This is the authors' abstract. We don't add key points for this paper.
Background/Objectives: To examine how the estimated association between baseline intraocular pressure (IOP) and technical failure after selective laser trabeculoplasty (SLT) depends on the definition of failure, and whether the observed pattern is compatible with criterion arithmetic and measurement error. Methods: In 144 eyes of 144 patients (156 patients and 168 treated eyes screened), technical failure was evaluated under six definitions (three algebraically coupled definitions based on reduction from baseline [a <20% reduction, an absolute reduction <3 mmHg and the primary composite], two fixed IOP thresholds, and an IOP-free comparator) using Cox proportional hazards models. A criterion had to be met at two consecutive visits, or at the last available visit when no confirming visit was possible. Additional analyses included a same-criterion confirmation rule, confirmed-only event validation, a time-varying coefficient model, extended multivariable adjustment, an individualized target-IOP definition, and a calibrated simulation varying the baseline-to-achieved slope, measurement error, and selection. Results: Technical failure occurred in 119 eyes (82.6%) during total follow-up and in 112 (77.8%) by month 24; of the 110 failed eyes with a 24-month measurement, 70 (63.6%) had IOP ≤ 21 mmHg. Baseline-IOP hazard ratios (HRs) were 0.873–0.950 under coupled definitions and 1.075–1.109 under absolute thresholds. Requiring the same criterion at two consecutive visits reduced the composite event count to 99 and gave HR 0.954 (95% CI 0.91–1.00); the cross-definition pattern was unchanged. Censoring the 19 terminal single-visit events (confirmed-only analysis; 100 events) gave HR 0.950 (0.91–0.99); extended adjustment gave 0.948 (0.90–0.99); the individualized target-IOP definition gave 121 events and HR 0.942 (0.90–0.98); and the IOP-free comparator gave 0.965 (0.90–1.03), which was not significant but could not be distinguished from the primary estimate (bootstrap p = 0.58). An exploratory time-varying coefficient model suggested that the protective association was concentrated in the early post-laser period (interaction p = 0.003). In simulation, the reversal required an observed baseline-to-achieved slope below unity; regression to the mean, treatment response and selection on the baseline reading each produced a spread of similar size and could not be distinguished. Conclusions: Outcome definition modifies the estimated direction of the baseline-IOP association, but this does not establish a biological or causal effect. Under classical-error assumptions, shared measurement error moved coupled estimates toward apparent protection rather than toward the null. The non-significant IOP-free comparator reflects limited precision, not evidence that the association disappears. Technical failure should be distinguished from clinical failure, and confirmation rules should be stated explicitly.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Evcimen et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: