice guideline statement, as well as much anticipation from physicians given the

belief that there were now multiple effective treatment options for patients with IPF.1, 6

Since approval, there have been several systematic reviews and pooled data analyses which have

largely confirmed the outcomes of the randomized controlled trials (RCTs).7,8 There have also

been network meta-analyses indirectly comparing the two medications that have suggested there

were no clear outcome differences between the treatments.9-11 Given these comparable results in

clinical trials and systematic reviews (and lack of direct comparisons), it is believed that most

physicians who treat patients with IPF ultimately make the decision for therapy following a

shared-decision making discussion with patients, largely based on the possible adverse effects of

the chosen medication as well as out-of-pocket costs.

4

Despite the initial excitement surrounding the approval of the anti-fibrotics, there are several

questions that remain, including whether these medications actually work in general clinical

practice to improve survival and decrease hospitalizations. To date, there has been no analysis of

the available real-world data of IPF patients on therapy to answer these clinically important

questions. Thus, in this study, we aim to assess the treatment benefits of the medications

compared to patients with IPF not receiving treatment as well as compare the effectiveness of the

two medications head-to-head using individuals enrolled in United States (US) commercial and

Medicare Advantage health plans. Some of the results of this study have been previously

reported in the form of an abstract.12

METHODS:

Data Source: This analysis was a retrospective cohort study using de-identified administrative

claims data from the OptumLabs® Data Warehouse, which includes data for commercially

insured and Medicare Advantage enrollees in a large US health plan.13 The database contains

claims-based information on individuals from all 50 states comprising all ages, ethnicities, and

racial groups. The health plan provides comprehensive insurance coverage for physician,

hospital, and outpatient prescription services. Pursuant to the Health Insurance Portability and

Accountability Act, the use of de-identified data does not require Institutional Review Board

Approval.

Study Populations: Two cohorts of patients with IPF were created for this study: an untreated

cohort and a treated group. For the treated cohort, we identified all adult patients (≥18 years) that

filled a prescription for either pirfenidone or nintedanib between October 1, 2014 and March 1,

2018. We required treated individuals to have a diagnosis of idiopathic pulmonary fibrosis using

5

the International Classification of Diseases, 9th edition (ICD-9) and International Classification

of Diseases, 10th edition (ICD-10) billing codes for IPF (9th Edition, 516.31; 10th Edition,

J84.112) in the 180 days prior to their first fill for either drug. For the untreated cohort, we

identified individuals with a diagnosis of IPF (based on the above codes) who did not fill a

prescription during our study period. For each patient, we defined the index date as the date of

first fill for either pirfenidone or nintedanib or the first IPF encounter in the study period after

patients met the 180 days enrollment requirement. We required all patients to have at least 180

days of continuous enrollment before the index date in order to have sufficient data to capture

patient’s baseline medical history. For patients not receiving anti-fibrotic medications, we

required patients to have an additional claim for the above codes for IPF to increase the

specificity of patient identification (including those patients with a single inpatient claim or at

least 2 outpatient claims on different dates).

Because there has been controversy over the specificity of billing codes used to classify the

disease in prior epidemiological analyses, a local cohort validation study using chart and imaging

examination by expert reviewers was performed by a separate set of clinical pulmonologists to

find the most specific billing codes (unpublished data), which were found to be ICD-9 516.31

and ICD-10 J84.112. Other codes (including those used in prior studies such as post-

inflammatory fibrosis [ICD-9 515, ICD-10 J84.10]) were found to be non-specific for the

diagnosis of IPF and were not included in our analysis.14

Independent Variables: Independent variables of interest at index included age, sex,

race/ethnicity, residence region, steroid use, oxygen use, health care utilization, and

comorbidities (as captured by ICD-9 and ICD-10 codes on claims in any position in the 6 months

before index date). Comorbidity burden was assessed using the Elixhauser sum of conditions.15

6

Prior hospitalizations and pulmonologist office visits in the 6-month baseline period were

captured using medical claims. Prior steroid use was defined as having a filled prescription

within 6 months prior to the index date. Oxygen use was identified using Healthcare Common

Procedure Coding System indicating oxygen supplies during the baseline period.

