Adjunctive ab-interno goniotomy in chronic angle-closure glaucoma: a retrospective proof-of-concept pilot study using...

To assess whether an anterior segment optical coherence tomography (AS-OCT)-informed causal framework could estimate the added benefit of adjunctive ab-interno goniotomy during phacoemulsification with goniosynechialysis (Phaco-GSL) in chronic angle-closure glaucoma (CACG).

Retrospective single-center comparative cohort pilot study.

In a single-center development cohort of 102 eyes, 54 underwent phacoemulsification with goniosynechialysis alone and 48 underwent adjunctive ab-interno goniotomy. Marginal treatment effects were estimated using stabilized inverse probability of treatment weighting (IPTW) and augmented inverse probability weighting (AIPW). Medication-free complete success was defined as intraocular pressure (IOP) 5-18 mmHg with at least 20% reduction from baseline without IOP-lowering medications or additional glaucoma surgery; qualified success used the same IOP criteria with or without medications. Qualified failure was defined as loss of qualified success or need for additional glaucoma surgery. Time to qualified failure through 24 months was analyzed using stabilized IPTW-weighted Kaplan-Meier curves and weighted Cox models. Predicted individualized benefit estimates for 24-month medication-free complete success were derived from baseline clinical and AS-OCT features using a doubly robust learner with ridge regularization and 5-fold cross-fitting.

Adjunctive goniotomy was associated with higher 24-month qualified-failure-free survival under stabilized IPTW adjustment (0.682 vs 0.425; weighted hazard ratio 0.475, 95% CI 0.257-0.880; p=0.018). Doubly robust estimates favored adjunctive goniotomy for 24-month medication-free complete success (risk difference 0.291; 95% CI 0.105-0.478) and qualified success (risk difference 0.233; 95% CI 0.064-0.403). Adjusted mean IOP differences were modest, whereas medication-free complete success and medication burden showed more apparent separation. Hyphema occurred numerically more often with adjunctive goniotomy. Predicted individualized benefit estimates (model-predicted absolute probability differences) were variable and were not interpreted as clinically actionable recommendations.

Adjunctive goniotomy was associated with more favorable 24-month surgical control in this retrospective pilot cohort. Imaging-informed doubly robust learning may support future studies of treatment-effect heterogeneity, but the current predicted individualized benefit estimates require external validation before clinical use.