Performance of referral strategies for patients with suspected axial spondyloarthritis: a systematic review with meta-analysis

Background Axial spondyloarthritis (axSpA) is a clinically important but often under-recognized condition, with substantial diagnostic delay. Structured referral strategies have been proposed to facilitate earlier identification of suspected cases; however, their diagnostic accuracy has not been comprehensively compared. We therefore conducted a systematic review to synthesize the diagnostic accuracy of referral strategies for identifying axSpA in patients with chronic back pain. Methods We performed a systematic review of observational studies evaluating referral strategies for suspected axSpA. The review protocol was prospectively registered in PROSPERO (CRD42025634153). MEDLINE, Embase, Web of Science, CENTRAL, and CINAHL Plus were searched from inception to December 2025. Study selection, data extraction using the "Checklist for Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modeling Studies" (CHARMS), risk of bias assessment using the "Prediction model Risk Of Bias Assessment Tool + Artificial Intelligence" (PROBAST+AI), and certainty of evidence rating using an adapted "Grading of Recommendations Assessment, Development and Evaluation" (GRADE) approach were conducted by two independent reviewers. Findings Twenty-four studies evaluating 40 referral strategies were included. Risk of bias was high across all strategies, resulting in low to very low certainty of evidence for most performance metrics. Meta-analysis was feasible for nine referral strategies. Statistically significant pooled likelihood ratios were observed for the negative likelihood ratio (LR−) of the ASAS referral criteria (0.18, 95% CI 0.04–0.81), and for both the positive and negative LRs of the Braun 2-step alternative strategy (LR+ 2.02, 95% CI 1.47–2.76; LR− 0.38, 95% CI 0.19–0.75), CaFaSpA ≥1 (LR+ 1.64, 95% CI 1.04–2.59; LR− 0.43, 95% CI 0.21–0.87), and RADAR 2/3 (LR+ 2.76, 95% CI 2.13–3.59; LR− 0.60, 95% CI 0.51–0.71). These results did not meet thresholds (e.g., LR+ >10, LR− <0.10) for strong diagnostic accuracy. Interpretation Current referral strategies for axSpA have limited diagnostic impact and are supported by low to very low-certainty evidence. Future research should prioritize the development of new referral strategies, particularly for primary care settings. Funding No specific funding was received for this work.