Follow-up: Follow-up started from the index date and continued until the end of treatment, end

of enrollment in health insurance plan, death, or end of study period (March 1, 2018). End of

treatment was defined as discontinuation of medication (not refilling a prescription within 30

days of the end of supply).

Study Outcomes: The primary outcome was all-cause mortality. Mortality was identified based

on the Social Security Death Master file and discharge status of the expired.16 The secondary

outcome was all-cause hospitalizations. Subgroup analysis was performed comparing

differences in all-cause mortality by drug.

Statistical Analysis: We used propensity score matching to balance the differences in baseline

characteristics between the treated and untreated cohorts. A propensity score was estimated using

logistic regression based on sociodemographic characteristics, medical history, prior steroid use,

year of index, oxygen use, and utilization. Specifically, we used one-to-one nearest-neighbor

caliper matching to match patients based on the logit of the propensity score using a caliper equal

to 0.2 of the standard deviation of the logit of the propensity score.17 We evaluated the

standardized difference to assess the balance of covariates after matching, and a standardized

difference less than 10% was considered acceptable.18 Cox proportional hazards regression was

then used to compare patients treated to those untreated for mortality and hospitalizations in the

propensity score-matched cohort, with robust sandwich estimates to account for the clustering

7

within matched sets.19 All analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC)

and Stata version 14.1 (StataCorp).

Sensitivity Analyses: Subgroup analyses for mortality were performed among those treated

stratified by index drug for drug-by-drug comparisons. In this analysis, if a covariate was not

balanced, we examined whether including it in the regression affected results. Falsification

endpoint analyses were performed to test for residual confounding; two endpoints were selected:

acute myocardial infarction and fracture.

Lastly, we used inverse probability of treatment weighting instead of propensity score matching

to repeat the main analyses. Specifically, we used a weight of 1/propensity score for patients

receiving a treatment and 1/ (1 − propensity score) for untreated patients.20

RESULTS

Baseline Characteristics

A total of 1255 treated and 6843 untreated adults with IPF were identified. The final propensity-

matched cohort included 1255 matched pairs of patients with IPF treated or untreated. Baseline

characteristics were well balanced between the two groups with standardized differences less

than 10% as seen in Table 1. Mean duration of follow-up was 9.93 (standard deviation [SD]

9.35) months. The mean age of the treated and untreated cohorts was 72.2 and 72.4 years,

respectively. The percentage of female patients was 36.9 in the untreated cohort and 37.1 in the

treated group. Roughly half of patients in the untreated group (48.8 percent) received steroids in

the six months prior to diagnosis compared to 49.6 percent of patients in the treated cohort, who

received steroids in the six months prior to treatment initiation. Between both groups, 58.5

percent of individuals required home oxygen supplementation. The most frequent co-morbidities

8

by diagnostic codes in the overall cohort were hypertension (66.4 percent), other chronic

pulmonary conditions (62.8 percent), and diabetes (32.9 percent).

The baseline characteristics of the subgroup analysis of patients treated with pirfenidone (n=593)

compared to nintedanib (n=662) are shown in Supplementary Table E1. There were no baseline

differences between patients taking the two medications in age, gender, steroid use, oxygen use,

or co-morbidities even without propensity score matching. The mean duration of treatment for

both medications was also similar (pirfenidone mean duration=241.32 days; nintedanib mean

duration=228.86 days).

All-Cause Mortality and Acute Hospitalizations

For the primary outcome of all-cause mortality, there was an association with reduced all-cause

mortality for those in the treated cohort compared to the untreated matched cohort (treated

cohort: 13.78 per 100 person-years versus untreated cohort: 16.34 per 100 person-years; Hazard

Ratio [HR] 0.77; 95% confidence interval [CI], 0.62 to 0.98; p value=0.034). This mortality

benefit was only present through the first two years of follow-up as seen in Figure 1A. For the

secondary outcome of acute hospitalizations, there was also a decrease in the treated cohort as

seen in Figure 1B (46.70 per 100 person-years versus 62.44 per 100 person-years; HR 0.70; 95%

confidence interval, 0.61 to 0.80; p value<0.001). The most common reason for hospitalization

identified in this cohort was respiratory failure (Supplementary Table E2).

In subgroup analysis, we observed no significant difference in all-cause mortality between

patients treated with pirfenidone and those on nintedanib as seen in Figure 2 (pirfenidone 12.83

per 100 person-years versus nintedanib 14.77 per 100 person-years; HR 1.14; 95% CI, 0.79 to

1.65; p=0.471).

