A health sciences librarian (MGR) designed a systematic search strategy in the databases PubMed (Medline), Embase (Elsevier), and Cochrane Library (Wiley). The search included MeSH terms and free text with their different combinations referring to “rheumatoid arthritis”, “interstitial lung disease” and “screening” until April 2020 (Appendix B, Annex 1 of the supplementary material). The search was limited to human studies and articles published in English, French, or Spanish. A hand search of the references of the articles retrieved was also performed.

The Covidence systematic review tool (www.covidence.org) was used to screen studies. Studies of any design using screening methods for interstitial lung disease in patients with RA were selected. As per protocol, studies were included in which: (a) the study population was adult patients with RA; (b) the intervention consisted of any method of screening for ILD (questionnaire, examination, complementary tests…); and (c) the validity or reliability of the screening method was assessed, or screening criteria were described. Two reviewers (SGC and LSF) first independently screened by title and abstract. Any discrepancies were resolved by consensus. The selected articles were then evaluated in detail to select those for final inclusion based on the above criteria. A third reviewer (DSM) resolved any disagreement on inclusion between the two reviewers. The reason for excluding all the rejected studies was recorded.

The two reviewers extracted information from the articles independently using specific templates. The retrieved articles were classified according to the screening method studied (HRCT, PFT, ultrasound, etc). Study characteristics were extracted in terms of design, population included, screening method, gold standard used, and results of comparison. A synthesis of the collected information was made in a narrative form with tabulation of the characteristics and results of each included study. Study quality was assessed using the scale for levels of scientific evidence and formulating recommendations for diagnostic questions (NICE [National Institute for Health and Care Excellence]14 adaptation of the levels of evidence of the Oxford Centre for Evidence Based Medicine and Centre for Reviews and Dissemination15).

Studies by Pancaldi et al.19 and Manfredi et al.24 compare the use of a digital stethoscope to automatically detect velcro-type lung crackles with HRCT findings. The digital audio collected was analysed using an algorithm for the binary classification of the findings (VECTOR). The target population comprised patients diagnosed with RA who had undergone HRCT for respiratory symptoms, velcro crackles, pathological lung function, or lung nodule study. Digital auscultation achieved diagnostic accuracy of 84%–90%, specificity of 77%–88%, sensitivity of 93%, positive predictive value (PPV) 75%–83%, and negative predictive value (NPV) 94%–95% in detecting ILD, and the accuracy obtained by auscultation by a rheumatologist was 67.2%.

The study by Abdel-Wahab et al. established a significant association between serum levels of interleukin-33 (IL-33) and the presence of DIDP in patients with RA.18 Castellanos-Moreira et al.28 assessed the presence of autoantibodies against 3 carbamylated antigens: foetal calf serum (FCS), fibrinogen (Fib), and chimeric fibrin/filaggrin homocitrullinated peptide (CFFHP) in RA patients. Included patients without a previous diagnosis of ILD were screened with HRCT. It was found that all anti-carbamylated protein antibodies studied were more frequent and with higher mean titres in the ILD group. In addition, logistic regression adjusted for age, RA duration, anti-cyclic citrullinated peptide antibodies (anti-CCP), rheumatoid factor (RF), sex and cumulative tobacco dose showed that anti-FCS, anti-CFFHP and anti-FCS-IgA were independently associated with ILD. Another study22 that evaluated different serum biomarkers in a Chinese cohort (later confirmed in a US cohort) showed an association between levels of extracellular matrix metalloproteinases 7 (MMP-7) and IFNγ-inducible protein-10 (IP-10) and the presence of (HRCT-diagnosed) ILD in RA patients. This finding was confirmed in patients with clinical and subclinical ILD. Biomarker levels were correlated with the severity of the ILD. The area under the ROC curve (Receiver Operating Characteristic curve) of these markers for the diagnosis of ILD reached values between 0.68 and 0.86. In a study26 with fewer RA patients who underwent lung HRCT, higher levels of CA15-3 and CA125 were laboratory determined in patients with ILD than in those without ILD. Finally, the study by Doyle et al.29 showed a significant increase in MMP-7, PARC, and surfactant protein D in patients with subclinical ILD in two different cohorts.

