Abstract
Rationale
Observational studies report a significant protective effect of antifibrotics on mortality among patients with idiopathic pulmonary fibrosis (IPF). Many of these studies, however, were subject to immortal time bias because of the mishandling of delayed antifibrotic initiation.
Objectives
To evaluate the antifibrotic effect on mortality among patients with IPF using appropriate statistical methods that avoid immortal time bias.
Methods
Using a large administrative database, we identified 10,289 patients with IPF, of whom 2,300 used antifibrotics. Treating delayed antifibrotic initiation as a time-dependent variable, three statistical methods were used to control baseline characteristics and avoid immortal time bias. Stratified analysis was performed for patients who initiated antifibrotics early and those who initiated treatment late. For comparison, methods that mishandle immortal time bias were performed. A simulation study was conducted to demonstrate the performance of these models in a wide range of scenarios.
Results
All three statistical methods yielded nonsignificant results for the antifibrotic effect on mortality, with the stratified analysis for patients with early antifibrotic initiation suggesting evidence for reduced mortality risk (for all patients, hazard ratio, 0.89; 95% confidence interval, 0.79–1.01; P = 0.08; for patients who were 65 years or older, hazard ratio, 0.85; 95% confidence interval, 0.73–0.98; P = 0.03). Methods that mishandle immortal time bias demonstrated significantly lower mortality risk for antifibrotic users. Bias of these methods was evident in the simulation study, where appropriate methods performed well with little to no bias.
Conclusions
Findings in this study did not confirm an association between antifibrotics and mortality, with a stratified analysis showing support for a potential treatment effect with early treatment initiation.
Keywords: antifibrotics, IPF, mortality, observational study
A progressive and fatal lung disease characterized by irreversible loss of lung function, idiopathic pulmonary fibrosis (IPF) has a poor prognosis and limited treatment options. Pirfenidone and nintedanib are the only antifibrotic medications approved for the treatment of IPF (1, 2). Their effect on the reduction of lung function decline is well established in randomized clinical trials (RCTs) (3, 4). Their impact on mortality, however, has been less compelling. Meta-analysis or pooled analysis of RCTs showed inconsistent findings (5–9). More recently, observational studies have used real-world data, including patient registries, clinical data, or administrative claims, to evaluate the mortality benefits of antifibrotics. These studies consistently reported lower mortality risk for patients with IPF who used antifibrotics. Synthesizing these studies, a meta-analysis conducted in 2021 reported a substantial protective effect of antifibrotics on mortality (10). However, many of these studies were flawed because of mishandling of delayed antifibrotic initiation that leads to the so-called immortal time bias (11–13). This raises the question of the validity of the antifibrotic effect estimation on mortality.
Observational studies evaluating the association between treatment and time to an outcome of interest may be subject to immortal time bias when the follow-up period includes an interval in which the outcome event cannot occur (14). This usually happens when there is a delay between cohort entry and treatment initiation. Patients who initiate treatment after the cohort entry must have survived the immortal time period between cohort entry and treatment initiation and are hence conferred a spurious survival advantage. In the literature, the immortal time period can be mishandled in two ways, mislabeled as the treatment period or excluded from the analysis, both of which produce distorted estimates of treatment effect (15).
Because patients with IPF often receive delayed antifibrotic treatment for reasons such as significant side effect profiles and high out-of-pocket cost (16–19), the bias described above could affect observational studies evaluating the effectiveness of antifibrotics on mortality. In a literature review, all 14 observational studies evaluating antifibrotic effect on mortality suffered from immortal time bias (11). Nine of these studies defined time zero as cohort entry and grouped patients on the basis of whether they ever received antifibrotics during the follow-up period (20–28), resulting in the mislabeling of the immortal time period between cohort entry and initiation of antifibrotics. Three of the studies required a minimal duration of antifibrotic use (21, 26, 28), intensifying the impact of immortal time bias. The remaining five studies defined time zero as treatment initiation for antifibrotic users, thereby excluding the immortal time period (29–33). The estimated antifibrotic effect on mortality in these studies is therefore biased.
In this paper, we employ statistical methods that avoid immortal time bias to evaluate the effect of antifibrotics on mortality in patients with IPF using data from a large U.S. administrative claims database. We compare these results with those derived from methods that mishandle the immortal time bias. A simulation study is performed to show the magnitude of the bias when using inappropriate methods in comparison with appropriate methods over a wide range of settings.
Methods
Data Source
A retrospective cohort study was performed using Optum’s deidentified Clinformatics Data Mart (CDM) Database derived from administrative health claims for members of large commercial and Medicare Advantage health plans. All enrollees in the CDM have both medical and prescription drug coverage. Enrollees of all ages, ethnicities, and racial groups from all 50 states are represented.
Study Population
The patient cohort was identified using the CDM database between October 2014 and March 2022. The International Classification of Diseases, Ninth Edition (ICD-9), code 516.31 and the International Classification of Diseases, Tenth Edition (ICD-10), code J84.112 validated in prior studies were used to identify patients with IPF (16, 17, 29). Prescription claims of the patients with IPF were searched to identify those who filled a prescription for nintedanib or pirfenidone between October 2014 and March 2022. For patients who never used antifibrotics, we defined the index date as the date of the first IPF encounter. For patients who used antifibrotics, the index date was defined as the earlier date of the first IPF encounter and the first fill of an antifibrotic prescription.
Patients were required to have continuous enrollment during the 12 months before the index date to ensure adequate data to capture their baseline medical history. Patients who did not use antifibrotics were required to have at least one inpatient claim or two outpatient claims with IPF diagnosis to increase the specificity of IPF patient identification. Patients with collagen vascular diseases in the baseline period were excluded. Following prior studies (34–38), we also excluded patients who had diagnosis codes for interstitial lung disease other than IPF (see Table E1 in the data supplement) on or after the date of the last IPF encounter. Also excluded were patients who were younger than 50 years of age or had missing demographics.
Independent Variables
Independent variables found to be associated with mortality in prior studies (29, 39) include demographics (age, sex, race, and region of residence). On the basis of medical services received during the 12-month baseline period, comorbidities, healthcare use, supplemental oxygen use, history of smoking, oral steroid use, and body mass index (BMI) classification were captured using ICD-9, ICD-10, Current Procedural Terminology codes, or Healthcare Common Procedure Coding System codes. Details of these variables are provided in the Supplemental Methods section of the data supplement, and specific codes used to derive these variables are provided in Table E2.
