Prevalence and mortality associations of interstitial lung abnormalities in rheumatoid arthritis within a multicentre prospective cohort of smokers

Abstract

Objective

To investigate the prevalence and mortality impact of interstitial lung abnormalities (ILAs) in RA and non-RA comparators.

Methods

We analysed associations between ILAs, RA, and mortality in COPDGene, a multicentre prospective cohort study of current and past smokers, excluding known interstitial lung disease (ILD) or bronchiectasis. All participants had research chest high-resolution CT (HRCT) reviewed by a sequential reading method to classify ILA as present, indeterminate or absent. RA cases were identified by self-report RA and DMARD use; non-RA comparators had neither an RA diagnosis nor used DMARDs. We examined the association and mortality risk of RA and ILA using multivariable logistic regression and Cox regression.

Results

We identified 83 RA cases and 8725 non-RA comparators with HRCT performed for research purposes. ILA prevalence was 16.9% in RA cases and 5.0% in non-RA comparators. After adjusting for potential confounders, including genetics, current/past smoking and other lifestyle factors, ILAs were more common among those with RA compared with non-RA [odds ratio 4.76 (95% CI 2.54, 8.92)]. RA with ILAs or indeterminate for ILAs was associated with higher all-cause mortality compared with non-RA without ILAs [hazard ratio (HR) 3.16 (95% CI 2.11, 4.74)] and RA cases without ILA [HR 3.02 (95% CI 1.36, 6.75)].

Conclusions

In this cohort of smokers, RA was associated with ILAs and this persisted after adjustment for current/past smoking and genetic/lifestyle risk factors. RA with ILAs in smokers had a 3-fold increased all-cause mortality, emphasizing the importance of further screening and treatment strategies for preclinical ILD in RA.

Rheumatology key messages.

• In this cohort of smokers, 17% of RA patients had preclinical interstitial lung abnormalities (ILAs). RA had 4-fold higher odds of ILAs than non-RA comparators, adjusted for current/past smoking, the MUC5B promoter variant and other factors.

• Participants with RA and no ILAs were not at increased mortality risk while those with ILAs or indeterminate for ILAs had a 3-fold increased risk of all-cause mortality compared with RA and non-RA patients without ILAs.

• Further studies are needed to establish the clinical utility of screening, prevention and treatment strategies targeting preclinical lung disease in current/past smokers with RA.

Introduction

RA-associated interstitial lung disease (RA-ILD) is a severe extra-articular disease manifestation and a major driver of morbidity and mortality among RA patients [1–7]. Although 5–10% of RA patients may develop clinically apparent RA-ILD, research screening studies have demonstrated that up to one-third of RA patients may have preclinical ILD based on evaluations of high-resolution CT (HRCT) scans [1, 8–12]. RA shares several risk factors with idiopathic pulmonary fibrosis, the stereotypical fibrotic ILD, including cigarette smoking, ageing and male sex [8, 13]. Thus it remains unknown whether the prevalence of lung abnormalities in RA simply reflects these overlapping risk factors or is explained by RA-related pulmonary inflammation, RA treatment effects or other mechanisms. Additionally, the prognostic implications of preclinical pulmonary abnormalities in RA remain unclear, as only a minority of RA patients with pulmonary abnormalities will develop clinically apparent lung disease [14].

Although there has been limited previous research on the implications of preclinical lung disease in RA, several large cohort studies have investigated preclinical pulmonary abnormalities in the general population [15–17]. These studies have largely focused on interstitial lung abnormalities (ILAs), a term that refers to specific radiologic patterns of increased lung density noted on HRCT chest imaging [18, 19]. ILAs share several characteristics and risk factors with idiopathic pulmonary fibrosis and other forms of ILD and may represent a preclinical form of these conditions in some patients. ILAs can occur in up to 7% of the general population [17] and have been associated with increased mortality in multiple cohorts [15, 16]. There have also been some investigations of ILAs in the RA population. One prior study of RA patients without known ILD noted that 15% had ILAs on screening HRCT [20]. Other studies investigating RA patients with clinically performed CT scans have noted a prevalence of preclinical ILD as high as 44–45% [14, 21]. Other studies, that did not specifically use an ILA definition, found HRCT evidence of preclinical ILD in 16.9–33% of RA patients [9, 10, 22, 23]. However, these studies did not specifically include non-RA controls and address known ILA risk factors, including age, smoking status and the promoter variant of the MUC5B gene.

