Identification of clinical phenotypes in Behçet's syndrome using latent class analysis: a step toward precision medicine.

Objectives: Behçet's syndrome (BS) is characterised by extreme clinical heterogeneity, underscoring the need for precise patient classification to enable personalised management. While traditional distance-based cluster analysis (CA) has provided new insights, its deterministic approach may not fully capture the complexity of BS. The primary objective of this study was to define BS clinical phenotypes using latent class analysis (LCA), a probabilistic, model-based clustering method that identifies hidden classes based on unobserved patterns. We also aimed to examine sex-related differences in clinical manifestations and treatment requirements across the identified classes.

Methods: We conducted a retrospective, observational, single-centre study including all adult BS patients followed in our department between 2012 and 2022, targeting a sample of 500 patients. LCA was performed using clinically relevant indicators (sex, oral and genital ulcers, skin lesions, articular involvement and major organ involvement). Models were compared based on fit indices, class number, separation, assignment and size. The final model was selected based on both clinical relevance and statistical performance.

Results: A total of 553 patients (409 males, 144 females) were enrolled, with a mean age of 32±7 years. Five latent classes (C1-C5) with distinct phenotypes were identified. C1 (n=215; 39%), 'vascular type': all patients had vascular lesions, with the highest prevalence of cardiac involvement (12%). C2 (n=171; 31%), 'ocular type': characterised by 100% uveitis and frequent mucocutaneous lesions. C3 (n=40; 7%), 'neurological type': all patients exhibited parenchymal neurological involvement, and 40% had concomitant uveitis. C4 (n=98; 18%), 'skin-mucosa and articular type': marked by 100% oral and genital ulcers, with the highest prevalence of papulopustular lesions (54%) and articular involvement (48%). C5 (n=29; 5%), 'uncertain BS': with 60% uveitis, 48% vascular lesions, and the lowest mucocutaneous involvement. Sex-related clinical differences were observed, with significant male predominance across all major organ classes (C1, C2, C3, and C5), whereas a near-equal sex distribution was noted in the skin-mucosa and articular class (p<0.001). Treatment patterns varied considerably, with higher corticosteroid doses and conventional immunosuppressant use in major organ classes, while biologics were mostly prescribed in the 'ocular class' (C2) and 'uncertain BS' (C5) (p< 0.001).

Conclusions: This study is the first to apply LCA for BS clinical phenotyping, providing a probabilistic classification that uncovers complex patient subgroups. Five latent classes were identified, with distinct clinical profiles, significant sex disparities, and varying therapeutic needs. These findings are crucial for advancing precision medicine in BS and ultimately improving patient outcomes.