Study Outcomes
The primary study outcome was time to death since the index date. Death information in the CDM was determined on the basis of the Death Master File maintained by the Social Security Office, Center for Medicare and Medicaid Services death data, enrollment discontinuation due to death, or claims with a discharge status of death. Patients were followed from the index date for up to 2 years until the end of enrollment in health insurance plan, lung transplant, death, or end of study on March 31, 2022, whichever happened first. Those who did not die were censored at the end of their follow-up. We chose the 2-year follow-up to allow comparison with prior studies (29). The secondary study outcome was time to all-cause hospitalization, with results reported in the data supplement.
Statistical Analysis
Data analysis
Baseline characteristics were summarized using mean and standard deviation or median and interquartile range for continuous variables and frequency and proportion for categorical variables. Differences between patients who did and did not use antifibrotics during follow-up were assessed using the standardized mean difference.
To evaluate the antifibrotic effect on mortality, we considered antifibrotic use as a time-dependent variable. For patients who never used antifibrotics, it was set to zero. For patients who used antifibrotics, it was equal to zero before the antifibrotic initiation and one after the antifibrotic initiation. Baseline characteristics, including demographics, BMI class, long-term oral steroid use, pulmonary office visit, supplemental oxygen use, comorbid conditions, smoking, and baseline respiratory hospitalizations, were adjusted using three methods: multivariable adjustment, propensity score adjustment, and propensity score matching. In the two latter methods, propensity score was estimated using a Cox regression model for time to antifibrotic initiation with baseline characteristics as covariates (40). Propensity score matching was performed sequentially as detailed in the Supplemental Methods section to match patients who did and did not use antifibrotics (41). A time-dependent Cox regression model was used for both multivariable adjustment and propensity score adjustment. For the matched sample, analysis was performed using the standard Cox regression with robust standard errors, where time zero was defined as antifibrotic initiation for antifibrotic users and their matching control subjects. Using the matched sample, we performed a stratified analysis to evaluate the effect of antifibrotic use on mortality for patients with early treatment initiation (within 2 mo of the index date) and those with late treatment initiation (after 2 mo following the index date), considering that late initiators may choose to start the medication for reasons related to disease progression. A sensitivity analysis was conducted using a threshold of 6 months for the definition of early or late initiation.
For comparison, we evaluated the antifibrotic effect using methods employed in prior IPF studies that did not account for immortal time bias (see Table E3 for a summary of the studies). Patients were assigned to treated or untreated groups on the basis of whether they used antifibrotics during the follow-up, and a standard Cox regression model was used. In the first analysis, time zero was set as the index date for all patients. This method introduced immortal time bias due to mislabeling of the untreated period before antifibrotic initiation as treated. In the second analysis, time zero for antifibrotic users was redefined as antifibrotic initiation. This method introduced immortal time bias due to the exclusion of the untreated period before antifibrotic initiation. For both analyses, multivariable adjustment of the baseline characteristics and propensity score approaches (weighting and matching) were used. Propensity score models were built on the basis of logistic regression with antifibrotic use as the dependent variable and baseline characteristics as covariates. Matching was performed at the 1:1 ratio. To improve the specificity of the case definition for IPF, we repeated the above analysis for patients who were aged 65 years or older in a sensitivity analysis.
Simulation study
The simulation study was performed to investigate whether findings in our real-world study based on methods that do and do not appropriately handle immortal time bias could be generalized to a wide range of settings. Details of the simulation study are provided in the Supplemental Methods section. Following Xu and colleagues (40), we simulated data under two scenarios, assuming that some patients would never get treated in scenario I but everyone had the potential to get treated in scenario II. In both scenarios, patients might not receive treatment because they had waited too long and death (or censoring) had already occurred. Treated patients were assumed to have increased, same, or decreased mortality risk compared with untreated patients. Under each scenario with a specific true treatment effect, we simulated 100 data sets, each with a sample size of 2,000 patients. All three types of analyses, each with three methods of covariate control, were performed on each data set to estimate the treatment effect. Estimated treatment effect was presented on the basis of the regression coefficient, which is the logarithm of the hazard ratio (HR), rather than HR because of the skewed asymptotic distribution of HRs. Treatment effect was summarized using the average estimate (mean), standard deviation of the estimates, average standard error, and mean squared error.
All statistical analyses were performed using SAS 9.4 (SAS Institute Inc.) and R version 4.2.3. R packages matchit and optmatch were used for propensity score matching.
Results
Between October 2014 and March 2022, the CDM contained approximately 50 million enrollees, of whom 41,100 had at least one claim with IPF diagnosis (Figure 1). The analytic sample included 10,289 patients with IPF meeting the inclusion/exclusion criteria, of whom 2,300 patients used antifibrotics during the first 2 years after the index date. The median follow-up was 1.73 years for patients who used antifibrotics and 1.28 years for patients who did not. Baseline characteristics are shown in Table 1. Patients who used antifibrotics were younger, more often male, and more likely obese. They were also more likely to be long-term oral steroid users, to have pulmonologist office visits, and to use oxygen in the baseline period. They had fewer comorbidities and respiratory hospitalizations.
Figure 1.
Cohort selection flow chart. CDM = Clinformatics Data Mart; ILD = interstitial lung disease; IPF = idiopathic pulmonary fibrosis.
Table 1.