9

All sensitivity analyses performed to confirm our primary findings showed results similar to the

main analysis (Supplementary Table E3). The reasons for censoring are provided in

Supplementary Table E4.

DISCUSSION

This is the first real-world study of the clinical effectiveness of the anti-fibrotic medications

pirfenidone and nintedanib for patients with IPF since their approval in 2014. There are a

number of important findings from this cohort analysis, most notably the association with an all-

cause mortality benefit observed in the cohort treated with antifibrotic medications.

All-cause mortality has been proposed as the most clinically meaningful endpoint for both

patients and clinicians when considering treatment of IPF.21 Given the relative rarity of the

disease and its progressive and ultimately fatal nature, powering a trial and ensuring appropriate

follow-up to demonstrate a mortality effect is incredibly difficult.22-24 Despite this, even before

pirfenidone and nintedanib were approved, there was debate about whether the use of surrogate

endpoints in the phase three trials of the drugs was appropriate.21,25 While the FDA ultimately

deemed the primary outcome of the trials and their findings strong enough for regulatory

approval, questions have remained as to whether the trial data translate into meaningful results in

every day clinical practice.

In this large cohort study of privately insured and Medicare Advantage patients, we have

attempted to start answering those questions by supplementing the trial data with real-world

evidence. We observed that patients treated with either pirfenidone or nintedanib had an

association with decreased risk of all-cause mortality and acute hospitalizations compared to a

matched cohort of untreated patients. This helps complement both the initial trial data for the

10

medications as well as more recent studies, which have largely relied on pooled analyses of trial

data.4,5,7,8,26-28 Impressively, this benefit was found in a population that was older and appeared to

have more severe disease than those studied in the randomized trials, which enrolled patients

with mostly mild to moderate IPF. In our cohort, more than 58 percent of patients in both the

treated and non-treated groups required home oxygen supplementation. In ASCEND, the

number of patients requiring oxygen was around 28 percent.4 All-cause mortality in our cohort

was 14.85 percent at 52 week follow-up, while in the two trials combined all-cause mortality at

one year in the placebo and treatment arms ranged from 5.6 to 6.5 percent. At least in this

cohort, it would appear physicians were using pirfenidone and nintedanib in patients with more

severe disease who would have been excluded from the RCTs. Despite those differences in

severity, these medications are improving outcomes in patients, such as in our study.

When comparing the effectiveness of the two medications at reducing all-cause mortality, there

was no difference; however, as with the published trial data, pirfenidone had a slightly more

favorable trend. Comparing the demographics of patients treated with either medication (even

without matching), the populations had nearly identical age, race, and co-morbidity profiles.

Given these similarities, it would seem to corroborate the belief that physicians are currently

deciding which medication to use based largely on shared-decision making discussions

surrounding factors such as side effects and cost considerations. Due to the lack of significant

clinical differences observed in this cohort, physicians should continue to solicit patient input

into which medication they prefer for initial treatment while awaiting further confirmation of

these results. Additional comparisons are also needed on patient-important factors such as

quality of life, symptom control, medication tolerability, and exercise tolerance.

11

One of the more intriguing findings from this study is the decline in the benefit of therapy over

time. This is consistent with prior pooled data and survival modeling for pirfenidone suggesting

a mortality benefit for up to two years.7,26 The reason for this later decline in efficacy is

currently unclear. These medications are not curative, meaning that fibrosis continues to

accumulate over time, as do clinical events such as acute exacerbations. One hypothesis is that

as lung fibrosis progresses and vascular remodeling occurs, deposition of the drugs in the lungs

decreases.29-31 Another possibility is that acute exacerbations occur despite drug therapy, which

prior research has shown portends a worse prognosis.32,33 In pooled analyses of trial data and

indirect comparison via network meta-analysis, pirfenidone consistently has the stronger

mortality benefit, while nintedanib has a stronger impact on acute exacerbations.3,7,34 Given these

findings, it is possible that a sequential treatment model utilizing pirfenidone initially and then

switching to nintedanib following the first acute exacerbation would help prevent this decreased

efficacy over time.35 In addition, combination therapy, which has been effective in other

pulmonary diseases like asthma, emphysema, and pulmonary hypertension, may also help sustain

the mortality benefit beyond two years. Thus far, studies have shown that combination therapy

is tolerable and possibly more efficacious than single drug treatment.35-38 Further research should

aim to discover the etiology for the reduction in clinical benefit over time as well as methods to

best attenuate this finding. Given the reported cost of the medications, if combination therapy is

to be considered, studies should also focus on out-of-pocket costs to patients as well as the

overall cost effectiveness of these agents.

Despite the large cohort and key observations of this analysis, there are several important

limitations of this study to consider. First, the dataset analyzed only includes patients enrolled in

private and Medicare Advantage health plans with pharmaceutical benefits, which limits the

12

generalizability of this study to other groups of patients such as the uninsured or those enrolled in

other government health plans such as Medicaid. These populations may have a different

prevalence of risk factors as well as lower socio-economic status than individuals with private or

Medicare Advantage coverage, which could lead to worse outcomes. These differences may be

mitigated in our cohort by the fact that insurance coverage rates are generally higher in older

Americans, the group most likely to be affected by IPF, thus making our findings largely

generalizable to most patients with the disease.39 Second, despite adequate matching and

adjustments across groups, confounding factors are still a distinct possibility given this study’s

observational nature. Additional sensitivity tests were undertaken and demonstrated similar

results to the main analysis, which further help to validate our main findings and to moderate

some of the concerns over confounding factors. Third, the treated cohort was identified by date

of first fill of a prescription for either anti-fibrotic medication. We acknowledge that filling a

prescription does not mean that a patient actually took the medication as prescribed. However,

the individuals in our cohort continued to fill their prescriptions; otherwise, they were censored.

As subsequent fills were tracked, this makes significant lack of adherence unlikely.

Additionally, the possibility that those in the treated cohort did not use the anti-fibrotics as

intended would presumably strengthen the observations of our analysis, as we would expect the

outcomes to further trend toward the treated cohort with stricter adherence.

A fourth limitation of our study is the use of the Social Security Death Master File to obtain the

primary outcome of mortality. Due to various circumstances and policies, since 2011, this

database has been limited in its ability to capture up to one-third of deaths.40 We attempted to

minimize this gap by using discharge status (i.e. in-hospital death) to supplement the Social

Security Death Master File as the majority of deaths occur in a hospital setting. Using this

13

approach, an additional 30 percent of deaths are typically captured in addition to those obtained

through the Death Master File. While this accounts for the majority of deaths missed by the

Death Master File, we acknowledge that a small proportion of patients that die outside of the

hospital setting could be missing, though this number is likely not high enough to influence our

overall observations.

Finally, billing codes were used to identify patients with IPF. There are a multitude of factors

that make IPF a complex diagnostic decision including the need for a multidisciplinary team, the

importance of considering other fibrotic interstitial lung diseases, and the ability to diagnose

based solely on radiographs without tissue pathology in some cases.41 Such qualities make the

use of billing codes prone to misidentification, something prior epidemiological studies in IPF

have acknowledged.42-45 We attempted to mitigate this limitation by using a local cohort

validation to identify and use only the most accurate codes for IPF (unpublished data). We also

required two outpatient codes for IPF to exclude the possibility of an evolving diagnosis after the

completion of further diagnostic testing. Even with these steps aimed at ensuring cohort

specificity, the possibility for misdiagnosis or miscoding is still possible.

In summary, in this real-world analysis of the clinical impact of the anti-fibrotic medications

pirfenidone and nintedanib, patients on treatment for idiopathic pulmonary fibrosis had an

association with a reduced risk of all-cause mortality and acute hospitalizations compared to

untreated matched individuals. Compared head-to-head, there was no difference in all-cause

mortality between the medications. Further research is needed to test the hypothesis that these

treatments reduce early but not long-term mortality as demonstrated in our study.

14

REFERENCES:

1. Raghu G, Collard HR, Egan JJ, Martinez FJ, Behr J, Brown KK, et al;

ATS/ERS/JRS/ALAT Committee on Idiopathic Pulmonary Fibrosis. An official

ATS/ERS/JRS/ALAT clinical practice guideline: treatment of idiopathic pulmonary

fibrosis. An update of the 2011 clinical practice guideline, Am J Respir Crit Care Med

2015;192:e3-19.

