Study cohort
In a national, registry-based prospective cohort study using multiple validated national registers, we identified all persons in Denmark who were alive and had T2D by December 1, 2018, or developed T2D no later than December 31, 2023. For identification, we combined data from the Danish National Patient Registry, including all diagnostic and treatment codes for in- and outpatient hospital care [9], and the Danish National Prescription Registry, containing information regarding redeemed prescriptions according to the Anatomical Therapeutical Chemicals (ATC) classification system [10]. T2D was defined by combining International Classification of Disease (ICD) version 10 codes [11] for T2D (E11*) and ACT-codes for insulin (A10A*) and non-insulin glucose-lowering medicine (A10B*) in accordance with definitions presented by Thykjaer et al. [12].
We included all persons with T2D above the age of 18 years free of NAION at the time of entry and excluded all persons that had previously used other forms of semaglutide (i.e. Rybelsus® and Wegovy®) than once-weekly semaglutide.
Exposures and outcome
In all persons with T2D, exposure was defined as redemption of at least one prescription of once-weekly semaglutide (Ozempic®, ATC: A10BJ06) as coded in the Danish National Prescription Registry between December 1, 2018, and December 31, 2023. The remaining persons with T2D were regarded as non-exposed. Index date was set as the day of the first redeemed prescription (exposed group) and December 1, 2018 (unexposed group). As we used time-varying exposure, patients exposed to once-weekly semaglutide participated as non-exposed prior to first redemption of semaglutide prescription.
The outcome was a diagnostic code of NAION (H470C) in the Danish National Patient Registry as registered between December 1, 2018, and June 30, 2024. In addition, we evaluated the overall number of persons with NAION (irrespective of T2D) between 1 January 2003 and 30 July 2024 to evaluate any overall trends over time.
Covariates
Covariates were evaluated at the time of entry into the study.
Information regarding age, sex, and marital status was obtained from the Danish Civil Registration System, which was also used to link data from all registers by a unique personal identifier given to all inhabitants in Denmark at birth or immigration [13]. Duration of diabetes was calculated as the time between the first diagnostic code or redeemed prescription indicating T2D and the date of entry in the study. We used the Register of Laboratory Results for Research [14] to obtain measurements of hemoglobin A1c, plasma creatinine, albumin/creatinine ratio in urine, and estimated glomerular filtration rate (Supplementary Table 1). We used the registration in closest proximity to study inclusion within an allowed range of one year. Use of cholesterol (C10*) and blood pressure lowering medicine (C03*, C07*, C08*, and C09*) was determined in the Danish National Prescription Registry. Cardiovascular disease was defined in the Danish National Patient Registry as the first day of any of the diagnostic coding of Supplementary Table 2 [15]. Finally, we used the Danish Registry of Diabetic Retinopathy [16] to assess information of most recent level of diabetic retinopathy (as given by worse eye) according to the International Classification of Diabetic Retinopathy Disease Severity Scale [17], which is a five-step scale ranging from no diabetic retinopathy, through mild, moderate and severe nonproliferative diabetic retinopathy to end-stage proliferative diabetic retinopathy.
Sensitivity analyses
To verify the robustness of the data, we performed three sensitivity analyses:
First, in order to account for potential unmeasured confounding by including an active comparator, we evaluated the risk of incident NAION in persons treated with sodium-glucose transport protein 2 (SGLT2) inhibitors (ATC: A10BK01 [n = 84,283], A10BK02 [n = 3048], A10BK03 [n = 85,892], A10BK04 [n = 96]) versus once-weekly semaglutide with or without redemption of prescriptions for SGLT2 inhibitors. In this analysis, patients were included from the redemption of the first prescription of a SGLT2 inhibitor or once-weekly semaglutide, whichever came first, until a month after the last prescript redemption or at the end of the study on June 30, 2024.
Second, we evaluated the risk of NAION among once-weekly semaglutide users in a model excluding persons with existing diabetic retinopathy to eliminate potential detection bias that may rise if patients who have been evaluated for diabetic retinopathy may be more prone to seek ophthalmic care.
Third, we excluded users of other glucagon-like peptide-1 receptor agonists to account for any potential class-effect of the drug.
Statistical analyses
We present data as counts (with proportions) or medians (with interquartile ranges [IQR]). Differences between persons exposed and non-exposed in Table 1 were tested by the k-sample test for equality of medians (continuous data) and chi-square tests (categorical data).
In Table 2, we evaluated the number of first-time NAION events, person-years at risk, incidence rates, and we performed a crude and a multivariable Cox proportional hazard regression adjusted for sex, age, marital status, duration of diabetes, hemoglobin A1c, estimated glomerular filtration rate, history of cardiovascular disease, use of insulin, use of cholesterol lowering medicine, and use of blood pressure lowering medicine.
All persons were followed from the time of inclusion until the day of the first diagnostic coding of NAION, death, emigration or end of follow-up, whichever came first.
We used Stata 18.0 (StataCorp, College Station, Texas) for statistical analysis, and statistical significance was considered as p-values lower than 0·05 and 95% CIs that did not include 1.
Leave a Reply