We used data from a large prospective cohort of current and past smokers to determine the prevalence of ILAs and investigate the long-term mortality risk of ILAs among patients with and without RA. Given the known importance of cigarette smoking as a risk factor for ILAs and ILD, this cohort was enriched for lung disease and represents a population of interest for potential screening strategies. We hypothesized that RA would be associated with ILAs, even after adjustment for current/past smoking and other risk factors. We further hypothesized that the combination of RA and ILAs would confer significantly higher mortality risk than either condition alone.

Methods

Study population and design

We analysed COPDGene, a multicentre prospective cohort of current and past smokers that has been described in detail elsewhere [24–26]. Briefly, the study enrolled non-Hispanic White or Black smokers with at least 10 pack-years of smoking history, 45–80 years of age, at 21 clinical centres in the USA between 2007 and 2011. The study enrolled smokers with and without baseline obstructive lung disease and reached its goal to include one-third Black participants. Baseline screening included health questionnaires, HRCT chest imaging, spirometry and genotyping. Participants with known respiratory diseases besides asthma were ineligible. Participants with evidence of ILD or diffuse bronchiectasis on initial CT scan were also excluded from the COPDGene cohort. We performed a cross-sectional study investigating the association of RA compared with non-RA with ILAs. We then performed a cohort study examining the associations of RA and ILAs with mortality risk. This substudy was approved by the Mass General Brigham Institutional Review Board (protocol 2020P000558) and adheres to the Declaration of Helsinki. All patients provided written informed consent at COPDGene enrolment.

RA cases and non-RA comparators

Within COPDGene we identified participants with RA based on a combination of self-reported physician-diagnosed RA and the use of at least one DMARD at baseline. A previous study noted low validity of self-report RA status alone, but reported a positive predictive value (PPV) of 88% using a similar case definition that combined self-report status with DMARD use [27]. We included medications approved for RA by the US Food and Drug Administration at the time COPDGene baseline visits were conducted and other DMARDs previously validated for identification of RA patients in cohort studies (Supplementary Table S1, available at Rheumatology online) [28]. We defined non-RA comparators as participants who did not report a history of RA and were not using any DMARDs at baseline. Participants who reported an RA history but did not report DMARD use and those on one or more DMARDs without reporting a history of RA were excluded as comparators. We performed an additional sensitivity analysis that included all participants who self-reported RA status as cases, regardless of DMARD use.

ILAs

The presence of ILAs on the research chest HRCT was identified using a previously described sequential reader review of HRCT images by up to three expert readers (two chest radiologists and one pulmonologist) to classify ILAs as present, indeterminate or absent [24, 29]. ILAs were defined as non-dependent subpleural changes affecting >5% of any lobar area. These included non-dependent ground-glass and reticular opacities, diffuse centrilobular nodules, non-emphysematous cysts, honeycombing and traction bronchiectasis. ILAs were further subcategorized into fibrotic and non-fibrotic subtypes, consistent with recent guidelines that recognized a higher rate of progression and poorer prognosis among those with fibrotic ILAs [18]. Scans with focal or unilateral changes, or patchy ground-glass opacities affecting <5% of a lobar region were considered indeterminate for ILAs.

Based on recent Fleischner Society guidelines [18], ILAs identified during the screening of high-risk individuals, including those with RA, can be considered ‘preclinical ILD’. Prior to this report, some studies referred to ILAs in RA as ‘subclinical ILD’ [30]. In order to maintain interpretability of the comparison of lung abnormalities in RA and non-RA comparators, we have referred to these findings as ‘RA with ILAs’ and also indicated where ‘preclinical ILD’ would also be appropriate terminology.