Baseline characteristics of study sample
| Total (N = 10,289) | Use of Antifibrotics in 2 yr |
SMD | ||
|---|---|---|---|---|
| No (n = 7,989) | Yes (n = 2,300) | |||
| Age, yr | 76.5 (8.1) | 77.3 (8.1) | 73.8 (7.4) | −0.45 |
| Age group | ||||
| 50–64 yr | 788 (7.7%) | 541 (6.8%) | 247 (10.7%) | 0.15 |
| 65–74 yr | 3,172 (30.8%) | 2,230 (27.9%) | 942 (41.0%) | 0.28 |
| 75+ yr | 6,329 (61.5%) | 5,218 (65.3%) | 1,111 (48.3%) | −0.35 |
| Sex | ||||
| Female | 4,212 (40.9%) | 3,445 (43.1%) | 767 (33.3%) | −0.20 |
| Male | 6,077 (59.1%) | 4,544 (56.9%) | 1,533 (66.7%) | 0.20 |
| Census region | ||||
| Midwest | 2,064 (20.1%) | 1,613 (20.2%) | 451 (19.6%) | −0.02 |
| Northwest | 1,190 (11.6%) | 947 (11.9%) | 243 (10.6%) | −0.04 |
| South | 4,357 (42.3%) | 3,355 (42.0%) | 1,002 (43.6%) | 0.03 |
| West | 2,678 (26.0%) | 2,074 (26.0%) | 604 (26.3%) | 0.01 |
| Race | ||||
| White | 7,322 (71.2%) | 5,700 (71.3%) | 1,622 (70.5%) | −0.02 |
| Black | 770 (7.5%) | 599 (7.5%) | 171 (7.4%) | 0.00 |
| Asian | 338 (3.3%) | 253 (3.2%) | 85 (3.7%) | 0.03 |
| Hispanic | 1,325 (12.9%) | 1,002 (12.5%) | 323 (14.0%) | 0.05 |
| Unknown | 534 (5.2%) | 435 (5.4%) | 99 (4.3%) | −0.05 |
| Body mass index class | ||||
| Underweight | 452 (4.4%) | 392 (4.9%) | 60 (2.6%) | −0.11 |
| Normal weight | 935 (9.1%) | 767 (9.6%) | 168 (7.3%) | −0.08 |
| Overweight | 1,491 (14.5%) | 1,124 (14.1%) | 367 (16.0%) | 0.05 |
| Obese | 2,475 (24.1%) | 1,808 (22.6%) | 667 (29.0%) | 0.15 |
| Unknown | 4,936 (48.0%) | 3,898 (48.8%) | 1,038 (45.1%) | −0.07 |
| Long-term oral steroid use | 1,228 (11.9%) | 884 (11.1%) | 344 (15.0%) | 0.12 |
| Pulmonary office visit | 4,643 (45.1%) | 3,210 (40.2%) | 1,433 (62.3%) | 0.45 |
| Oxygen use | 3,044 (29.6%) | 2,283 (28.6%) | 761 (33.1%) | 0.10 |
| Conditions | ||||
| Cardiac arrhythmia | 3,838 (37.3%) | 3,164 (39.6%) | 674 (29.3%) | −0.21 |
| Congestive heart failure | 3,026 (29.4%) | 2,512 (31.4%) | 514 (22.3%) | −0.20 |
| Chronic pulmonary disease | 6,427 (62.5%) | 4,962 (62.1%) | 1,465 (63.7%) | 0.03 |
| Depression | 2,088 (20.3%) | 1,663 (20.8%) | 425 (18.5%) | −0.06 |
| Diabetes | 3,684 (35.8%) | 2,862 (35.8%) | 822 (35.7%) | 0.00 |
| Hypertension | 8,016 (77.9%) | 6,301 (78.9%) | 1,715 (74.6%) | −0.10 |
| Pulmonary circulation disorder | 1,898 (18.4%) | 1,487 (18.6%) | 411 (17.9%) | −0.02 |
| Peripheral vascular disorders | 3,336 (32.4%) | 2,743 (34.3%) | 593 (25.8%) | −0.18 |
| Hypothyroidism | 2,615 (25.4%) | 2,092 (26.2%) | 523 (22.7%) | −0.08 |
| Renal failure | 2,759 (26.8%) | 2,323 (29.1%) | 436 (19.0%) | −0.23 |
| Liver disease | 880 (8.6%) | 688 (8.6%) | 192 (8.3%) | −0.01 |
| Lymphoma | 175 (1.7%) | 149 (1.9%) | 26 (1.1%) | −0.06 |
| Solid tumor without metastasis | 1,462 (14.2%) | 1,199 (15.0%) | 263 (11.4%) | −0.10 |
| Metastatic cancer | 264 (2.6%) | 241 (3.0%) | 23 (1.0%) | −0.13 |
| Valvular disease | 2,856 (27.8%) | 2,285 (28.6%) | 571 (24.8%) | −0.08 |
| Weight loss | 1,197 (11.6%) | 1025 (12.8%) | 172 (7.5%) | −0.17 |
| Deficiency anemia | 1,158 (11.3%) | 969 (12.1%) | 189 (8.2%) | −0.12 |
| Alcohol abuse | 305 (3.0%) | 258 (3.2%) | 47 (2.0%) | −0.07 |
| Coagulopathy | 877 (8.5%) | 746 (9.3%) | 131 (5.7%) | −0.13 |
| Elixhauser comorbidity index | ||||
| Mean (SD) | 5.6 (3.5) | 5.8 (3.6) | 4.9 (3.2) | −0.27 |
| Median (IQR) | 5.0 (3.0–8.0) | 5.0 (3.0–8.0) | 4.0 (2.0–7.0) | N/A |
| Smoker | 3,886 (37.8%) | 2,988 (37.4%) | 898 (39.0%) | 0.03 |
| Baseline respiratory hospitalizations | ||||
| 0 | 7,556 (73.4%) | 5,776 (72.3%) | 1,780 (77.4%) | 0.12 |
| 1 | 1,812 (17.6%) | 1,425 (17.8%) | 387 (16.8%) | −0.03 |
| 2+ | 921 (9.0%) | 788 (9.9%) | 133 (5.8%) | −0.14 |
Definition of abbreviations: IQR = interquartile range; N/A = not applicable; SD = standard deviation.
A total of 3,197 patients died during the 2-year follow-up period, of whom 525 used antifibrotics. The remaining 7,092 patients were censored: 61 at the time of lung transplant, 1,284 at the end of health plan enrollment, 1,688 at March 2022, and the remaining 4,059 at 2 years since the index date.
Antifibrotic Effect on Mortality
On the basis of the multivariable time-dependent Cox regression model, antifibrotic use was not associated with mortality (HR, 0.97; 95% confidence interval [CI], 0.88–1.07; P = 0.5) after adjusting the baseline characteristics (Table 2). Among the adjusted baseline characteristics, older age, male sex, long-term oral steroid use, oxygen use, cardiac arrhythmia, congestive heart failure, depression, diabetes, pulmonary circulation disorders, renal failure, metastatic cancer, weight loss, alcohol abuse, coagulopathy, and history of respiratory hospitalization were associated with a higher risk of mortality (Table E4). Pulmonologist office visit and greater BMI were associated with a lower risk of mortality. Results from the two propensity score methods were similar. The propensity score adjustment yielded an HR of 1.04 (95% CI, 0.94–1.14; P = 0.49). The sequential propensity score matching led to a matched sample with 1,935 antifibrotic users and 1,935 control subjects, where the baseline characteristics of the matched sample are presented in Table E5. The Kaplan-Meier curve in Figure E1 showed no difference in survival between antifibrotic users and their matching control subjects. The HR was 0.99 (95% CI, 0.88–1.12; P = 0.91), again suggesting no antifibrotic effect on mortality.
