Effectiveness and Tolerability of Antifibrotics in Rheumatoid Arthritis-associated Interstitial Lung Disease

Abstract

Objective:

Our aim was to investigate the effectiveness and tolerability of antifibrotics in a real-world cohort of patients with rheumatoid arthritis-associated interstitial lung diseases (RA-ILD).

Methods:

In this retrospective cohort study, we identified RA-ILD patients initiating antifibrotics at Mass General Brigham Integrated Health Care System, a large multi-hospital healthcare system in Boston, MA, USA. We used electronic query to identify all patients with at least 2 RA diagnosis codes and a prescription for either nintedanib or pirfenidone (2014–2023). All analyzed patients met 2010 American College of Rheumatology/European Alliance of Associations for Rheumatology classification criteria for RA and had definite RA-ILD according to Bongartz criteria. Data regarding pulmonary function test (PFT) results, adverse events (AEs), tolerability, and clinical data were collected. A linear mixed model with random intercept was used to compare the within-patient trajectory of the percent predicted forced vital capacity (FVCpp) within 18-months before to 18-months after antifibrotic initiation among those with these PFT data. Lung transplant-free survival and drug retention was estimated in a Kaplan-Meier analysis and a Cox regression analysis was performed to identify independent baseline factors associated with lung transplant or mortality.

Results:

We analyzed 74 patients with RA-ILD that initiated antifibrotics (mean age 67.8 years, 53% male); 40 patients initiated nintedanib and 34 initiated pirfenidone. Median follow-up was 89 weeks (min 4, max 387). There was a significant improvement in the estimated slope of FVCpp after antifibrotic initiation (−0.3% per year after initiation compared to −6.2% per year before antifibrotic initiation, p=0.03). Nintedanib and pirfenidone had similar FVCpp trajectory. Twenty-six patients (35%) died and 4 (5%) had undergone lung transplantation during follow-up. Male sex and heavy smoking were each associated with the composite outcomes of lung transplant or mortality. AEs were reported in 41 (55%) patients, with gastrointestinal (GI) AEs being most common (n=30). The initial antifibrotic was discontinued in 34 (46%) patients mostly due to GI AEs (n=19). The median drug retention time was 142 weeks (95%CI 56, 262) with no difference between nintedanib and pirfenidone (p=0.68).

Conclusion:

In this first real-world study of antifibrotic use dedicated to RA-ILD, antifibrotic initiation was associated with a modestly improved trajectory of FVCpp. AEs were frequently reported, particularly GI, and discontinuation was common. However, lung transplant and mortality rates were still high, emphasizing the need for further therapeutic strategies in patients with severe RA-ILD. These real-world data complement previous trial data that has investigated efficacy and adverse events.

1. INTRODUCTION

Interstitial lung diseases (ILD) can be observed in 10–30% of patients with rheumatoid arthritis (RA) (1, 2). Compared to ILD associated with other systemic autoimmune rheumatic diseases (SARDs), RA-ILD is characterized by a high frequency of fibrosing ILD subtypes, with usual interstitial pneumonia (UIP) being the most frequent, followed by other subtypes such as fibrotic non-specific interstitial pneumonia (NSIP) (3). RA-ILD is a severe manifestation and ILD progression has been reported in about half of the patients (4, 5). Once clinically significant, it is associated with an increased mortality, estimated to be 2- to 10-fold higher than in RA without ILD (6–10).

Antifibrotic therapies include nintedanib and pirfenidone, which were initially approved by the US Food and Drug Agency (FDA) to slow the progression of idiopathic pulmonary fibrosis (IPF) (11, 12). Recently, each drug has been tested in several randomized controlled trials (RCT) for other causes of ILD, including SARDs such as RA-ILD (13–15). In the INBUILD trial, nintedanib was associated with a significantly lower annual rate of decline of the forced vital capacity (FVC) among patients with progressive fibrosing ILD (PF-ILD), regardless of the underlying etiology, leading to US FDA approval in this population (13). Sub-analyses of the 89 patients with RA-ILD in the INBUILD trial showed similar efficacy compared to other causes of PF-ILD (16). Pirfenidone has been tested in RA-ILD in TRAIL1, a placebo-controlled trial dedicated to RA-ILD (14). Although the primary endpoint (a composite endpoint including a decline in percent predicted FVC [FVCpp] of 10% or more or death at one year) was not met, the use of pirfenidone was associated with a clinically significant slower rate of FVCpp decline in patients with RA-ILD compared with placebo. Post-hoc analyses found that the effect of pirfenidone on the decline in FVC was more pronounced in patients with an UIP pattern (14).

Despite their positive impact on slowing ILD progression, nintedanib and pirfenidone are both associated with adverse events (AEs) that can limit their use. In the above-mentioned clinical trials, gastrointestinal (GI) AEs were frequently reported, including diarrhea (67% for nintedanib), nausea (29% for nintedanib and 53% for pirfenidone), decreased appetite and weight loss, and rash for pirfenidone (13, 14). Elevated liver enzymes were also observed in 11% of the patients treated with nintedanib (13). In the clinical trials, these AEs led to drug discontinuation (20% for nintedanib and 24% for pirfenidone) or drug dose reduction (33% for nintedanib) (13, 14). While information provided by clinical trials is important, real-world studies are also of interest since they are more representative of the population seen in clinical practice and potentially have longer follow-up. Outside of RCTs, literature about the real-world effectiveness and safety of nintedanib and pirfenidone has mostly been described in IPF patients (17–21). Studies investigating other causes of ILD have included a limited number of patients with RA-ILD and mostly treated with nintedanib (22–25). Therefore, our objective was to investigate the effectiveness and tolerability of antifibrotics in a real-world cohort of patients with RA-ILD.

2. METHODS

2.1. Study design and population

In this retrospective cohort study, we identified patients with RA-ILD initiating antifibrotics at Mass General Brigham (MGB) Integrated Health Care System in, Boston, MA, USA. MGB is a multicenter health-care system that includes 14 hospitals and primary care or specialty outpatient centers. We identified patients with RA initiating antifibrotics using electronic query. This study was approved by the Mass General Brigham Institutional Review Board. Patient informed consent was not required for this retrospective study.

2.2. Identification of patients with RA-ILD initiating antifibrotics

First, we used electronic query to identify all adults with at least two RA diagnosis codes and at least one prescription of either nintedanib or pirfenidone from January 1, 2014, to February 28, 2023. Next, we performed medical record review to only include patients having i) confirmed RA meeting 2010 American College of Rheumatology/European Alliance of Associations for Rheumatology classification criteria for RA (26), and ii) definite ILD according to Bongartz classification criteria for RA-ILD (1). We excluded RA-ILD patients that were prescribed antifibrotics without initiating them according to medical records.

2.3. Effectiveness measures

We collected FVC in mL and FVCpp as well as diffusion capacity of the lung for carbon monoxide (DLCO) in percent predicted (DLCOpp) from all available pulmonary function tests (PFT) ordered through routine clinical care. We also collected presence and dates of lung transplant and deaths. We supplemented ascertainment of death that may have occurred outside MGB by reviewing obituaries. We also collected cause of death from medical record review when available, particularly whether ILD contributed to death.

2.4. Tolerability measures

We performed medical record review to identify AEs and categorized into organ system and type of event. We also recorded the date of discontinuation as well as reason for discontinuation or last date verified to be on drug. If a patient initiated a second antifibrotic, we also reviewed this drug for tolerability.

2.5. Covariates

The index date was date of initiation of the initial antifibrotic, confirmed by medical record review. Some covariates were identified using electronic query (age, sex, self-reported race/ethnicity, rheumatoid factor [RF] and anti-citrullinated peptides antibodies [ACPA] serostatus, and disease-modifying antirheumatic drugs [DMARDs] at antifibrotic initiation). We collected other variables by medical record review. These included RA and ILD diagnosis dates, ILD pattern on clinically-indicated high-resolution computed tomography (HRCT; definite UIP, probable UIP, fibrotic NSIP), supplemental oxygen use/dose/setting (continuous, activity, nocturnal), RA disease activity category at index date (remission, low, moderate, high), glucocorticoid use and dose, and smoking status and pack-years. Besides PFT, auto-antibodies data, RA disease activity at index date, and steroids dose there were no missing data.

2.6. Statistical analysis

Characteristics at initial antifibrotic initiation (index date) were described using mean (standard deviation, SD) or median (interquartile range, IQR) for continuous variables and number (percentage) for categorical variables.

For the effectiveness analysis, patients having at least one value for FVCpp 18-months before and 18-months after antifibrotic initiation were included in a linear mixed-effect model with random intercept. Similar to prior research, a knot at date of antifibrotic initiation (time 0) was used to estimate and compare FVCpp trajectory before and after antifibrotic initiation. Since patients served as their own control subjects, analyses were performed without covariate adjustment (27). Subgroup analyses stratified by the initial antifibrotic were also performed. Similar analyses were conducted for FVC in mL and DLCOpp.

Lung transplant-free survival was assessed using Kaplan-Meier curves where patients were censored at the time of their last visit, death, or lung transplant (whichever came first). Analyses were performed in the overall population and stratified by the initial antifibrotic. We also performed Cox proportional hazard models used to identify potential baseline factors associated with the composite outcome of lung transplant or death. Calendar date of the announcement of results of the INBUILD study (September 30, 2019) was also included in the analysis to assess for possible temporal trends (28). Initial results were unadjusted. The multivariable model adjusted for age, sex, initial antifibrotic, and smoking pack-years by tertile.

For the tolerability analysis, Kaplan-Meier curves were constructed to assess drug retention from antifibrotic initiation. Patients were censored at the time of their last visit confirmed to be antifibrotic, death, lung transplant, or initial antifibrotic cessation (whichever came first). A Cox proportional hazard model with log-rank were used to compare drug retention according to initial antifibrotic (nintedanib vs. pirfenidone). We also quantified the reasons for discontinuations. Among those who initiated a second antifibrotic, we performed a similar descriptive analysis summarizing drug retention and reasons for discontinuation.

We considered a two-sided p value of less than 0.05 as statistically significant in all analyses. All analyses were performed using SAS (version 9.4).

3. Results

3.1. Study sample and characteristics

We identified 114 adults with at least two ICD codes for RA and at least one prescription for an antifibrotic between January 1, 2014 to February 28, 2023. We excluded 31 patients that did not meet ACR/EULAR criteria for RA and 9 patients that never initiated an antifibrotic (Supplementary Figure 1). The final sample included 74 patients who initiated an antifibrotic for RA-ILD.

The characteristics at initial antifibrotic initiation of the 74 included patients are available in Table 1. Nintedanib was initiated first for 40 (54%) patients and 34 (46%) initiated pirfenidone. Mean age was 67.8 (SD 8.6) years, 39 (53%) were male, 61 (82%) were white, 51 (69%) were ever smokers, and 52 (70%) had UIP pattern on HRCT. At initial antifibrotic initiation, mean RA duration was 8.8 (SD 10.9) years, mean ILD duration was 5.3 (SD 5.0) years, mean FVCpp was 73.1% (SD 20.1), and mean DLCOpp was 42.4% (SD 17.7). While the number of patients initiating pirfenidone remained relatively stable over time, the number of patients initiating nintedanib increased after 2019 (Supplementary Figure 2).

Table 1.

Characteristics of patients with RA-ILD at initiation of initial antifibrotic medication.