The search yielded 2 studies conducted in the late 1980s30,38 in which BAL was used to screen for ILD in RA patients. The diagnosis of ILD was established by plain chest X-ray and patients underwent BAL for cellular and biochemical analysis. The study by Tishler et al.30 showed that patients with radiological abnormalities had a higher percentage of lymphocytes in the BAL than patients with normal X-rays. Gilligan et al.38 detected subclinical ILD in 15% of patients by chest X-ray and PFT. In the BAL of patients with established ILD, there was a significant difference in the number of neutrophils compared to controls, as well as an increase in collagenase and N-terminal procollagen type III. Gochuico et al.8 compared BAL findings between RA patients with biopsy-confirmed pulmonary fibrosis, RA and subclinical ILD diagnosed by HRCT and RA patients without lung disease. Platelet-derived growth factor (PDGF)-AB and PDGF-BB levels were higher in patients with ILD versus those with RA without ILD. Moreover, significantly higher levels of interferon gamma and transforming growth factor beta 1 (TFG-β1) were detected in patients with ILD with progression of lung damage versus those without.

PFTs have been assessed in several studies for the diagnosis of ILD. The study by Mohd Noor et al.34 estimated that PFT abnormalities exist in 95% of patients with RA of more than 5 years’ disease duration, 66.6% of them with restrictive pattern. Zhang et al.33 observed that patients with ILD had significantly more PFT abnormalities consisting of restrictive pattern and decreased diffusion. Mori et al.35 found that patients with ILD had significantly lower DLCO and DLCO/VA (DLCO corrected for alveolar volume) values than patients without ILD. Wang et al.26 proposed a DLCO cut-off point of less than 52.95% for the detection of HRCT-confirmed ILD. This DLCO screening method achieved sensitivity of 100% and specificity of 61%. Yilmazer et al.36 found a significant association between DLCO values <75% and respiratory symptoms with the presence of ILD in multivariate analysis. In the study by Leonel et al.,37 PFT was statistically significant as a predictor of HRCT abnormalities in patients with RA. Manfredi et al.24 studied the accuracy of PFT for the diagnosis of ILD with FVC cut-off points below 70% and DLCO below 47%. They reported an accuracy of 52.8% and 54.9%, sensitivity of 20% and 30.8%, and specificity of 82.1% and 80%, respectively.

Three studies assessed the usefulness of transthoracic lung ultrasound in screening for ILD.16,17,20 Cogliati et al.16 evaluated the presence of B-lines in 72 lung segments (28 anterior and 44 posterior) following anatomical lines. The presence of more than 10 B-lines in a single segment was considered diagnostic of ILD. In this study, the sensitivity of ultrasound compared to HRCT was 92% and specificity 56% in a population of RA patients with suspected ILD. The group of Moazedi-Fuerst et al.,17,20 in addition to B-lines (pathological if more than two lines), considered pleural thickening (pathological if more than 3 mm), and the presence of subpleural nodules (pathological if at least one nodule). Ultrasound was estimated to have a sensitivity of 92%–97%, specificity of 56%–97%, PPV of 94%, and NPV of 99% for the diagnosis of HRCT-confirmed ILD in consecutive RA patients without respiratory symptoms.17,20

HRCT was the most sensitive technique for the diagnosis of ILD compared to chest X-ray or PFT in several studies, although a relationship between HRCT findings and respiratory symptoms could not be established.31,32 In the series of Hassan et al.,25 HRCT detected 5% of ILD in respiratory asymptomatic RA patients. In the study by Gabbay et al.,6 33% of RA patients with RA of less than 2 years' duration had ILD on HRCT, clinically insignificant in 3 out of 4 patients. In the paper by Gochuico et al.,8 33% of patients had subclinical ILD on inclusion in the study. After a mean follow-up of 1.5 years, 57% of these patients were found to have progression of lung damage as assessed by HRCT.

The combination of HRCT and PFT detected subclinical ILD in 17%–55% of RA patients in the different studies.6,8,21,23,27,33 Gabbay et al.6 found that 58% of patients with RA of less than 2 years’ duration had some abnormality suggestive of ILD in the tests performed (HRCT, chest X-ray,PFT, BAL, scintigraphy). Forty-four percent had subclinical ILD (mild abnormalities on HRCT, FFP and BAL).

The studies by Gabbay et al.6 and Dawson et al.27 concluded that plain X-ray is not a highly sensitive technique for the diagnosis of ILD in RA.