Table 2.
Estimated effect of antifibrotic use on all-cause mortality
| Estimate (95% CI) | HR (95% CI) | P Value | |
|---|---|---|---|
| Time-dependent antifibrotics use | |||
| Multivariable adjustment | −0.03 (−0.13, 0.06) | 0.97 (0.88, 1.07) | 0.5 |
| Propensity score adjustment | 0.03 (−0.06, 0.13) | 1.04 (0.94, 1.14) | 0.49 |
| Sequential propensity score matching | −0.01 (−0.13, 0.11) | 0.99 (0.88, 1.12) | 0.91 |
| Immortal time period mislabeled | |||
| Multivariable adjustment | −0.36 (−0.45, −0.26) | 0.70 (0.64, 0.77) | <0.001 |
| Propensity score weighting | −0.40 (−0.51, −0.29) | 0.67 (0.60, 0.75) | <0.001 |
| Propensity score matching | −0.29 (−0.41, −0.17) | 0.75 (0.67, 0.84) | <0.001 |
| Immortal time period excluded | |||
| Multivariable adjustment | −0.20 (−0.30, −0.10) | 0.82 (0.74, 0.91) | <0.001 |
| Propensity score weighting | −0.23 (−0.35, −0.12) | 0.79 (0.71, 0.88) | <0.001 |
| Propensity score matching | −0.13 (−0.25, −0.01) | 0.88 (0.78, 0.99) | 0.029 |
Definition of abbreviations: CI = confidence interval; HR = hazard ratio.
Covariates included in the multivariable regression and propensity score modeling are age, sex, census region, race, body mass index class, long-term oral steroid use, pulmonary office visits, oxygen use, cardiac arrythmia, congestive heart failure, depression, diabetes, hypertension, pulmonary circulation disorder, peripheral vascular disorders, hypothyroidism, renal failure, liver disease, lymphoma, solid tumor without metastasis, valvular disease, weight loss, deficiency anemia, alcohol abuse, coagulopathy, smoker, and baseline respiratory hospitalizations.
In the stratified analysis, 1,354 antifibrotic users in the matched sample initiated treatment within 2 months of their index date, and the remaining 581 antifibrotic users initiated treatment after 2 months. Compared with their matching control subjects, antifibrotic users who initiated treatment within 2 months had an 11% lower mortality risk, providing weak evidence for the effectiveness of antifibrotic treatment, although it did not reach statistical significance (HR, 0.89; 95% CI, 0.79–1.01; P = 0.08). By contrast, patients who initiated antifibrotics after 2 months had a higher mortality risk (HR, 1.35; 95% CI, 0.98–1.86; P = 0.07) than their matching control subjects.
In the sensitivity analysis, 1,695 antifibrotic users initiated treatment within 6 months of their index date and the remaining 240 antifibrotic users initiated treatment more than 6 months later. Patients who initiated antifibrotics within 6 months did not show a significantly different mortality risk compared with their matching control subjects (HR, 0.94; 95% CI, 0.83–1.05; P = 0.28), whereas those who initiated antifibrotics more than 6 months later had a significantly higher mortality risk (HR, 2.02; 95% CI, 1.21–3.38; P = 0.008).
Methods that mishandled the immortal time periods all yielded significant results (Table 2). Regardless of whether immortal time periods were mislabeled or excluded, all analyses suggested that antifibrotic use was associated with a significantly lower risk of mortality, with mislabeling of the immortal time periods producing larger effect sizes.
A sensitivity analysis on 9,501 patients who were aged 65 years or older (2,053 antifibrotic users) yielded similar results (Table E6). Sequential propensity score matching yielded a sample of 1,741 antifibrotic users and 1,741 matching control subjects. Of these, 1,224 patients initiated antifibrotics within 2 months of the index date and had a 15% lower mortality risk than their matching control subjects (HR, 0.85; 95% CI, 0.73–0.98; P = 0.03). Among the remaining 517 patients who initiated their treatment after 2 months of the index date, the mortality risk was higher relative to their matching control subjects (HR, 1.40; 95% CI, 1.06–1.85; P = 0.02).
Simulation Study
Tables 3 and 4 show the simulation results under simulation scenarios I and II, respectively. All three models incorporating the time-dependent treatment performed well across different settings with little to no bias. In comparison, models that mishandled the immortal time periods produced substantial biases, which were all in the same direction toward the protective treatment effect. For example, in simulation scenario I, when immortal time periods were mislabeled, the treatment effect was estimated to be approximately −0.8, on average, corresponding to an HR of 0.45, when treatment had no effect (true effect = 0, HR = 1). Excluding the immortal time periods produced smaller biases than mislabeling the immortal time periods, as evidenced by its estimated average treatment effect of approximately −0.2 that corresponded to an HR of 0.82.
Table 3.