	Overall n=74	Nintedanib n=40 (54%)	Pirfenidone n=34 (46%)
Demographics and lifestyle
Mean age, years (SD)	67.8 (8.7)	69.2 (9.1)	66.3 (7.9)
Male sex	39 (53%)	19 (48%)	20 (59%)
Race
- White	61 (82%)	34 (85%)	27 (79%)
- Black	7 (9%)	2 (5%)	5 (15%)
- Asian	2 (3%)	1 (3%)	1 (3%)
- Other	4 (5%)	3 (8%)	1 (3%)
Smoking status
- Never	23 (31%)	11 (28 %)	12 (35%)
- Past	48 (65%)	28 (7 0%)	20 (59%)
- Current	3 (4%)	1 (3%)	2 (6%)
Median pack-years (IQR)	14.5 (0, 30)	15 (0, 30)	9.5 (0, 35)
Smoking level, pack-years
- Lowest tertile (0 – 1.9)	24 (32%)	11 (28%)	13 (38%)
- Middle tertile (2 – 27.9)	25 (34 %)	15 (38%)	10 (29%)
- Highest tertile (28 – 100)	25 (34%)	14 (35%)	11 (32%)
RA characteristics
Mean RA duration, years (SD)	8.8 (10.9)	11.1 (12.1)	6.1 (8.7)
Seropositive, missing data = 1	53 (74%)	32 (80%)	21 (66%)
ACPA-positive, missing data = 3	40 (59%)	22 (59%)	18 (58%)
RF-positive, missing data = 2	47 (67%)	29 (76%)	18 (56%)
RA disease activity, mis. ing data = 4
- Remission	44 (68%)	24 (71%)	20 (65%)
- Low	12 (18%)	7 (21%)	5 (16%)
- Moderate	8 (12%)	3 (9%)	5 (16%)
- High	1 (2%)	0 (0%)	1 (3%)
Concomitant dMARDs
Any DMARD	33 (45%)	19 (48%)	14 (41%)
Any DMARD or glucocorticoid	50 (66%)	30 (75%)	20 (59%)
Methotrexate	5 (7%)	2 (5%)	3 (9%)
Other csDMARDs	17 (23%)	9 (23%)	8 (24%)
Rituximab	10 (14%)	8 (20%)	2 (6%)
Other biologic DMARDs	6 (8%)	2 (5%)	4 (12%)
JAK inhibitors	1 (1%)	0 (0%)	1 (3%)
Mycophenolate mofetil	4 (5%)	1 (3%)	3 (9%)
Glucocorticoids	37 (50%)	24 (60%)	13 (38%)
Median glucocorticoid dose, mg daily prednisone (IQR), missing data = 20	10 (5, 10)	10 (5, 10)	20 (7.5, 20)
ILD characteristics
Mean ILD duration, years (SD)	5.3 (5.0)	5.4 (5.3)	5.3 (4.7)
HRCT ILD pattern
- Definite UIP (%)	48 (65%)	28 (70%)	20 (59%)
- Probable UIP (%)	4 (5%)	3 (8%)	1 (3%)
- Fibrotic NSIP (%)	13 (18%)	6 (15%)	7 (21%)
- Unknown (%)	9 (12%)	3 (8%)	6 (18%)
Mean FVCpp (SD), missing data = 2	73.1 (20.1)	73.4 (23.0)	72.8 (16.4)
Mean FVC, mL (SD), missing data = 3	2,630 (940)	2,560 (1,040)	2,720 (790)
Mean DLCOpp (SD), missing data = 7	42.4 (17.7)	41.6 (17.5)	43.5 (18.3)
O2 current use	22 (30%)	12 (30%)	10 (29%)
- Continuous	15/22 (68%)	9/12 (75%)	6/10 (60%)
- Only with activity	5/22 (23%)	2/12 (17%)	3/10 (30%)
- Only nocturnal	2/22 (9%)	1/12 (8%)	1/10 (10%)
If continuous, median O2 flow (L/min) (IQR)	2 (2, 4)	2 (2, 4)	3 (2, 5)

ACPA: anti-citrullinated peptides auto-antibodies, csDMARDs: conventional disease modifying drugs, DLCO: diffusing lung capacity of the lungs for carbon monoxide, FVC: forced vital capacity, HRCT: high resolution computed tomography, ILD: Interstitial lung disease, NSIP: non-specific interstitial pneumonia, RA: rheumatoid arthritis, RF: rheumatoid factor, UIP: usual interstitial pneumonia.

