Longitudinal GGT Trajectories Identify Prognostic Phenotypes in Paediatric Primary Sclerosing Cholangitis

Primary sclerosing cholangitis (PSC) is characterised by substantial clinical heterogeneity. While normal gamma-glutamyl transferase (GGT) at one year predicts favourable outcomes, the heterogeneous trajectories underlying this remain uncharacterised. We applied latent class mixed modelling to identify distinct GGT trajectory phenotypes in paediatric PSC and evaluate their prognostic significance.

We analysed 782 children with PSC from the Pediatric PSC Consortium who had baseline GGT and at least one additional measurement over three years. Latent class mixed models were fit to log-transformed GGT values, with optimal model selection based on Bayesian Information Criterion and entropy. Cox proportional hazards models assessed associations between trajectory clusters and transplant-free survival, adjusting for baseline characteristics. We used multinomial logistic regression to assess whether cluster membership could be predicted using early markers.

Four trajectory clusters were identified: persistently elevated (23%), moderate-stable (30%), elevated-declining (26%), and persistently low (21%). 5-year transplant-free survival differed markedly: 74%, 86%, 98%, and 96%, respectively (p<0.0001). The elevated-declining and moderate-stable clusters both had elevated baseline GGT but divergent outcomes, with the elevated-declining cluster achieving normalisation by year one. In adjusted analyses, persistently elevated and moderate-stable clusters had significantly higher transplant hazard versus persistently low (HR 6.45, 95% CI 2.87-14.5; HR 3.00, 95% CI 1.31-6.90). Transplant-free survival in the elevated-declining cluster was similar to the persistently low cluster (p=0.22). Favourable clusters had younger age with higher IBD prevalence. Cluster membership could be predicted using baseline and 6-month GGT values (AUC 0.83).

GGT trajectory phenotypes identify clinically meaningful PSC subgroups with distinct prognoses. Early biochemical response, rather than baseline values alone, predicts long-term outcomes.

Prognostic assessment in PSC has relied on cross-sectional biomarker measurements at single timepoints, limiting the ability to capture dynamic changes over time. By applying trajectory modelling to repeated GGT measurements in 782 children with PSC, we identified four distinct groups with significantly different transplant-free survival. Two clusters with similarly elevated baseline GGT followed divergent trajectories and had very different outcomes, information that cross-sectional assessment would miss. These trajectory phenotypes can be predicted using baseline and 6-month GGT values, offering clinicians an early tool to identify high-risk patients. Trajectory-based phenotyping may also facilitate clinical trial design by enabling patient enrichment strategies and providing an early surrogate endpoint.