Summary of treatment effect estimates in simulation scenario I, where intermediate treatment status is first generated as a binary variable and then timing of treatment initiation is simulated
| True Treatment Effect | Models | Mean | SD | SE | MSE |
|---|---|---|---|---|---|
| −0.5 | Time-dependent treatment | ||||
| Multivariable adjustment | −0.494 | 0.148 | 0.142 | 0.022 | |
| Propensity score adjustment | −0.493 | 0.148 | 0.142 | 0.022 | |
| Sequential propensity score matching | −0.469 | 0.203 | 0.189 | 0.042 | |
| Immortal time period mislabeled | |||||
| Multivariable adjustment | −1.316 | 0.142 | 0.141 | 0.685 | |
| Propensity score weighting | −1.104 | 0.152 | 0.107 | 0.388 | |
| Propensity score matching | −1.287 | 0.133 | 0.150 | 0.636 | |
| Immortal time period excluded | |||||
| Multivariable adjustment | −0.913 | 0.144 | 0.140 | 0.191 | |
| Propensity score weighting | −0.712 | 0.159 | 0.106 | 0.070 | |
| Propensity score matching | −0.883 | 0.136 | 0.148 | 0.165 | |
| −0.2 | Time-dependent treatment | ||||
| Multivariable adjustment | −0.188 | 0.131 | 0.130 | 0.017 | |
| Propensity score adjustment | −0.188 | 0.131 | 0.130 | 0.017 | |
| Sequential propensity score matching | −0.162 | 0.183 | 0.177 | 0.035 | |
| Immortal time period mislabeled | |||||
| Multivariable adjustment | −1.021 | 0.125 | 0.129 | 0.690 | |
| Propensity score weighting | −0.824 | 0.136 | 0.097 | 0.408 | |
| Propensity score matching | −1.006 | 0.123 | 0.137 | 0.665 | |
| Immortal time period excluded | |||||
| Multivariable adjustment | −0.600 | 0.126 | 0.128 | 0.176 | |
| Propensity score weighting | −0.416 | 0.141 | 0.097 | 0.066 | |
| Propensity score matching | −0.582 | 0.123 | 0.136 | 0.161 | |
| 0.0 | Time-dependent treatment | ||||
| Multivariable adjustment | 0.011 | 0.125 | 0.123 | 0.016 | |
| Propensity score adjustment | 0.011 | 0.125 | 0.123 | 0.016 | |
| Sequential propensity score matching | 0.038 | 0.173 | 0.171 | 0.031 | |
| Immortal time period mislabeled | |||||
| Multivariable adjustment | −0.832 | 0.119 | 0.123 | 0.706 | |
| Propensity score weighting | −0.642 | 0.124 | 0.092 | 0.428 | |
| Propensity score matching | −0.826 | 0.117 | 0.131 | 0.696 | |
| Immortal time period excluded | |||||
| Multivariable adjustment | −0.397 | 0.120 | 0.122 | 0.172 | |
| Propensity score weighting | −0.221 | 0.129 | 0.092 | 0.066 | |
| Propensity score matching | −0.389 | 0.118 | 0.130 | 0.165 | |
| 0.2 | Time-dependent treatment | ||||
| Multivariable adjustment | 0.209 | 0.117 | 0.118 | 0.014 | |
| Propensity score adjustment | 0.209 | 0.117 | 0.118 | 0.014 | |
| Sequential propensity score matching | 0.236 | 0.169 | 0.165 | 0.030 | |
| Immortal time period mislabeled | |||||
| Multivariable adjustment | −0.645 | 0.114 | 0.117 | 0.727 | |
| Propensity score weighting | −0.460 | 0.113 | 0.088 | 0.448 | |
| Propensity score matching | −0.648 | 0.115 | 0.125 | 0.733 | |
| Immortal time period excluded | |||||
| Multivariable adjustment | −0.196 | 0.114 | 0.116 | 0.169 | |
| Propensity score weighting | −0.025 | 0.119 | 0.087 | 0.065 | |
| Propensity score matching | −0.196 | 0.113 | 0.124 | 0.170 |
Definition of abbreviations: MSE = mean squared error; SD = standard deviation; SE = standard error.
Table 4.
Summary of treatment effect estimates in simulation scenario II, where all patients have a potential treatment initiation time
| True Treatment Effect | Models | Mean | SD | SE | MSE |
|---|---|---|---|---|---|
| −0.5 | Time-dependent treatment | ||||
| Multivariable adjustment | −0.510 | 0.135 | 0.144 | 0.018 | |
| Propensity score adjustment | −0.508 | 0.134 | 0.144 | 0.018 | |
| Sequential propensity score matching | −0.502 | 0.177 | 0.185 | 0.031 | |
| Immortal time period mislabeled | |||||
| Multivariable adjustment | −1.315 | 0.122 | 0.139 | 0.679 | |
| Propensity score weighting | −1.340 | 0.120 | 0.110 | 0.720 | |
| Propensity score matching | −1.341 | 0.151 | 0.157 | 0.730 | |
| Immortal time period excluded | |||||
| Multivariable adjustment | −0.733 | 0.127 | 0.138 | 0.070 | |
| Propensity score weighting | −0.724 | 0.127 | 0.109 | 0.066 | |
| Propensity score matching | −0.725 | 0.150 | 0.156 | 0.073 | |
| −0.2 | Time-dependent treatment | ||||
| Multivariable adjustment | −0.196 | 0.120 | 0.129 | 0.014 | |
| Propensity score adjustment | −0.195 | 0.119 | 0.129 | 0.014 | |
| Sequential propensity score matching | −0.187 | 0.169 | 0.171 | 0.029 | |
| Immortal time period mislabeled | |||||
| Multivariable adjustment | −1.022 | 0.106 | 0.123 | 0.687 | |
| Propensity score weighting | −1.051 | 0.106 | 0.101 | 0.736 | |
| Propensity score matching | −1.052 | 0.140 | 0.143 | 0.745 | |
| Immortal time period excluded | |||||
| Multivariable adjustment | −0.419 | 0.111 | 0.123 | 0.060 | |
| Propensity score weighting | −0.412 | 0.112 | 0.100 | 0.057 | |
| Propensity score matching | −0.413 | 0.138 | 0.142 | 0.064 | |
| 0.0 | Time-dependent treatment | ||||
| Multivariable adjustment | 0.004 | 0.123 | 0.121 | 0.015 | |
| Propensity score adjustment | 0.004 | 0.122 | 0.121 | 0.015 | |
| Sequential propensity score matching | 0.009 | 0.174 | 0.165 | 0.030 | |
| Immortal time period mislabeled | |||||
| Multivariable adjustment | −0.839 | 0.109 | 0.115 | 0.715 | |
| Propensity score weighting | −0.869 | 0.109 | 0.095 | 0.767 | |
| Propensity score matching | −0.869 | 0.140 | 0.136 | 0.775 | |
| Immortal time period excluded | |||||
| Multivariable adjustment | −0.220 | 0.116 | 0.114 | 0.062 | |
| Propensity score weighting | −0.213 | 0.118 | 0.095 | 0.059 | |
| Propensity score matching | −0.214 | 0.139 | 0.135 | 0.065 | |
| 0.2 | Time-dependent treatment | ||||
| Multivariable adjustment | 0.203 | 0.116 | 0.113 | 0.013 | |
| Propensity score adjustment | 0.204 | 0.115 | 0.113 | 0.013 | |
| Sequential propensity score matching | 0.205 | 0.169 | 0.159 | 0.028 | |
| Immortal time period mislabeled | |||||
| Multivariable adjustment | −0.659 | 0.103 | 0.108 | 0.748 | |
| Propensity score weighting | −0.690 | 0.103 | 0.091 | 0.802 | |
| Propensity score matching | −0.690 | 0.136 | 0.130 | 0.811 | |
| Immortal time period excluded | |||||
| Multivariable adjustment | −0.023 | 0.110 | 0.107 | 0.062 | |
| Propensity score weighting | −0.017 | 0.112 | 0.091 | 0.060 | |
| Propensity score matching | −0.018 | 0.133 | 0.129 | 0.065 |
Definition of abbreviations: MSE = mean squared error; SD = standard deviation; SE = standard error.