3.2. Antifibrotic effectiveness

A total of 49 patients with available PFTs 18-months prior to 18-months after first antifibrotic initiation were included in this analysis. Characteristics of the 49 patients are available in Supplementary Table 1. The median number of PFTs performed per patient was 8 (IQR 5, 10). We identified a significantly slower decline in FVCpp trajectory after initial antifibrotic initiation: −6.2% per year before compared to −0.3% per year after antifibrotic initiation, p=0.03 (Figure 1). A similar change in trajectory was observed for both nintedanib (n=27) and pirfenidone (n=22), although this did not reach statistical significance in the subgroup analyses: −6.1% per year prior compared to +0.5% per year after nintedanib initiation (p=0.11) and −6.5% per year prior compared to −1.5% per year after pirfenidone initiation (p=0.15).

Figure 1. Trajectories of FVCpp, absolute FVC and DLCOpp before and after initial antifibrotic prescription for RA-ILD.

The shaded grey represents the 95%CI of the FVCpp (Panels A, B and C), the absolute FVC (Panels C, D and E) and the DLCOpp (Panels G, H and I) trajectory. The dotted line represents the hypothetical decline based on values prior to antifibrotic initiation. Panel A, D and G represent patients that initiated either antifibrotic (nintedanib or pirfenidone), Panel B, E and H represent patients that initiated nintedanib, and Panel C, F and I represent patients that initiated pirfenidone. Trajectories were estimated using a linear mixed model with random intercept with a knot at initial antifibrotic prescription (time 0). DLCOpp: diffusion capacity of the lung for carbon monoxide percent predicted, FVCpp: forced vital capacity percent predicted.

Similar change of trajectories was observed for absolute FVC without statistical significance (−208.2 mL per year prior compared to −43.7 mL per year after initial antifibrotic initiation, p=0.29). The impact of antifibrotics on DLCOpp trajectory was also not statistically significant (−5.8% per year prior compared to −3.5% after initial antifibrotic initiation, p=0.47). Lastly, we observed similar results for the patients with UIP pattern (n=35), Supplementary Figure 3. We were unable to perform the analysis for non-UIP RA-ILD (n=14) due to limited sample size.

3.3. Lung transplant-free survival

After a median follow-up of 90 weeks (min 4, max 387; IQR 47, 172), 4 (5%) patients had undergone a lung transplant and 26 patients (35%) had died. ILD contributed to the cause of death for 17/26 (65%) patients who died. Median lung transplant-free survival was 265.1 weeks (95%CI 129.7, 305.8; Figure 2). No difference was observed when the analysis was stratified by type of initial antifibrotic: median survival of 267.5 weeks (95%CI 118.6, NA) for nintedanib and 187.5 weeks (95%CI 87.7, NA) for pirfenidone (p=0.87).

Figure 2. Kaplan-Meier curves for lung transplant-free survival.

The survival estimates are represented by the Kaplan-Meier curve with the corresponding 95%CI in shading. Panel A represents patients that initiated either antifibrotic (nintedanib or pirfenidone), Panel B represents the analysis stratified by the initial antifibrotic. CI: confidence interval.

Unadjusted Cox regression identified that male sex was associated with risk of lung transplant or death (HR 2.28, 95%CI 1.07, 4.89, p=0.049; Table 2). In the multivariable regression including age, sex, type of initial antifibrotic, and smoking pack-year level, an independent association was found between mortality and smoking pack-year level (highest vs. lowest/middle tertile, HR = 2.59 95% IC 1.21, 5.54, p=0.015). No association was observed comparing nintedanib to pirfenidone for risk of lung transplant or death (HR 1.03, 95%CI 0.46, 2.29, p=0.95).

Table 2.

