Association of Metformin Use and New-Onset International Classification of Diseases Coding of Neovascular Age-Related...

To investigate whether metformin use is associated with decreased odds of new-onset International Classification of Diseases (ICD) coding of neovascular age-related macular degeneration (nAMD).

Case-control study.

A total of 22,205 cases with new-onset ICD coding of nAMD and 22,126 matched controls without age-related macular degeneration (AMD) were identified in the Merative MarketScan Research Databases between 2008 and 2017. A subgroup of patients diagnosed with diabetes included 6664 cases and 5513 controls.

Cases with new-onset ICD coding of nAMD were propensity score matched to controls without AMD. Multivariable conditional logistic regression analyzed the association between new-onset ICD coding of nAMD and metformin use controlling for: (1) AMD risk factors, including diabetes, diabetic retinopathy, hyperlipidemia, obesity, and smoking; (2) exposures to other antidiabetic medications, including insulin, sulfonylureas, glitazones, meglitinides, other diabetic medications, and statins; and (3) the number of antidiabetic medications.

Adjusted odds ratios (aORs) of new-onset ICD coding of nAMD for any metformin use and cumulative 2-year metformin dose in grams.

Any metformin use was associated with decreased adjusted odds of new-onset nAMD (aOR, 0.84; 95% confidence interval [CI], 0.74-0.95). In an exploratory dosing analysis, associations across cumulative-dose categories were heterogeneous, with the largest magnitude in the middose range (between 271 and 600 g; aOR, 0.73; 95% CI, 0.63-0.85); because adherence cannot be verified from claims, dose-response findings are interpreted as hypothesis-generating. These associations persisted among diabetic patients (any metformin use: aOR, 0.83; 95% CI, 0.72-0.94; 271 to 600 g: aOR, 0.72; 95% CI, 0.61-0.85; >1080 g: aOR, 0.85; 95% CI, 0.72-0.999). Any metformin use in diabetic patients without retinopathy was associated with decreased odds of new-onset ICD nAMD (aOR, 0.79; 95% CI, 0.69-0.92); however, this association was absent in diabetic patients with retinopathy.

In this claims-based analysis, metformin use was associated with lower odds of incident ICD-coded nAMD, and an exploratory dosing analysis suggests a low-to-moderate cumulative dose may be associated with the greatest decrease. These associations persisted among diabetic patients without retinopathy. Causal inference is limited by the observational design, potential diagnostic misclassification, imperfect exposure measurement, and residual confounding. Replication with validated case definitions, active-comparator new-user designs, and adjudicated outcomes is warranted.

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.