Chronic obstructive pulmonary disease (COPD)

We used each participant’s post-bronchodilator forced expiratory volume in 1 s (FEV₁) and forced vital capacity (FVC) to categorize them into severity grades according to the Global Initiative for Chronic Lung Disease (GOLD) 2007 classification system.

Mortality

Death outcomes in COPDGene were assessed through longitudinal study follow-up by biannual phone calls and periodic searches of the social security death index. The mortality dataset was last updated in August 2022.

Covariates

We obtained age at the COPDGene baseline visit, sex and self-reported race. BMI was obtained from the baseline examination. Smoking status (current, past) and pack-years of smoking were obtained from a baseline respiratory health questionnaire. As the MUC5B promoter variant rs35705950 has been associated with RA in some cohorts [31] and is the strongest known genetic risk factor for ILAs and RA-ILD [17, 32], this genetic risk factor was included as a covariate. Patients were genotyped using the TaqMan real-time PCR assay (Thermo Fisher Scientific, Waltham, MA, USA). In analyses that included this genetic marker, we used genotyping data to determine principal components of ancestry and included the first six principal components of ancestry as additional covariates [33]. This technique is widely used to minimize the possibility of confounding by genetic ancestry but does not eliminate other potential sources of confounding in the study design or related to assay or batch effects [33].

Statistical analysis

We compared baseline characteristics between RA cases and non-RA comparators using frequencies, proportions and means with s.d.s. We examined the association between RA case vs non-RA comparator status with ILAs, indeterminate for ILAs, and the fibrotic ILA subtype using unadjusted and multivariable logistic regression adjusted for age, sex, current/past smoking status, pack-years of smoking and BMI. We performed an additional model that also adjusted for genetic ancestry and the MUC5B rs35705950 genotype, treating the number of risk alleles for each patient as an ordinal variable. Participants with a missing MUC5B promoter variant genotype (n = 5 RA cases and n = 259 non-RA comparators) were excluded from that analysis. We assessed for a possible association between RA and COPD, as defined by GOLD class 2–4, using multivariable logistic regression adjusted for age, sex, current/past smoking status and pack-years of smoking. There were 51 non-RA comparators who were missing spirometry data and excluded from these analyses. There were no other missing data.

For the mortality analyses, we constructed Kaplan–Meier curves with cumulative incidence of mortality based on RA and ILA status. We also performed Cox regression models to estimate hazard ratios (HRs) for mortality based on RA vs non-RA status as well as a cross-classified variable based on ILA presence/absence and RA/non-RA status. We performed unadjusted models as well as multivariable models adjusted for age, sex, current/past smoking status, pack-years and BMI. For our main analyses, the reference group was non-RA comparators without ILA to demonstrate the impact of the combination of RA and ILAs on mortality compared with a comparable general population of smokers without lung abnormalities. We also performed our analyses limited to RA cases using RA without ILAs as the reference group. In secondary analyses, we repeated the Cox proportional hazards models after excluding indeterminate for ILAs and another analysis after excluding ILAs.

We assessed the proportional hazards assumptions in all analyses using Schoenfeld residuals. There were no significant violations of the proportional hazards assumptions except for the age and BMI covariates, which showed interactions with follow-up time and the mortality outcome (Supplementary Table S1 and Supplementary Figs S1–S8, available at Rheumatology online). In models that included interaction terms of these variables with follow-up time, the point estimates of the main exposures were unchanged so we reported results without the interaction terms. Two-sided P-values <0.05 were considered statistically significant. All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA).

Results

Study sample

Of 10 371 COPDGene participants, we identified 83 RA cases and 8725 non-RA comparators scored for ILAs (Fig. 1). RA cases self-reported a history of RA and were using at least one DMARD at baseline. Details of specific DMARD use are detailed in Supplementary Table S2, available at Rheumatology online. A total of 656 participants who reported a history of RA but were not using DMARDs and 45 participants who reported DMARD use but no RA history were excluded. We also excluded 2 participants with RA and 611 non-RA comparators who were not scored for ILAs.

Figure 1.

