Lower HLA-B27 positivity and heterogeneous late-onset case proportions across axial spondyloarthritis cohorts: a systematic review and meta-analysis.

To estimate the proportion of late-onset ankylosing spondylitis (AS)/axial spondyloarthritis (axSpA) cases defined by study-reported symptom, axial-symptom, or back-pain onset using > 45- or ≥ 45-year thresholds in clinical cohorts or registries, and to compare later- versus earlier-onset phenotypes. The nine original databases were searched through 20 August 2026, and DOAJ was added as a supplementary source. The prespecified primary study-family set was synthesised using a logistic-normal mixed model. Compatible phenotypes were pooled using REML random-effects models with Hartung-Knapp intervals. Risk of bias and certainty were assessed using design-matched tools and an adapted prognostic-factor framework. Fifty-four reports (42 independent studies) were included. Six prespecified study families contributed 745 late-onset cases among 6,631 patients with assessable onset age; the pooled proportion was 13.43% (95% CI 9.23%-19.14%; prediction interval 3.63%-39.02%; I²=93.3%; very low certainty). HLA-B27 positivity was lower with later onset (4 studies; n = 2,593; OR 0.47, 95% CI 0.34-0.67; I²=0%; moderate certainty). Male sex (5 studies; OR 0.66, 95% CI 0.32-1.33), radiographic sacroiliitis/radiographic axSpA (3 studies; OR 0.70, 95% CI 0.43-1.15), and supplementary BASDAI evidence (3 studies; MD 1.45, 95% CI - 1.18 to 4.09) were inconclusive. Other incompatible phenotypes were not pooled. Late-onset cases averaged approximately 13% in included cohorts and registries, but heterogeneity was high, the prediction interval wide, and certainty very low. Lower HLA-B27 positivity was the most consistent association but relied on four unadjusted comparisons and cannot be interpreted causally or diagnostically. PROSPERO registration: CRD420261469724 (https://www.crd.york.ac.uk/PROSPERO/view/CRD420261469724).