Discussion
On the basis of a large real-world administrative health claims database and using approaches that appropriately handle the delayed initiation of antifibrotic treatment, our study did not find an association between antifibrotic use and mortality risk in patients with IPF. Despite the null finding in the primary analysis, the stratified analysis demonstrated an HR of 0.89 (P = 0.08) for death among antifibrotic users initiating treatment within 2 months. When restricting the analysis to patients aged 65 years or older (because these are the individuals most likely to receive a diagnosis of IPF), antifibrotic users initiating treatment within 2 months had a significantly lower hazard for death (HR, 0.85; P = 0.03), providing evidence for the effectiveness of antifibrotics on mortality. The clinical implications of these findings support early initiation of antifibrotic medications in IPF rather than waiting for clinical progression.
Results in the stratified analysis and the sensitivity analysis imply that the null finding in the primary analysis may be related to time-dependent confounding that occurs after the IPF diagnosis. Through patient and pulmonologist surveys, prior research has suggested that some patients may receive delayed antifibrotic therapy until disease progression (42). Markers for disease progression, such as forced vital capacity, may therefore be time-dependent confounders because decreased forced vital capacity would impact both antifibrotic initiation and mortality. Although we are unable to adjust the time-dependent confounding effects due to the lack of detailed clinical information in administrative databases, the stratified analysis for patients with antifibrotic initiation within 2 months of the index date minimizes such confounding effect. Furthermore, we found that patients who initiated antifibrotics late had a higher mortality risk than their matching control subjects. Antifibrotic users and their control subjects were matched on the basis of baseline characteristics at the index date and the timing of the antifibrotic initiation, so the higher mortality risk for patients who initiated antifibrotics late may be attributable to their disease progression after the index date. Further investigation based on more comprehensive data is therefore warranted.
The findings of our study are different from those in many prior observational studies that consistently reported a reduced mortality risk for patients with IPF who used antifibrotics. In the meta-analysis performed by Suissa and colleagues (11), the pooled HR was 0.55 (95% CI, 0.47–0.63) using nine studies that mislabeled the immortal time period and 0.68 (95% CI, 0.61–0.77) using five studies that excluded the immortal time period. With immortal time periods mishandled in the analysis, these studies suffered from immortal time bias that distorts the treatment effect estimation, making their findings less reliable. More recently, Dempsey and colleagues (43) used time-dependent Cox regression to address the immortal time bias and found a lower mortality risk (HR, 0.69; 95% CI, 0.66–0.72; P < 0.001) associated with antifibrotic use. This study, however, censored antifibrotic users at the time of treatment stop. Because patients with IPF may stop antifibrotic therapy owing to acute exacerbation or disease progression (44), censoring at treatment stop may be correlated with death. Informative censoring is known to cause bias and result in less reliable findings.
Pooled analyses or meta-analyses of data from RCTs have shown divergent findings, with some studies showing significantly lower mortality for one antifibrotic but not the other, whereas others show neither was associated with reduced mortality (11). Nonetheless, effect sizes were all in the direction for the beneficial effect of antifibrotics. These effect sizes were based on the relatively short follow-up period of 12 to 18 months in the RCTs, similar to the 2-year follow-up period in our study. Although our primary analysis yielded a null finding, our stratified analyses for antifibrotic users who initiated medication within 2 months of their index date provided evidence for the beneficial effect of antifibrotics. This is consistent with the findings in the RCT studies.
Limitations
The results in this study should be interpreted with caution because of the following limitations. First, the case definition to identify patients with IPF was not validated using medical charts, and it is not known whether billing codes are reflective of diagnosis based on multidisciplinary discussions. Second, our analysis is based on an administrative database for patients enrolled in private and Medicare Advantage insurance plans with prescription drug coverage. Caution should be exercised when trying to generalize our findings to different groups of patients. Third, despite adjustment of baseline characteristics that were known to be associated with survival, there may still be unadjusted confounding factors, given the nature of the data. Baseline disease severity may be an important confounder that needs to be adjusted. In addition, time-dependent confounding, such as disease progression, is another important factor that may contribute to the null finding of the primary analysis. However, such information is not available in the administrative data and hence cannot be evaluated.
Conclusions
In conclusion, results in our study suggest that early initiation of antifibrotics is associated with lower mortality risk. This finding has important clinical implications because it suggests that antifibrotics should be initiated as soon as a diagnosis of IPF is made. Despite scientific evidence from RCTs that antifibrotics are effective in slowing the progression of IPF, many patients adopt a “watch and wait” approach, especially when they are diagnosed at an early stage (45), and delay their treatment until disease progresses (46). In real-world studies, the adoption rate of antifibrotics was found to be remarkably low at only 25% (16). The low adoption and delayed initiation of antifibrotics may be related to the lack of awareness of the unpredictability of the IPF disease course for both patients and clinicians, concerns of potential adverse effects, and high out-of-pocket cost. The reduced mortality risk established in our study for patients who initiated antifibrotics early provides evidence against the “wait and watch” approach and supports the early start of antifibrotic treatment.
Footnotes
Supported by National Institutes of Health grant R01 HL168126 (J.S.L.).
Author Contributions: Study conceptualization and study design: H.X. and R.D.B. Data management and statistical analysis: H.X. and Z.Z. Results interpretation, manuscript writing, and critical revisions: H.X., S.L.H., J.S.L., Z.Z., and R.D.B. All authors approved the final manuscript.
This article has a data supplement, which is accessible at the Supplements tab.
Author disclosures are available with the text of this article at www.atsjournals.org.