Hazard ratios for the composite of lung transplant or mortality by baseline factors at initiation of antifibrotic for RA-ILD (n=74)

	Unadjusted HR (95%CI)	Multivariable HR (95%CI)
Demographics and lifestyle
Age (per 10 years)	1.16 (0.75, 1.79)	0.96 (0.58, 1.58)
Female sex	1.0 (Ref)	1.0 (Ref)
Male sex	2.28 (1.07, 4.89)	2.25 (1.002, 5.06)
Smoking status
- Never	1.0 (Ref)	--
- Ever	1.74 (0.80, 3.78)	--
Smoking pack-years (per unit)	1.012 (1.000, 1.025)	--
Smoking level
- Lowest/middle tertile	1.0 (Ref)	1.0 (Ref)
- Highest tertile)	2.63 (1.26, 5.49)	2.59 (1.21, 5.54)
RA characteristics
RA duration (per year)	1.01(0.98, 1.04)	--
Seronegative	1.0 (Ref)	--
Seropositive	0.78 (0.35, 1.72)	--
RA disease activity
- Remission/low	1.0 (Ref)	--
- Moderate/high	0.44 (0.10, 1.90)	--
Concomitant DMARDs
No DMARDs	1.0 (Ref)	--
csDMARDs only	0.29 (0.07, 1.23)	--
Rituximab or mycophenolate mofetil	0.63 (0.21, 1.85)	--
Other biologic DMARDs and JAK inhibitors	1.38 (0.41, 4.71)	--
No glucocorticoids	1.0 (Ref)	--
Any glucocorticoids	1.47 (0.70, 3.07)	--
Glucocorticoid category
- No glucocorticoid	1.0 (Ref)	--
- >0 to 5 mg	1.35 (0.47, 3.87)	--
- >5 to 15 me	1.38 (0.54, 3.50)	--
- >15 mg	1.76 (0.62, 4.95)	--
ILD characteristics
Nintedanib	0.94 (0.45, 1.96)	1.03 (0.46, 2.29)
Pirfenidone	1.0 (Ref)	1.0 (Ref)
ILD duration (per year)	1.02 (0.96, 1.10)	--
ILD pattern		--
- Definite or probable UIP	1.46 (0.64, 3.30)	--
- Fibrotic NSIP or unknown	1.0 (Ref)	--
FVCpp (per unit)	0.99 (0.97, 1.01)	--
DLCOpp (per unit)	0.97 (0.94, 0.998)	--
No O2 use	1.0 (Ref)	--
O2 use	1.25 (0.58, 2.68)	--
Calendar date
Before September 30, 2019	1.0 (Ref)	--
September 30, 2019 or later	0.94 (0.42, 2.13)	--

ACPA: anti-citrullinated peptides proteins, csDMARDs: conventional disease modifying drugs, DLCOpp: diffusing lung capacity of the lungs for carbon monoxide percent predicted, FVCpp: forced vital capacity percent predicted, HRCT: high resolution computed tomography, ILD: Interstitial lung disease, NSIP: non-specific interstitial pneumonia, RA: rheumatoid arthritis, RF: rheumatoid factor, UIP: usual interstitial pneumonia.

3.4. Antifibrotic tolerability and retention

At the end of follow-up, AEs were reported in 41 patients (55%). The most frequent AEs was GI events (n=30), followed by ILD progression (n=6), skin rash (n=3), and hepatitis (n=2). No severe AEs or death related to antifibrotic use were reported. The initial antifibrotic was discontinued in 34 patients (46%), with a median time to discontinuation of 27 weeks (IQR 12, 74). Reasons for discontinuation were GI events (n=19), rash (n=3), hepatitis (n=2), financial (n=1), and patient decision, not due to AEs (n=3). No difference in reasons for discontinuation were observed between nintedanib and pirfenidone (Supplementary Table 2). Overall antifibrotic median retention was 142 weeks (95%CI 56, 262; Figure 3). No statistically significant difference was observed depending on the initial antifibrotic: median antifibrotic retention was 148 weeks (95%CI 37, NA) for nintedanib and 106 weeks (95%CI 56, 262) for pirfenidone (p=0.68).

Figure 3. Kaplan-Meier curves for retention of initial antifibrotic used to treat RA-ILD.

The survival estimates are represented by the Kaplan-Meier curve with the corresponding 95% CI in shaded. Panel A represents patients that initiated either antifibrotic (nintedanib or pirfenidone), Panel B represents the analysis stratified by the initial antifibrotic. CI: confidence interval.