2. Raghu G, Behr J, Brown KK, Egan JJ, Kawut SM, Flaherty KR, Martinez FJ, et al;

ARTEMIS-IPF Investigators. Treatment of idiopathic pulmonary fibrosis with

ambrisentan: a parallel, randomized trial. Ann Intern Med 2013;158:641-649.

3. Noth I, Anstrom KJ, Calvert SB, de Andrade J, Flaherty KR, Glazer C, et al; Idiopathic

Pulmonary Fibrosis Clinical Research Network. A placebo-controlled randomized trial of

warfarin in idiopathic pulmonary fibrosis. Am J Respir Crit Care Med 2012;186:88-95.

4. King TE Jr, Bradford WZ, Castro-Bernardini S, Fagan EA, Glaspole I, Glassberg MK, et

al.; ASCEND Study Group. A phase 3 trial of pirfenidone in patients with idiopathic

pulmonary fibrosis. N Engl J Med 2014;370:2083–2092.

5. Richeldi L, du Bois RM, Raghu G, Azuma A, Brown KK, Costabel U, et al.; INPULSIS

Trial Investigators. Efficacy and safety of nintedanib in idiopathic pulmonary fibrosis. N

Engl J Med 2014;370:2071-82.

6. Hunninghake GM. A new hope for idiopathic pulmonary fibrosis. N Engl J Med

2014;370:2142-2143.

7. Nathan SD, Albera C, Bradford WZ, Costabel U, Glaspole I, Glassberg MK, et al. Effect

of pirfenidone on mortality: pooled analyses and meta-analyses of clinical trials in

idiopathic pulmonary fibrosis. Lancet Respir Med 2017;5:33-41.

15

8. Richeldi L, Cottin V, du Bois RM, Selman M, Kimura T, Bailes Z, Schlenker-Herceg R,

Stowasser S, Brown KK. Nintedanib in patients with idiopathic pulmonary fibrosis:

combined evidence from the TOMORROW and INPULSIS trials. Respir Med

2016;113:74-79.

9. Rogliani P, Calzetta L, Cavalli F, Matera MG, Cazzola M. Pirfenidone, nintedanib and

N-acetylcysteine for the treatment of idiopathic pulmonary fibrosis: a systematic review

and meta-analysis. Pulm Pharmacol Ther 2016;40:95-103

10. Loveman E, Copley VR, Scott DA, Colquitt JL, Clegg AJ, O’Reilly KM. Comparing new

treatments for idiopathic pulmonary fibrosis-a network meta-analysis. BMC Pulm Med

2015;15:37.

11. Canestaro WJ, Forrester SH, Raghu G., Ho L, Devine BE. Drug treatment of idiopathic

pulmonary fibrosis: systematic review and network meta-analysis. Chest 2016;149;756–

766.

12. Dempsey TM, Sangaralingham LR, Yao X, Limper AH. Mortality effect of the anti-

fibrotics pirfenidone and nintedanib in clinical practice [abstract]. Am J Respir Crit Care

Med 2019; 18639.

13. Wallace PJ, Shah ND, Dennen T, Bleicher PA, Crown WH. Optum Labs:building a novel

node in the learning health care system. Health Aff 2014;33:1187-1194.

14. Vu A, Vasireddy A, Moua T, Baqir M, Ryu J. Clarifying the Diagnosis of Post-

Inflammatory Pulmonary Fibrosis: A Population-based Study. Eur Respir J. (in press)

15. Quan H, Sundararajan V, Halfon P et al. Coding algorithms for defining comorbidities in

ICD-9-CM and ICD-10 administrative data. Med Care 2005;43:1130-9.

16

16. Yao X, Gersh BJ, Holmes DR, Jr. et al. Association of Surgical Left Atrial Appendage

Occlusion With Subsequent Stroke and Mortality Among Patients Undergoing Cardiac

Surgery. JAMA 2018;319:2116-2126.

17. Austin PC. Optimal caliper widths for propensity-score matching when estimating

differences in means and differences in proportions in observational studies. Pharm Stat

2011;10:150-61

18. Austin PC. Balance diagnostics for comparing the distribution of baseline covariates

between treatment groups in propensity-score matched samples. Stat Med 2009;28:3083-

107.

