Meta-analysis identifies the major impact of patient classification on the ACPA association with rheumatoid arthritis-associated interstitial lung disease (RA-ILD)

We need screening for RA patients at high risk of RA-ILD to prevent the associated decrease in life quality and survival. The proposed screenings disagree regarding the anti-citrullinated protein antibodies (ACPA) because of their inconsistent association across recent studies. Therefore, we hypothesized that meta-analysis of the published reports should reveal clues explaining the heterogeneity of results, helping us progress in RA-ILD early detection. We aimed to discover the factors accounting for the variability of the ACPA association in the published reports. We searched the Web of Science and PubMed databases for studies reporting ACPA in RA-ILD and RA-control groups. The identified studies were analyzed using meta-analysis and meta-regression to identify moderators of the ACPA association. We found 513 unique records, containing 31 eligible data sets. The meta-analysis preceding the search for moderators showed a remarkable heterogeneity (pQ = 5.7 × 10–7). Appropriate tests showed that it was largely attributable (58.1%) to an outlier study, which had recruited cases and controls in different place and time contexts. The exclusion of this outlier from subsequent analyses did not completely remove heterogeneity (pQ = 0.004). However, it permitted the identification of the patient classification method as a significant moderator: The 14 studies using chest CT showed stronger ACPA association with RA-ILD (OR = 3.05 [95%CI: 2.12–4.38]) than the 16 employing multifactorial criteria (1.55 [95%CI: 1.18–2.03]; p = 0.0047 for the contrast). This moderator accounted for the significant heterogeneity (pQ = 0.079), was robust in sensitivity analyses, and was the only one found. Our results validate the ACPA association with RA-ILD, reinforce the importance of study design, and suggest the need to consider if studies relying on chest CT for classification could be more fruitful in the search for RA-ILD biomarkers.