Identification of participants with RA and non-RA comparators in COPDGene

Baseline characteristics

Baseline characteristics of RA and non-RA participants are shown in Table 1. Participants with RA were older at enrolment (63.8 vs 59.3 years) and more predominantly female (65.1 vs 45.9%) and White race (79.5 vs 67.5%). The RA cases had a higher proportion of past smokers (66.3 vs 46.8%) and a higher mean BMI (30.0 vs 28.7 kg/m²).

Table 1.

Demographics, lifestyle factors, pulmonary disease and genetics of RA patients and non-RA comparators at baseline (N = 8808).

Characteristics	RA cases (n = 83)	Non-RA comparators (n = 8725)
Demographics
Age at enrolment, years, mean (s.d.)	63.8 (8.7)	59.3 (9.0)
Female, n (%)	54 (65.1%)	4002 (45.9%)
White race, n (%)	66 (79.5%)	5890 (67.5%)
Black race, n (%)	17 (20.5%)	2835 (32.5%)
Lifestyle
Current smoker, n (%)	28 (33.7%)	4645 (53.2%)
Past smoker, n (%)	55 (66.3%)	4080 (46.8%)
Pack-years, mean (s.d.)	42.2 (17.9)	44.0 (24.6)
BMI, kg/m2, mean (s.d.)	30.0 (7.8)	28.7 (6.2)
MUC5B promoter variant (rs35705950) genotype, n/N (%)
GG	63/78 (80.8)	7266/8466 (85.8)
GT	15/78 (19.2)	1136/8466 (13.4)
TT	0/78 (0.0)	64/8466 (0.8)
Chronic obstructive pulmonary diseasea, n/N (%)
GOLD class 0 (no COPD)	32 (38.6)	3790/8674 (43.7)
PRISm	13 (15.7)	1063/8674 (12.3)
GOLD class 1	3 (3.6)	697/8674 (8.0)
GOLD class 2	17 (20.5)	1627/8674 (18.8)
GOLD class 3	13 (15.7)	986/8674 (11.4)
GOLD class 4	5 (6.0)	511/8674 (5.9)
Pulmonary function testing (post-bronchodilator)
FEV1, % predicted (s.d.)	70.8 (23.9)	76.7 (25.5)
FVC, % predicted (s.d.)	83.3 (17.4)	87.2 (18.3)
FEV1/FVC (s.d.)	0.641 (0.17)	0.668 (0.16)

PRISm, preserved ratio impaired spirometry (subset of GOLD class 0).

^aBased on GOLD 2007 classification using post-bronchodilator FEV₁/FVC only.

COPD associations

The proportion of RA cases and non-RA comparators in each GOLD class was similar (Table 1). There were no statistically significant associations between RA and COPD. Odds ratios (ORs) were 1.21 (95% CI 0.78, 1.88) in the unadjusted analysis and 0.93 (95% CI 0.58, 1.50) in the analysis adjusted for age, sex, current/past smoking status and pack-years (Supplementary Table S3, available at Rheumatology online).

ILA prevalence and associations

A higher proportion of RA cases had ILAs (16.9%) compared with the non-RA comparators (5.0%; Fig. 2). In the multivariable logistic regression model adjusted for age, sex, current/past smoking status, pack-years and BMI, RA cases had an OR of 4.15 (95% CI 2.17, 7.97) for ILAs compared with those without RA (Table 2). Additionally, 8.4% of the RA cases had fibrotic ILAs compared with 1.1% of the non-RA comparators. For RA cases, the OR for fibrotic ILAs was 9.26 (95% CI 3.82, 22.44). Results were similar after adjustment for the number of risk alleles for the MUC5B promoter variant (rs35705950), as shown in Table 2. There was also a higher proportion of participants who were indeterminate for ILAs among the RA cases (41.0%) compared with those without RA (35.8%).

Figure 2.

Frequency and subtypes of radiologic ILAs among RA patients and non-RA comparators at baseline (N = 8808). *Among RA cases, presence of ILAs may be considered ‘preclinical ILD’

Table 2.

Cross-sectional association of RA and radiologic ILAs* on HRCT at baseline (N = 8808).