19. Gayat E, Resche-Rigon M, Mary JY, Porcher R. Propensity score applied to survival data

analysis through proportional hazards models: a Monte Carlo study. Pharm Stat

2012;11:222-9

20. Li F, Morgan KL, Zaslavsky AM. Balancing covariates via propensity score weighting. J

Am Stat Assoc 2018;113:390-400.

21. Raghu G, Collard HR, Anstrom KJ, Flaherty KR, Fleming TR, King TE Jr, et al.

Idiopathic pulmonary fibrosis: clinically meaningful primary endpoints in phase 3 clinical

trials. Am J Respir Crit Care Med 2012;185:1044-1048.

22. King TE Jr, Albera C, Bradford WZ, Costabel U, du Bois RM, Leff JA, et al. All-cause

mortality rate in patients with idiopathic pulmonary fibrosis. Implications for the design

and execution of clinical trials. Am J Respir Crit Care Med 2014;189:825-831.

23. Bradford WZ, Cohen AH, Leff JA. Selection of clinically meaningful primary endpoints

in phase 3 clinical trials in idiopathic pulmonary fibrosis [letter]. Am J Respir Crit Care

Med 2013;187:1269-1270.

17

24. Wells AU, Behr J, Costabel U, Cottin V, Poletti V, Richeldi L. Hot off the breath:

mortality as a primary end-point in IPF treatment trials: the best is the enemy of the good.

Thorax. 2012;67(11):938-940.

25. Du Bois RM, Nathan SD, Richeldi L, Schwarz MI, Noble PW. Idiopathic pulmonary

fibrosis: lung function is a clinically meaningful endpoint for phase 3 trials. Am J Respir

Crit Care Med 2012;186:712-715.

26. Fisher M, Nathan SD, Hill C, Marshall J, Dejonckheere F, Thuresson PO, et al.

Predicting life expectancy for pirfenidone in idiopathic pulmonary fibrosis. J Manag Care

Spec Pharm 2017;23(3-b Suppl):S17-S24.

27. Jo HE, Glaspole I, Grainge C, Goh N, Hopkins PM, Moodley Y, et al. Baseline

characteristics of idiopathic pulmonary fibrosis: analysis from the Australian Idiopathic

Pulmonary Fibrosis Registry. Eur Respir J 2017;49:1601592.

28. Lancaster L, Crestani B, Hernandez P, Inoue Y, Wachtlin D, Loaiza L, Quaresma M, et

al. Safety and survival data in patients with idiopathic pulmonary fibrosis treated with

nintedanib: pooled data from six clinical trials. BMJ Open Respiratory Research

2019;6:e000397.

29. Spagnolo P. Pirfenidone and mortality in idiopathic pulmonary fibrosis. Lancet Respir

Med 2017;5:3-5.

30. Barratt S, Millar A. Vascular remodelling in the pathogenesis of idiopathic pulmonary

fibrosis. QJM. 2014;107:515-519.

31. Plantier L, Cazes A, Dinh-Xuan A-T, Bancal C, Marchand-Adam S, Crestani B.

Physiology of the lung in idiopathic pulmonary fibrosis. Eur Respir Rev 2018;27:170062.

18

32. Song JW, Hong SB, Lim CM, Koh Y, Kim DS. Acute exacerbation of idiopathic

pulmonary fibrosis: incidence, risk factors and outcome. Eur Respir J 2011;37:356-363.

33. Kondoh Y, Taniguchi H, Katsuta T, Kataoka K, Kimura T, Nishiyama N, et al. Risk

factors of acute exacerbation of idiopathic pulmonary fibrosis. Sarcoidosis Vasc Diffuse

Lung Dis 2010;27:103-110.

34. Costabel U, Inoue Y, Richeldi L, Collard HR, Tschoepe I, Stowasser S, Azuma A.

Efficacy of nintedanib in idiopathic pulmonary fibrosis across prespecified subgroups in

INPULSIS. Am J Respir Crit Care Med 2016;193:178-185.

35. Wuyts WA, Antoniou KM, Borensztajn K, Costabel U, Cottin V, et al. Combination

therapy: the future of management for idiopathic pulmonary fibrosis? Lancet Respir. Med

2014;2:933-942.

