Objective:

To raise technical concerns about reported risk-set accounting and the timing of NAION risk in a study comparing semaglutide with SGLT2 inhibitors among US veterans.

Approach:

• Risk-set accounting: The letter compares the SGLT2i risk-set counts at day 0 and day 182.5 with reported NAION events and censoring.
• Temporal risk: It notes when the Kaplan-Meier curves separate and asks whether the hazard ratio varies between the first treatment year and later follow-up.

Key Findings:

• The SGLT2i at-risk population reportedly declined by 3,748 patients over the first 182.5 days, while 35 NAION events and 510 censored patients account for 545; the letter states that 3,203 patients remain unexplained.
• The curves reportedly begin to separate at approximately one year, raising concern that a single hazard ratio may not represent risk consistently over time.
• The letter cites a positive predictive value of approximately 74.5% for administrative NAION coding.

Interpretation:

The author says unexplained risk-set attrition could reflect informative censoring or a data-cleaning artifact and could bias the reported hazard ratio of 2.33. The letter asks whether the authors assessed time-varying hazard ratios.

Limitations:

• The letter does not establish the cause of the apparent risk-set discrepancy.
• It does not report whether the proportional hazards assumption was tested or provide period-specific hazard ratios.
• The concerns are based on reported figures and cited prior investigations; no new analyses are presented.

Conclusion:

The author requests clarification of cohort accounting and the temporal pattern of risk to inform assessment of the reported NAION risk.

Sources:

• Article context supplied in the prompt