References
- 1. Raghu G, Rochwerg B, Zhang Y, Garcia CA, Azuma A, Behr J, et al. Latin American Thoracic Association 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–e19. doi: 10.1164/rccm.201506-1063ST. [DOI] [PubMed] [Google Scholar]
- 2. Raghu G, Remy-Jardin M, Richeldi L, Thomson CC, Inoue Y, Johkoh T, et al. Idiopathic pulmonary fibrosis (an update) and progressive pulmonary fibrosis in adults: an official ATS/ERS/JRS/ALAT clinical practice guideline. Am J Respir Crit Care Med . 2022;205:e18–e47. doi: 10.1164/rccm.202202-0399ST. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. 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;370:2071–2082. doi: 10.1056/NEJMoa1402584. [DOI] [PubMed] [Google Scholar]
- 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. doi: 10.1056/NEJMoa1402582. [DOI] [PubMed] [Google Scholar]
- 5. 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. doi: 10.1016/j.chest.2015.11.013. [DOI] [PubMed] [Google Scholar]
- 6. 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. doi: 10.1016/j.pupt.2016.07.009. [DOI] [PubMed] [Google Scholar]
- 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. doi: 10.1016/S2213-2600(16)30326-5. [DOI] [PubMed] [Google Scholar]
- 8. Lancaster L, Crestani B, Hernandez P, Inoue Y, Wachtlin D, Loaiza L, et al. Safety and survival data in patients with idiopathic pulmonary fibrosis treated with nintedanib: pooled data from six clinical trials. BMJ Open Respir Res . 2019;6:e000397. doi: 10.1136/bmjresp-2018-000397. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Di Martino E, Provenzani A, Vitulo P, Polidori P. Systematic review and meta-analysis of pirfenidone, nintedanib, and pamrevlumab for the treatment of idiopathic pulmonary fibrosis. Ann Pharmacother . 2021;55:723–731. doi: 10.1177/1060028020964451. [DOI] [PubMed] [Google Scholar]
- 10. Petnak T, Lertjitbanjong P, Thongprayoon C, Moua T. Impact of antifibrotic therapy on mortality and acute exacerbation in idiopathic pulmonary fibrosis: a systematic review and meta-analysis. Chest . 2021;160:1751–1763. doi: 10.1016/j.chest.2021.06.049. [DOI] [PubMed] [Google Scholar]
- 11. Suissa S, Suissa K. Antifibrotics and reduced mortality in idiopathic pulmonary fibrosis: immortal time bias. Am J Respir Crit Care Med . 2023;207:105–109. doi: 10.1164/rccm.202207-1301LE. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Suissa S, Assayag D. Mortality benefit with antifibrotics in idiopathic pulmonary fibrosis: real world evidence or bias? Eur Respir J . 2021;57:2004562. doi: 10.1183/13993003.04562-2020. [DOI] [PubMed] [Google Scholar]
- 13. Zheng Q, Otahal P, Cox IA, de Graaff B, Campbell JA, Ahmad H, et al. The influence of immortal time bias in observational studies examining associations of antifibrotic therapy with survival in idiopathic pulmonary fibrosis: a simulation study. Front Med (Lausanne) . 2023;10:1157706. doi: 10.3389/fmed.2023.1157706. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Yadav K, Lewis RJ. Immortal time bias in observational studies. JAMA . 2021;325:686–687. doi: 10.1001/jama.2020.9151. [DOI] [PubMed] [Google Scholar]
- 15. Suissa S, Dell’Aniello S. Time-related biases in pharmacoepidemiology. Pharmacoepidemiol Drug Saf . 2020;29:1101–1110. doi: 10.1002/pds.5083. [DOI] [PubMed] [Google Scholar]
- 16. Dempsey TM, Payne S, Sangaralingham L, Yao X, Shah ND, Limper AH. Adoption of the antifibrotic medications pirfenidone and nintedanib for patients with idiopathic pulmonary fibrosis. Ann Am Thorac Soc . 2021;18:1121–1128. doi: 10.1513/AnnalsATS.202007-901OC. [DOI] [PubMed] [Google Scholar]
- 17. Dempsey TM, Thao V, Moriarty JP, Borah BJ, Limper AH. Cost-effectiveness of the anti-fibrotics for the treatment of idiopathic pulmonary fibrosis in the United States. BMC Pulm Med . 2022;22:18. doi: 10.1186/s12890-021-01811-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Maher TM, Lancaster LH, Jouneau S, Morrison L, Lederer DJ, Molina-Molina M, et al. Pirfenidone treatment in individuals with idiopathic pulmonary fibrosis: impact of timing of treatment initiation. Ann Am Thorac Soc . 2019;16:927–930. doi: 10.1513/AnnalsATS.201810-720RL. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Maher TM, Strek ME. Antifibrotic therapy for idiopathic pulmonary fibrosis: time to treat. Respir Res . 2019;20:205. doi: 10.1186/s12931-019-1161-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. 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. doi: 10.1183/13993003.01592-2016. [DOI] [PubMed] [Google Scholar]
- 21. Kaunisto J, Salomaa ER, Hodgson U, Kaarteenaho R, Kankaanranta H, Koli K, et al. Demographics and survival of patients with idiopathic pulmonary fibrosis in the Finnish IPF registry. ERJ Open Res . 2019;5:00170-2018. doi: 10.1183/23120541.00170-2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Nathan SD, Brown AW, Mogulkoc N, Soares F, Collins AC, Cheng J, et al. The association between white blood cell count and outcomes in patients with idiopathic pulmonary fibrosis. Respir Med . 2020;170:106068. doi: 10.1016/j.rmed.2020.106068. [DOI] [PubMed] [Google Scholar]
- 23. Alhamad EH, Cal JG, Alrajhi NN, Aharbi WM, AlRikabi AC, AlBoukai AA. Clinical characteristics, comorbidities, and outcomes in patients with idiopathic pulmonary fibrosis. Ann Thorac Med . 2020;15:208–214. doi: 10.4103/atm.ATM_230_20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Adegunsoye A, Alqalyoobi S, Linderholm A, Bowman WS, Lee CT, Pugashetti JV, et al. Circulating plasma biomarkers of survival in antifibrotic-treated patients with idiopathic pulmonary fibrosis. Chest . 2020;158:1526–1534. doi: 10.1016/j.chest.2020.04.066. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Gao J, Kalafatis D, Carlson L, Pesonen IHA, Li CX, Wheelock A, et al. Baseline characteristics and survival of patients of idiopathic pulmonary fibrosis: a longitudinal analysis of the Swedish IPF Registry. Respir Res . 2021;22:40. doi: 10.1186/s12931-021-01634-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Moon SW, Kim SY, Chung MP, Yoo H, Jeong SH, Kim DS, et al. Longitudinal changes in clinical features, management, and outcomes of idiopathic pulmonary fibrosis. A nationwide cohort study. Ann Am Thorac Soc . 2021;18:780–787. doi: 10.1513/AnnalsATS.202005-451OC. [DOI] [PubMed] [Google Scholar]
