We thank Qiu and Cen for their thoughtful comme-ntary on our study and welcome the opportunity to clarify our approach.
Regarding the first point on reverse causation, we respectfully note that our definition of delayed follow-up (DFU) was designed to minimize this concern. DFU was defined as the deviation from each patient’s physician-recommended follow-up interval rather than the absolute interval between visits. Specifically, DFU was computed as: Actual Visit Date − (Prior Visit Date + Provider-Recommended Follow-Up Interval). Because the recommended interval already reflects the treating physician’s assessment of disease severity, stability, and progression risk, DFU reflects departures from individualized care plans rather than physician risk stratification itself. For example, a patient judged to be at high risk may receive a shorter recommended follow-up interval yet still accumulate substantial DFU if that recommendation is not followed. Thus, DFU is measured relative to the physician’s intended follow-up schedule rather than the underlying disease severity. As acknowledged in our manuscript, residual confounding cannot be fully excluded in any retrospective design.
We considered time-dependent analyses during study design before selecting our final methodology. While time-dependent models are valuable for evaluating changing exposures over time, they require assumptions that DFU has an immediate or proximal effect on subsequent visual field progression, which is difficult to justify for a slowly progressive disease such as glaucoma. Moreover, the resulting model coefficients are less readily interpretable for clinicians because follow-up intervals are determined before future glaucoma progression is known. Furthermore, we used three complementary metrics (percentage of visits with DFU, mean days of DFU, and maximum days of DFU), which capture different patterns of delayed care. For example, maximum days of DFU isolates the effect of a single prolonged interruption, while mean days of DFU may better capture multiple intermittent short delays. The consistent findings across all three metrics strengthen confidence in our conclusions.
We agree that treatment escalation may mediate the relationship between delayed follow-up and glaucoma progression. However, the dataset used in this study did not contain structured longitudinal information on treatment escalation. Accurately identifying changes in medical therapy, laser procedures, or glaucoma surgery would therefore require extensive manual chart review, which is unlikely to provide reliable ascertainment for a cohort of this size. We are currently addressing this question using the SOURCE dataset, which contains detailed longitudinal treatment information in a substantially larger population. Together with our ongoing studies evaluating OCT and additional visual field outcomes, these analyses will provide a more comprehensive assessment of the consequences of delayed care than visual field trend analysis alone.
We are grateful to Qiu and Cen for their thoughtful comments and for advancing this discussion. We look forward to continued research that will further refine evidence-based, individualized follow-up strategies for glaucoma patients.
CRediT authorship contribution statement
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
