Clinical and Systemic Predictors of Primary Trabeculectomy Success at Three Months: A Retrospective Single-Centre Study of 44 Eyes in Indonesia
TL;DR
Primary trabeculectomy achieved three-month qualified success in seven eyes in ten, and diabetes mellitus was the only variable with a clinically meaningful effect size, but the evidence is hypothesis-generation once exact testing and multiplicity are accounted for.
Abstract
Background: Trabeculectomy is the reference filtration procedure when medical therapy fails to control intraocular pressure (IOP), but the weight of its candidate predictors in Indonesian referral practice is poorly quantified. Objective: This study evaluated three-month success of primary trabeculectomy and seven predictors at a West Sumatran tertiary centre. Methods: Retrospective cross-sectional analysis of consecutive eyes undergoing primary trabeculectomy at Dr. M. Djamil General Hospital, Padang, per STROBE. Success was IOP below 21 mmHg on at most two topical agents at three months. Success was estimated with Wilson intervals, association tested exactly, effect size expressed as the conditional odds ratio for failure, and the seven p values Benjamini–Hochberg corrected. Antifibrotic protocol, complications and visual acuity were not recorded. Result: Among 44 eyes, success was 31/44 (70.5%, 95% CI 55.8–81.8). Mean IOP fell from 28.06 to 15.70 mmHg (Friedman p < 0.001; d(av) = 0.90). Diabetes mellitus carried the largest effect: success 45.5% (5/11) versus 78.8% (26/33), conditional odds ratio for failure 4.28 (exact 95% CI 0.82–24.16). The association did not reach significance on an exact test (p = 0.057) and did not survive correction for seven comparisons (q = 0.398). No other variable approached significance. The design could detect an odds ratio of 8.9 at 80% power. Conclusion: Primary trabeculectomy achieved three-month qualified success in seven eyes in ten. Diabetes mellitus was the only variable with a clinically meaningful effect size, but the evidence is hypothesis-generating once exact testing and multiplicity are accounted for. Around 92 eyes are needed to test it.