	Unadjusted OR (95% CI)	Model 1a OR (95% CI)	Model 2b OR (95% CI)
Outcome: ILAs or indeterminate for ILAs* vs no ILAs
Non-RA comparators	1.0 (Ref)	1.0 (Ref)	1.0 (Ref)
RA cases	1.99 (1.28, 3.08)	1.87 (1.20, 2.92)	1.84 (1.16, 2.92)
Outcome: ILAs* vs no ILAs
Non-RA comparators	1.0 (Ref)	1.0 (Ref)	1.0 (Ref)
RA cases	4.76 (2.54, 8.92)	4.15 (2.17, 7.97)	3.90 (1.95, 7.80)
Outcome: Fibrotic ILAs* vs no ILAs
Non-RA comparators	1.0 (Ref)	1.0 (Ref)	1.0 (Ref)
RA cases	11.23 (4.86, 25.94)	8.41 (3.49, 20.25)	8.18 (3.17, 21.08)

^aAdjusted for age, sex, smoking status (current/past), pack-years and BMI.

^bAdditionally adjusted for MUC5B rs35705950 genotype (copy number of T risk alleles; missing excluded) and the first six principal components of ancestry.

^*Among RA cases, presence of ILAs may be considered ‘preclinical ILD’.

RA, ILAs and mortality risk

In the mortality analysis, the median follow-up time was 10.5 years [interquartile range (IQR) 5.8–12.4] for RA cases and 10.7 years (IQR 5.4–12.4) for non-RA comparators. There were 33 deaths among the RA cases and 2225 among non-RA comparators during 740.9 and 76 594.5 person-years, respectively (Supplementary Table S4, available at Rheumatology online). The 10-year mortality was 50.0% (95% CI 36.4, 64.6) in participants with RA and ILAs or indeterminate for ILAs compared with 22.0% (95% CI 7.5, 36.5) for RA cases without ILAs and 21.2% (95% CI 20.6, 21.9) for non-RA comparators without ILAs.

Kaplan–Meier cumulative mortality curves stratified by RA and ILA status are shown in Figs 3 and 4. In the Cox regression models adjusted for age, sex, current/past smoking status, pack-years and BMI, RA had an HR for mortality of 1.52 (95% CI 1.07, 2.14), as shown in Fig. 3. After stratifying by ILA status, the HR for mortality was 0.89 (95% CI 0.24, 3.27) for the RA cases with ILAs and 4.21 (95% CI 1.03, 9.20) for the RA participants indeterminate for ILA findings compared with non-RA comparators without ILAs (Fig. 4A). When we combined RA with ILAs or indeterminate for ILAs and compared with participants without ILAs, the group with ILAs/indeterminate for ILAs had an HR of 2.64 (95% CI 1.76, 3.97) for death compared with participants without RA or ILAs (Fig. 4B). When we limited our analysis to RA cases, RA cases with ILAs or indeterminate for ILA findings had an HR of 2.86 (95% CI 1.33, 6.16) compared with RA cases without ILAs. There were no significant differences in the HR for mortality comparing RA and non-RA participants without ILAs [HR 0.92 (95% CI 0.48, 1.78)]. Mortality comparisons using non-RA with ILAs as the reference group are reported in Supplementary Tables S5 and S6, available at Rheumatology online.

Figure 3.

Cumulative mortality and HRs stratified by RA/comparator in COPDGene (N = 8808)

Figure 4.

(A) Cumulative mortality and HRs stratified by RA case/non-RA comparator and ILA status in COPDGene (N = 8808). *Among RA cases, presence of ILAs may be considered ‘preclinical ILD’. The multivariable model was adjusted for age, sex, smoking status (current/past), pack-years and BMI. (B) Cumulative mortality and HRs stratified by RA/comparator and ILA status in COPDGene (N = 8808). *Among RA cases, presence of ILAs may be considered ‘preclinical ILD’. The multivariable model was adjusted for age, sex, smoking status (current/past), pack-years and BMI

Sensitivity analysis

Results from a sensitivity analysis using a less stringent RA case definition that included any COPDGene participant who self-reported an RA history regardless of DMARD use (n = 683) are reported in Supplementary Tables S7–S9 and Supplementary Figs S9 and S10, available at Rheumatology online. Overall, these results were attenuated compared with the main results. The prevalence of ILAs and indeterminate for ILAs was higher in RA cases compared with non-RA comparators (7.9% and 41.6% vs 5.0% and 33.8%, respectively). Mortality was higher among RA cases with ILAs or indeterminate for ILAs compared with non-RA without ILAs [adjusted HR 1.87 (95% CI 1.56, 2.23)] and RA without ILAs [adjusted HR 1.31 (95% CI 1.03, 1.66)].