36. Hagmeyer L, Treml M, Priegnitz C, Randerath WJ. Successful Concomitant Therapy

with Pirfenidone and Nintedanib in Idiopathic Pulmonary Fibrosis: A Case Report.

Respiration 2016; 91:327-332.

37. Vancheri C, Kreuter M, Richeldi L, Ryerson CJ, Valeyre D, Grutters JC, et al.

Nintedanib with add-on pirfenidone in idiopathic pulmonary fibrosis: results of the

INJOURNEY Trial. Am J Respir Crit Care Med 2018;197:356-363.

38. Flaherty KR, Fell CD, Huggins JT, Nunes H, Sussman R, Valenzuela C, et al. Safety of

nintedanib added to pirfenidone treatment for idiopathic pulmonary fibrosis. Eur Resp J

2018; 52:1800230.

39. Barnett JC, Vornovitsky MS for the US Census Bureau. Health insurance coverage in the

United States: 2015. Washington D.C. U.S. Government Printing Office; 2016;P60-257

(RV).

19

40. da Graca B, Filardo G, Nicewander D. Consequences for healthcare quality and research

of the exclusion of records from the Death Master File. Circ Cardiovasc Qual Outcomes

2013;6:124-128.

41. Moua T, Ryu JH. Obstacles to early treatment of idiopathic pulmonary fibrosis: current

perspectives. Ther Clin Risk Manag 2019;15:73-81.

42. Raghu G, Chen SY, Hou Q, Yeh WS, Collard HR. Incidence and prevalence of idiopathic

pulmonary fibrosis in US adults 18–64 years old. Eur Respir J 2016; 48:179-186.

43. Raghu G, Weycker D, Edelsberg J, Bradford WZ, Oster G. Incidence and prevalence of

idiopathic pulmonary fibrosis. Am J Respir Crit Care Med 2006;174:810-816.

44. Collard HR, Ward AJ, Lanes S, Hayflinger CD, Rosenberg DM, Hunsche E. Burden of

illness in idiopathic pulmonary fibrosis. J Med Econ 2012;15:829-835.

45. Collard HR, Chen SY, Yeh WS, Li Q, Lee YC, Wang A, Raghu G. Health care

utilization and costs of idiopathic pulmonary fibrosis in U.S. Medicare beneficiaries aged

65 years and older. Ann Am Thorac Soc 2015;12:981-987.

20

FIGURE LEGENDS:

Figure 1A: Mortality Cumulative Risk Percent. The cumulative risk of all-cause mortality in

patients on treatment for idiopathic pulmonary fibrosis (treated) compared to an untreated IPF

matched cohort (untreated) during 24-month follow-up. For the number of patients at risk during

each time interval, please see Supplementary Table E5.

Figure 1B. Hospitalization Cumulative Risk Percent. The cumulative risk of acute

hospitalizations in patients on treatment for idiopathic pulmonary fibrosis (treated) compared to

an untreated IPF matched cohort (untreated) during 24-month follow-up. For the number of

patients at risk during each time interval, please see Supplementary Table E5.

Figure 2. Comparison of Mortality Risk by Drug. The cumulative risk of all-cause mortality in

patients on pirfenidone compared to those on nintedanib during 24-month follow-up. For the

number of patients at risk during each time interval, please see Supplementary Table E5.

21

TABLES: Table 1. Baseline Demographics of Patients with Idiopathic Pulmonary Fibrosis Before and After Propensity Score Matching

Non-matched 1:1 Matched

No-treatment

(N=6843) Treated

(N=1255) Std. Diff

No treatment (N=1255)

Treated (N=1255)

