The prevalence of interstitial lung disease in rheumatoid arthritis: a systematic review

Interstitial lung disease (ILD) is an important comorbidity in RA. Its true prevalence remains unclear. We assessed reported prevalences across different diagnostic approaches.

We conducted a systematic literature review and meta-analysis to assess RA-ILD prevalence. Embase, MEDLINE (PubMed), Cochrane Library and Web of Science were searched (from inception to 2022). Heterogeneity was assessed (I ²). Subgroup differences were assessed (Q-test). Bias appraisal was performed via a funnel plot and Egger's test. Individual studies were evaluated for the presence of bias using the AXIS score. Mixed effects meta-regression explored sample size, diagnostic modality, geography and AXIS score as moderators.

We included 74 studies and obtained an overall random effects pooled prevalence of 20.6% (95% CI 18.3, 23.2; I ² = 99.7%). The random effects pooled prevalence varied by diagnostic method: 31% (95% CI 25, 38; I ² = 97%) when using high-resolution CT, 21% (95% CI 14, 31; I ² = 95%) with pulmonary function tests, 13% (95% CI 5, 30; I ² = 99%) when combining multiple diagnostic tools, 12% (95% CI 7, 20; I ² = 96%) with chest X-ray and 5% (95% CI 4, 8; I ² = 100%) when based on International Classification of Diseases codes. Low-bias recent studies (AXIS >16; 2013-2022) yielded a pooled prevalence of 10.1% (95% CI 6.2, 16.3; I ² = 99.9%). Prevalences significantly differed between tools (Q = 73.16, df = 4, P < 0.0001). The meta-regression model (QM = 89.96, df = 11, P < 0.0001) had a moderate explanatory value (R ² = 47.82%). The asymmetric funnel plot (t = 2.59, df = 98, P = 0.011) pointed to significant bias (estimate of 4.86, SE = 1.88).

Due to extreme heterogeneity and evidence of small-study effects, the overall pooled random effects prevalence (20%) should be interpreted cautiously and primarily reflects the wide variation in diagnostic approaches. Estimates derived from low-bias and multimodal diagnostic studies (≈10-13%) may provide a more clinically meaningful approximation, although they remain informed estimates rather than definitive prevalence rates.