Discussion

In this large prospective, comparative study, RA was strongly associated with ILAs, even after adjustment for cigarette smoking, the MUC5B promoter variant and other risk factors. The prevalence of ILAs in RA was 17% compared with 5% in non-RA in this cohort composed of smokers without clinical evidence of ILD or bronchiectasis. Furthermore, we found that RA with ILAs or indeterminate for ILAs had a 3-fold increased mortality compared with without ILA. These findings suggest that ILAs have important prognostic significance in RA patients and emphasize the urgent need for additional research investigating the screening, monitoring and early treatment strategies for preclinical ILD in RA, especially in high-risk patients with a significant smoking history.

Our study adds to the existing literature investigating preclinical lung disease in RA. One study of 177 RA patients in the ESCAPE RA cohort found that 32% had at least one ILD feature on cardiac CT scan [34]. However, this study did not include non-RA comparators. An additional investigation compared ESCAPE RA patients with non-RA controls in the Multi-Ethnic Study of Atherosclerosis (MESA) and found evidence of an association between RA and increased high attenuation areas in the lung parenchyma captured on cardiac CT scans [35]. The MESA study also noted an association between high-attenuation areas and RA-related autoantibodies in a separate study [36]. However, these studies did not investigate ILAs, use dedicated full-lung HRCT imaging or specifically exclude patients with known ILD. Furthermore, none specifically investigated only smokers and instead included a combination of smokers and non-smokers. Several additional prospective studies have investigated the prevalence of preclinical lung disease in populations of both smoking and non-smoking RA patients using lung HRCT. One screening study in the BRASS registry performed research HRCT scans on RA patients without a known ILD diagnosis and found evidence of ILAs in 15% [20]. Another recent investigation performed in the French ESPOIR and TRANSLATE2 cohorts obtained HRCTs of RA patients without an ILD history, respiratory symptoms or signs of pulmonary disease on examination [22]. This study noted preclinical ILD in 16.9% and 19.0% of the subjects. Similar investigations performed in small cohorts of RA patients without known lung disease at the National Institutes of Health (Bethesda, MD, USA) and Cedars-Sinai (Los Angeles, CA, USA) noted preclinical ILD on HRCT in 33% and 39%, respectively [37, 38]. Perhaps not surprisingly, the prevalence of preclinical ILD may be higher on HRCT performed for clinical purposes. A separate study performed in the BRASS registry used a sequential reader method to assess for evidence of ILAs on clinically-performed CT scans of RA patients [20]. This investigation noted evidence of ILA changes in 45% of RA patients, including 40% of patients without a known history of pulmonary fibrosis [21]. These studies did not include non-RA comparators or account for potential risk factors like smoking and the MUC5B promoter variant.

Although a high prevalence of preclinical lung abnormalities has been well established in RA patients, less is known about the progression of these abnormalities and their impact on mortality. One study from Cedars-Sinai noted progression of ILAs in 86% of RA patients after 1 year [38]. A retrospective investigation performed in Brazil noted progression of ILAs in 29% of RA patients who had clinical HRCTs performed [14]. However, neither study included non-RA comparators. Furthermore, although risk factors for progression of ILAs have been identified in the general population, including the fibrotic subtype of ILA, risk factors for ILA progression in RA remain largely unknown. One study did not find any impact of the MUC5B promoter variant on ILD progression in RA patients [39].