- 27. Lee EG, Lee TH, Hong Y, Ryoo J, Heo JW, Gil BM, et al. Effects of low-dose pirfenidone on survival and lung function decline in patients with idiopathic pulmonary fibrosis (IPF): results from a real-world study. PLoS One . 2021;16:e0261684. doi: 10.1371/journal.pone.0261684. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Jegal Y, Park JS, Kim SY, Yoo H, Jeong SH, Song JW, et al. Korea ILD Study Group Clinical features, diagnosis, management, and outcomes of idiopathic pulmonary fibrosis in Korea: analysis of the Korea IPF Cohort (KICO) Registry. Tuberc Respir Dis (Seoul) . 2022;85:185–194. doi: 10.4046/trd.2021.0123. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Dempsey TM, Sangaralingham LR, Yao X, Sanghavi D, Shah ND, Limper AH. Clinical effectiveness of antifibrotic medications for idiopathic pulmonary fibrosis. Am J Respir Crit Care Med . 2019;200:168–174. doi: 10.1164/rccm.201902-0456OC. [DOI] [PubMed] [Google Scholar]
- 30. Behr J, Prasse A, Wirtz H, Koschel D, Pittrow D, Held M, et al. Survival and course of lung function in the presence or absence of antifibrotic treatment in patients with idiopathic pulmonary fibrosis: long-term results of the INSIGHTS-IPF registry. Eur Respir J . 2020;57:1902279. doi: 10.1183/13993003.02279-2019. [DOI] [PubMed] [Google Scholar]
- 31. Kang J, Han M, Song JW. Antifibrotic treatment improves clinical outcomes in patients with idiopathic pulmonary fibrosis: a propensity score matching analysis. Sci Rep . 2020;10:15620. doi: 10.1038/s41598-020-72607-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Leuschner G, Klotsche J, Kreuter M, Prasse A, Wirtz H, Pittrow D, et al. INSIGHTS-IPF Registry Group Idiopathic pulmonary fibrosis in elderly patients: analysis of the INSIGHTS-IPF observational study. Front Med (Lausanne) . 2020;7:601279. doi: 10.3389/fmed.2020.601279. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Wright WA, Crowley LE, Parekh D, Crawshaw A, Dosanjh DP, Nightingale P, et al. Real-world retrospective observational study exploring the effectiveness and safety of antifibrotics in idiopathic pulmonary fibrosis. BMJ Open Respir Res . 2021;8:e000782. doi: 10.1136/bmjresp-2020-000782. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. 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. doi: 10.1183/13993003.01653-2015. [DOI] [PubMed] [Google Scholar]
- 35. Ley B, Urbania T, Husson G, Vittinghoff E, Brush DR, Eisner MD, et al. Code-based diagnostic algorithms for idiopathic pulmonary fibrosis. Case validation and improvement. Ann Am Thorac Soc . 2017;14:880–887. doi: 10.1513/AnnalsATS.201610-764OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Mortimer KM, Bartels DB, Hartmann N, Capapey J, Yang J, Gately R, et al. Characterizing health outcomes in idiopathic pulmonary fibrosis using US health claims data. Respiration . 2020;99:108–118. doi: 10.1159/000504630. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Cottin V, Schmidt A, Catella L, Porte F, Fernandez-Montoya C, Le Lay K, et al. Burden of idiopathic pulmonary fibrosis progression: a 5-year longitudinal follow-up study. PLoS One . 2017;12:e0166462. doi: 10.1371/journal.pone.0166462. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Bellou V, Belbasis L, Evangelou E. Tobacco smoking and risk for pulmonary fibrosis: a prospective cohort study from the UK Biobank. Chest . 2021;160:983–993. doi: 10.1016/j.chest.2021.04.035. [DOI] [PubMed] [Google Scholar]
- 39. Ley B, Collard HR, King TE., Jr Clinical course and prediction of survival in idiopathic pulmonary fibrosis. Am J Respir Crit Care Med . 2011;183:431–440. doi: 10.1164/rccm.201006-0894CI. [DOI] [PubMed] [Google Scholar]
- 40. Xu R, Luo Y, Glynn R, Johnson D, Jones KL, Chambers C. Time-dependent propensity score for assessing the effect of vaccine exposure on pregnancy outcomes through pregnancy exposure cohort studies. Int J Environ Res Public Health . 2014;11:3074–3085. doi: 10.3390/ijerph110303074. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Lu B. Propensity score matching with time-dependent covariates. Biometrics . 2005;61:721–728. doi: 10.1111/j.1541-0420.2005.00356.x. [DOI] [PubMed] [Google Scholar]
- 42. Maher TM, Swigris JJ, Kreuter M, Wijsenbeek M, Cassidy N, Ireland L, et al. Identifying barriers to idiopathic pulmonary fibrosis treatment: a survey of patient and physician views. Respiration . 2018;96:514–524. doi: 10.1159/000490667. [DOI] [PubMed] [Google Scholar]
- 43. Dempsey TM, Thao V, Helfinstine DA, Jr, Chang YH, Sangaralingham L, Limper AH. Real-world cohort evaluation of the impact of the antifibrotics in patients with idiopathic pulmonary fibrosis. Eur Respir J . 2023;62:2301299. doi: 10.1183/13993003.01299-2023. [DOI] [PubMed] [Google Scholar]
- 44. Takehara K, Koga Y, Hachisu Y, Utsugi M, Sawada Y, Saito Y, et al. Differential discontinuation profiles between pirfenidone and nintedanib in patients with idiopathic pulmonary fibrosis. Cells . 2022;11:143. doi: 10.3390/cells11010143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Sugino K, Ono H, Watanabe N, Ando M, Tsuboi E, Homma S, et al. Efficacy of early antifibrotic treatment for idiopathic pulmonary fibrosis. BMC Pulm Med . 2021;21:218. doi: 10.1186/s12890-021-01595-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Torrisi SE, Pavone M, Vancheri A, Vancheri C. When to start and when to stop antifibrotic therapies. Eur Respir Rev . 2017;26:170053. doi: 10.1183/16000617.0053-2017. [DOI] [PMC free article] [PubMed] [Google Scholar]