A second antifibrotic was prescribed in 14 patients. After a median follow-up time of 69 weeks (min 8, max 208; IQR 49, 162), 4/14 patients discontinued the second antifibrotic (3 for GI events, 1 for ILD progression).

4. DISCUSSION

Our study is the first real-world study dedicated to investigating tolerability and effectiveness of antifibrotic medications in RA-ILD. We found that the initiation of an antifibrotic was associated with a modestly slower decline in FVCpp. After a median follow-up of 90 weeks, AEs were frequent for both antifibrotics and resulted in drug discontinuation in more than half of the patients. Despite a potential positive effect on lung function decline, RA-ILD was still associated with high mortality rate, especially in men and heavy smokers, emphasizing that additional effective and tolerable treatments are needed.

To date, effectiveness and tolerability of antifibrotics have mostly been investigated in IPF (17–21, 29). However, the applicability of IPF data to RA-ILD is unknown due to difference in patient characteristics, comorbidities, and concomitant use of glucocorticoids and DMARDs in RA that could increase the risk of AEs. The currently available literature for antifibrotics use outside IPF is mostly restricted to the INBUILD and TRAIL1 trials, both having included a relatively limited RA-ILD sample size (n=89 and n=123, respectively) (13, 14). RCTs have the strongest design to investigate efficacy. However, they may underestimate AEs and drug retention and overestimate effectiveness compared to real-world studies related to strict inclusion/exclusion criteria and willingness/ability to participate in a trial. Studies dedicated to real-world use of antifibrotics outside IPF mostly focused on nintedanib with a very small number of patients with RA-ILD (ranging from 7 to 17 patients) (22–25).

The efficacy of antifibrotics in RA-ILD has been evaluated in the INBUILD study that was not dedicated to RA-ILD and the TRAIL1 study which primary outcome was not met, perhaps underpowered due to not meeting the enrollment goal (13, 14). Although sub-analyses of the INBUILD study and secondary outcomes of the TRAIL1 study suggested a significant impact of antifibrotics on FVC decline in RA-ILD, additional data confirming these results may be anticipated (14, 15, 30). In our study, both nintedanib and pirfenidone use was associated with less decline in FVC and DLCO trajectory over 18-months. This change of trajectory was significant for FVCpp but not for absolute FVC and DLCO with a potentially limited power of detection due to small sample size. We estimated a difference of FVC of 164.5 mL per year after antifibrotic initiation which was similar to that observed in the INBUILD trial (107 mL per year, 95% CI [65.4, 148.5]) and the TRAIL1 trial (80 mL per year [SD 30] for the overall population, and 126 mL per year [SD 39] for the UIP subgroup analysis) (13, 14). Similar results were observed between nintedanib and pirfenidone, suggesting comparable efficacy between the two drugs, though we were underpowered to detect modest differences.

In our study, more than half of the patients reported AEs, with the most frequent being GIs events (41% of the patients), followed by ILD progression, rash, and hepatitis. These results are similar to reports from clinical trials and the previous real-world studies in IPF. Nausea and diarrhea are the most frequent AEs reported in IPF and non-IPF clinical trials for both pirfenidone and nintedanib (24.3% and 61.5% for nintedanib in the pooled INPULSIS trials and 35.5% and 24.6% for pirfenidone in the pooled CAPACITY trials, respectively) (11, 12). Real-world studies confirmed the high frequency of GI events, for example in the PASSPORT study (1009 patients with IPF treated with pirfenidone across Europe) that reported GI AEs in 38% of the patients (18). Even if these AEs are rarely severe, they may lead to persistent dose reduction or drug discontinuation. In IPF, real-world studies observed a higher drug discontinuation rate compared to the clinical trials. In the CAPACITY trial, 11.9% of the patients stopped pirfenidone compared to 28.7% in the PASSPORT study (11, 18). For nintedanib, drug discontinuation was reported in 17.6% in INPULSIS compared to up to 26.3% in real-world studies (12, 31, 32). Likewise, our study revealed a high drug discontinuation rate (46%), about twice higher than the one reported by the INBUILD and the TRAIL1 trials (20% and 24%, respectively) (13, 14). As reported in IPF, GI events were the most frequent cause of discontinuation (57%) (31). Interestingly, even if 45% of the patients were using a concomitant DMARDs, we did not observe an increased frequency of hepatitis (3% in our study compared to 2.9% to 13% in the literature).