Although multiple studies have demonstrated increased mortality associated with clinically apparent RA-ILD, none have investigated the impact of preclinical RA-ILD on mortality. Unexpectedly, we found that mortality was highest in the RA group that was indeterminate for ILAs and we did not see a significant association between ILA and mortality in the RA subgroup. This may be due to the small sample size of the RA with ILAs subgroup. Alternatively, it remains possible that the lung abnormalities classified as indeterminate for ILAs have particular importance among RA patients and lead to higher mortality, particularly in this subgroup. Notably, in one previous study that included several large cohorts, indeterminate for ILA findings were associated with increased mortality in two of the four cohorts [15]. In one of the cohorts, the indeterminate for ILA group had higher mortality than the ILA group, similar to our findings [15]. Our finding of increased mortality in the RA with indeterminate for ILAs subgroup may suggest unique importance for indeterminate findings in this subgroup. Additional studies validating our findings and investigating the mortality impact of specific radiologic interstitial abnormalities in studies with sufficient sample size that include both RA cases and non-RA comparators are needed.

Our study offers epidemiologic evidence of an important link between RA and ILAs and found a particularly strong association with fibrotic ILAs. The mucosal origins hypothesis for RA pathogenesis posits that damage to the airway and respiratory mucosa leads to aberrant inflammation, dysbiosis, citrullination of peptides and the generation of RA-related autoantibodies [40]. Evidence for this hypothesis derives from several studies that have noted evidence of pulmonary inflammation and RA-related autoantibodies in sputum samples prior to the onset of joint inflammation [36, 41]. Despite recognition of the important role of the lung as a mucosal site of autoantibody production, we have limited understanding of why only a minority of RA patients with ILAs will progress to clinically apparent RA-ILD. These mechanisms may relate to underlying genetic risk factors like the MUC5B promoter variant or involve ongoing injury to the pulmonary parenchyma through the direct effects of inflammation, medications or recurrent respiratory infections, smoking or other inhalants as a trigger in the setting of immunosuppression [42, 43]. Further studies are needed to characterize risk factors for progression and determine pathways that can be targeted by medications that arrest progression of lung disease.

Our study has several strengths. This investigation was performed using a large, multicentre, prospective, longitudinal cohort that collected extensive details on smoking history and respiratory risk factors. We specifically investigated the presence of ILAs in RA compared with non-RA and carefully adjusted for smoking status and pack-years. Furthermore, we determined ILA status using a validated sequential reading method with expert reviewers who were blinded to the exposures of interest of this study. Lastly, there was a long duration of follow-up to examine mortality risk, and deaths were systematically captured by a combination of longitudinal follow-up and queries of the social security death index.

Our study also has several limitations to consider. First, we relied on a combination of self-reported RA status and the use of DMARD medications to identify participants with and without RA. Other research studies have demonstrated a positive predictive value of 88% for RA using similar methods [27, 28]. Second, because COPDGene was not investigating RA, we have limited information on RA-related covariates. Thus we were unable to examine how immunosuppressive medication use, disease activity, serostatus, bone erosions or RA duration may impact ILA risk in RA [44]. However, we were able to account for important lifestyle, demographic and genetic risk factors that are known risk factors for RA-ILD and may influence mortality [32, 45]. Third, the prevalence of RA in COPDGene was 0.8%, which may be lower than expected. While the prevalence of RA is sometimes quoted as 1%, the prevalence ranges have been lower than this in recent systematic reviews [46, 47]. We may have missed some RA cases since we relied on self-reports and DMARD use. While smoking is a risk factor for RA, the majority of COPDGene participants are men, so may have altered the expected prevalence of RA. Finally, COPDGene is a prospective study but is not population-based. It is also possible that having a chronic disease such as RA may have affected the willingness to participate. Fourth, since the COPDGene study was limited to White and Black smokers, our results may not be generalizable to other populations, including non-smokers. However, smoking remains an important risk factor for RA development and smokers represent a high-risk group that would be a likely target of future screening strategies [6, 12, 38, 45]. Fifth, despite the large sample size of the COPDGene cohort, the RA group was relatively small. While we were able to detect associations of RA with specific ILA subtypes such as fibrotic ILA, the mortality investigation was relatively limited due to the small number of deaths in this subgroup. Similarly, due to the size of our RA cohort, we were only able to study all-cause mortality rather than cause-specific mortality [12]. Larger studies with longer follow-up are needed to determine the progression of specific ILA subtypes among RA patients.

