Divergent risk of central visual loss after severe-stage glaucoma surgery: the opposing roles of early postoperative hypotony and hypertension, stratified by baseline vision

Background: Central visual loss (CVL) is a devastating complication after surgery for severe-stage glaucoma. The mechanisms underlying CVL remain poorly understood. We hypothesized that early postoperative intraocular pressure (IOP) dynamics, rather than the choice of surgery, primarily determine visual prognosis, with risk influenced by baseline visual reserve. Methods: This retrospective cohort study included 523 eyes of 523 patients with severe-stage glaucoma (mean deviation <−20 dB) who underwent trabeculectomy or Ahmed glaucoma valve (AGV) implantation. The primary outcome was CVL at 3 months postoperatively. To address baseline confounding, the primary analysis focused on identifying risk factors for CVL using Firth penalized logistic regression, with a prespecified subgroup analysis stratified by baseline best-corrected visual acuity (BCVA) and prespecified sensitivity analyses. Results: The study cohort exhibited significant confounding by indication, with the AGV group representing a higher-risk population. Despite these disparities, 1-year outcomes, including CVL incidence, were comparable (18.2% vs. 14.0%, P = 0.198), with consistent findings in propensity score-adjusted analyses. The most powerful predictors of CVL were not the surgery type but early postoperative IOP instability. A dichotomous risk profile emerged from the stratified analysis: in eyes with good baseline vision (BCVA >20/200), 1-month hypotony (IOP <8 mmHg) was the dominant risk factor (OR 3.46; 95% CI 1.56–7.51; P = 0.003). Conversely, in eyes with poor baseline vision (BCVA ≤20/200), 1-month hypertension (IOP >21 mmHg) was the sole independent predictor (OR 2.89; 95% CI 1.03–8.11; P = 0.038). A persistently high IOP trajectory predicted the poorest visual recovery. Conclusions: The risk of CVL after severe-stage glaucoma surgery is not monolithic but is fundamentally conditioned by baseline visual reserve. Visual prognosis depends less on the surgical procedure and more on maintaining early postoperative IOP stability within a tailored range. These findings compel a shift from a uniform IOP target toward an individualized, risk-stratified postoperative management strategy.