Despite the recent availability of two antifibrotics with impact on lung function decline, RA-ILD remains a severe disease associated with high mortality. The impact of antifibrotics on quality of life remains uncertain. In our study, about half of the patients had died or had undergone a lung transplant after a median follow-up of 90 weeks. While the patients included in our study may reflect the most severe cases of progressive fibrosing RA-ILD, they also highlight the important burden of the disease. We identified male sex and heavy smoking as independent risk factors for mortality. There were no association between choice of antifibrotic or DMARDs and mortality, though our study may have been underpowered to detect modest differences. Interestingly, while often associated with ILD progression and mortality in RA-ILD in previous literature (10, 33), we did not identify the UIP HRCT pattern as a risk factor for mortality. This finding supports those from a previous study identifying fibrotic features and ILD HRCT extent as independent risk factors for mortality, regardless of the HRCT pattern (34). Altogether, these results suggest that patients with non-UIP RA-ILD having a progressive fibrotic ILD phenotype should be considered at similar risk for mortality that the one with a UIP pattern.

Our study has several strengths. This is the first real-world study investigating antifibrotic use in RA-ILD in a contemporary cohort. We were able to systematically identify all RA-ILD cases in this large healthcare system. In addition, all cases met research criteria for both RA and ILD.

Our study also has some limitations to consider. First, the interpretation of the change in lung function after antifibrotics prescription is limited due to the lack of control group (patients with RA-ILD without antifibrotic use). While we considered including a control group, the RA-ILD patients whose clinicians chose to initiate antifibrotic treatment are likely the most severe and progressive patients from an ILD perspective. Untreated patients likely had less severe and less progressive RA-ILD which would likely have confounded results due to the indication of antifibrotics. However, most patients had low RA disease activity or remission at antifibrotic initiation, which may limit generalizability and also suggests that patients with active articular RA were either not treated or initiated anti-inflammatory medications for ILD. Secondary analyses compared each antifibrotic drug but were likely underpowered to detect modest differences due to small sample size. Second, although this is the first real-world study of antifibrotic use in RA-ILD patients, our sample size may still have been too small to detect significant differences. the small sample size may have decreased the power to detect significant differences in drug retention, change of lung function, and survival. Indeed, we could only identify a significant change in trajectory for FVCpp. A similar change in absolute FVC was observed without reaching statistical significance. Besides, the limited number of included patients did not allow several additional analyses such as effectiveness analyses restricted to patients having a non-UIP RA-ILD, and identification of risk factors for drug discontinuation. Third, this was a retrospective study from a single healthcare system with non-standardized data collection from clinical care. Some factors such as RA disease activity over time could not be assessed. Genetic analyses such as MUC5B rs35705950 genotyping could not be performed although a previous study did not find any association with progression nor mortality (35). Likewise, we could not identify patients meeting the INBUILD definition for progressive ILD, though it should be noticed that 32 (43%) of them initiated antifibrotic before the release of the INBUILD publication (13). However, the MGB includes 14 distinct hospitals and centers, we were able to obtain detailed data from medical record review that would not be available in administrative data and minimized missing data. In any case, additional studies with more diverse patient populations in different geographic regions are needed. Last, our retrospective design is still considered as less robust than an RCT. However, results from both methods are of interest for clinical practice. They should be interpreted considering the strengths and the limitations of each design.

In conclusion, in this first real-world study of antifibrotic use dedicated to RA-ILD, antifibrotic initiation was associated with a modestly improved trajectory of FVC. However, AEs were frequently reported, particularly GI, and discontinuation was common. Also, the mortality rate was still high and the impact of antifibrotics on quality of life is unclear, emphasizing the need for further therapeutic strategies in patients with severe RA-ILD. Further prospective studies with larger sample size are needed to better estimate the benefit and risks of antifibrotics in RA-ILD.