In conclusion, we found that RA was strongly associated with ILAs in this large comparative study. The association between RA and ILAs persisted after adjustment for smoking intensity, smoking duration and the MUC5B promoter variant. RA patients with ILAs or indeterminate for ILAs had a 3-fold increased mortality compared with those without ILAs. These findings emphasize the significance of ILAs among RA patients and the need for further research to establish the clinical utility of screening, monitoring and early treatment strategies for preclinical ILD.

Supplementary material

Supplementary material is available at Rheumatology online.

Data availability

Data are available upon reasonable request and with appropriate institutional review board approval.

Authors’ contributions

G.C.M. and J.A.S. conceptualised and designed the study. M.M., M.H.C., G.L.K., A.H., Ta.H., To.H., M.Y., M.N., E.A.R., H.H., G.M.H., and E.K.S. collected the data. K.H. and K.Y. analsed the data. G.C.M. and J.A.S. prepared the initial manuscript. All authors interpreted the results and read and approved the final manuscript. All authors had full access to all the data in the study and accept responsibility to submit for publication. G.C.M., K.H., and J.A.S. have accessed and verified all the data in the study.

Funding

G.C.M. is supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases (grant T32 AR007530) and the VERITY Pilot and Feasibility Award. J.A.S. is supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases (grants R01 AR077607, P30 AR070253 and P30 AR072577), the R. Bruce and Joan M. Mickey Research Scholar Fund and the Llura Gund Award for Rheumatoid Arthritis Research and Care. M.H.C. was supported by grants from the National Heart, Lung, and Blood Institute R01 HL153248, R01 HL149861, R01 HL147148. T.J.D. is supported by the National Institutes of Health, National Health, Lung, and Blood Institute (grant R01 HL155522). COPDGene was supported by awards U01 HL089897 and U01 HL089856 from the National Heart, Lung, and Blood Institute. COPDGene is also supported by the COPD Foundation through contributions made to an industry advisory board that has included AstraZeneca, Bayer Pharmaceuticals, Boehringer Ingelheim, Genentech, GlaxoSmithKline, Novartis, Pfizer and Sunovion.

Disclosure statement: K.Y. reports consulting fees to OM1 unrelated to this work. M.M. reports institutional grant support from Bayer. M.H.C. has received grant funding from GlaxoSmithKline and Bayer and speaking or consulting fees from AstraZeneca, Illumina and Genentech unrelated to this work. T.J.D. reports grant funding and other support from Bayer and Bristol Myers Squibb, consulting fees from Boehringer Ingelheim and L.E.K. Consulting and has been part of a clinical trial funded by Genentech, all unrelated to this study. P.F.D. reports grant support from Genentech and Bristol Myers Squibb, royalties or licences from Up To Date and participation on a Food and Drug Administration advisory board. R.S.J.E. received grant support from Boehringer Ingelheim and is a co-founder and equity holder of Quantitative Imaging Solutions. M.N. reports research grants from Canon Medical Systems, AstraZeneca and Daiichi Sankyo and consulting fees from AstraZeneca and Daiichi Sankyo. G.R.W. reports grants from Boehringer Ingelheim; consultancy for Pulmonx, Janssen Pharmaceuticals, Novartis and Vertex and is founder and co-owner of Quantitative Imaging Solutions. H.H. reports grants or contracts and consulting fees from Canon Medical Systems and Konica-Minolta and consulting fees from Mitsubishi Chemical. G.M.H. reports consulting fees from Boehringer Ingelheim, the Gerson Lehrman Group and Chugai Pharmaceuticals unrelated to this work. E.K.S. has received grant support from GlaxoSmithKline and Bayer unrelated to this work. J.A.S. has received research support from Bristol Myers Squibb and performed consultancy for AbbVie, Amgen, Boehringer Ingelheim, Bristol Myers Squibb, Gilead, Inova Diagnostics, Janssen, Optum and Pfizer unrelated to this work. All other authors report no conflicts of interests.