Std. Diff

Age, years

Mean (SD) 72.9 (10.8) 72.2 (8.6) 1.6% 72.4 (9.2) 72.2 (8.6) 0.5%

Median 74 73 73 73

Age group

18-54 417 (6.1%) 42 (3.3%) 13.3% 43 (3.4%) 42 (3.3%) 0.6%

55-64 885 (12.9%) 172 (13.7%) 2.4% 176 (14.0%) 172 (13.7%) 0.9%

65-74 2131 (31.1%) 507 (40.4%) 19.5% 501 (39.9%) 507 (40.4%) 1.0%

75+ 3410 (49.8%) 534 (42.5%) 14.7% 535 (42.6%) 534 (42.5%) 0.2%

Sex

Female 3364 (49.2%) 465 (37.1%) 24.6% 463 (36.9%) 465 (37.1%) 0.4%

Male 3479 (50.8%) 790 (62.9%) 24.6% 792 (63.1%) 790 (62.9%) 0.4%

Census Region

MIDWEST 1901 (27.8%) 372 (29.6%) 4.0% 352 (28.0%) 372 (29.6%) 3.5%

NORTHEAST 988 (14.4%) 187 (14.9%) 1.4% 182 (14.5%) 187 (14.9%) 1.1%

SOUTH 3301 (48.3%) 590 (47.0%) 2.6% 596 (47.5%) 590 (47.0%) 1.0%

WEST 653 (9.5%) 106 (8.4%) 3.9% 125 (10.0%) 106 (8.4%) 5.5%

Race

Asian 190 (2.8%) 36 (2.9%) 0.6% 39 (3.1%) 36 (2.9%) 1.2%

Black 836 (12.2%) 136 (10.8%) 4.4% 141 (11.2%) 136 (10.8%) 1.3%

Hispanic 893 (13.0%) 131 (10.4%) 8.1% 129 (10.3%) 131 (10.4%) 0.3%

22

Unknown 335 (4.9%) 50 (4.0%) 4.4% 53 (4.2%) 50 (4.0%) 1.0%

White 4589 (67.1%) 902 (71.9%) 10.4% 893 (71.2%) 902 (71.9%) 1.6%

Steroid Use (6m baseline) 2837 (41.5%) 623 (49.6%) 16.3% 612 (48.8%) 623 (49.6%) 1.6%

Pulmonary Office Visit 3646 (53.3%) 1045 (83.3%) 68.1% 1043 (83.1%) 1045 (83.3%) 0.5%

DME, Oxygen 2756 (40.3%) 738 (58.8%) 37.7% 731 (58.2%) 738 (58.8%) 1.2%

Conditions

Cardiac Arrhythmia 2241 (32.7%) 339 (27.0%) 12.5% 346 (27.6%) 339 (27.0%) 1.4%

Congestive Heart Failure 2016 (29.5%) 253 (20.2%) 21.6% 264 (21.0%) 253 (20.2%) 2.0%

Chronic Pulmonary Disease 4313 (63.0%) 800 (63.7%) 1.5% 778 (62.0%) 800 (63.7%) 3.5%

Depression 903 (13.2%) 149 (11.9%) 3.9% 137 (10.9%) 149 (11.9%) 3.2%

Diabetes 2237 (32.7%) 412 (32.8%) 0.2% 415 (33.1%) 412 (32.8%) 0.6%

Hypertension 4734 (69.2%) 827 (65.9%) 7.1% 842 (67.1%) 827 (65.9%) 2.5%

Pulmonary Circulation Disorder

1357 (19.8%) 261 (20.8%) 2.5% 270 (21.5%) 261 (20.8%) 1.7%

Renal Failure 1195 (17.5%) 153 (12.2%) 15.0% 149 (11.9%) 153 (12.2%) 0.9%

Rheumatoid Arthritis 1072 (15.7%) 106 (8.4%) 22.6% 113 (9.0%) 106 (8.4%) 2.1%

Solid Tumor without Metastasis

745 (10.9%) 95 (7.6%) 11.4% 96 (7.6%) 95 (7.6%) 0.0%

Valvular Disease 1370 (20.0%) 247 (19.7%) 0.8% 238 (19.0%) 247 (19.7%) 1.8%

Elixhauser Sum of Conditions

Mean (SD) 4.7 (2.9) 4.0 (2.5) 3.4% 4.0 (2.5) 4.0 (2.5) 0%

Median 4 4 4 4

Smoker 2390 (34.9%) 534 (42.5%) 15.6% 520 (41.4%) 534 (42.5%) 2.2%

Baseline hospitalizations

0 4568 (66.8%) 865 (68.9%) 4.5% 892 (71.1%) 865 (68.9%) 4.8%

23

1 1576 (23.0%) 295 (23.5%) 1.2% 281 (22.4%) 295 (23.5%) 2.6%

2+ 699 (10.2%) 95 (7.6%) 9.1% 82 (6.5%) 95 (7.6%) 4.3%