Differential Effect of Glaucoma Medications on Dry-Eye Incidence
Narrative Responses:
Purpose
Use of glaucoma medications is associated with dry eye syndrome (DES), however the relative incidence of DES with various glaucoma medications is poorly characterized. We sought to test the hypothesis that glaucoma medications differentially affect DES incidence, and estimate the risks of different medications on incident DES.
Methods
A retrospective case-control study was designed utilizing 13 consecutive years of coding data from the Veterans Affairs (VA) national patient database. A 2.5% random sample of patients seen in VA eye clinics during this period was analyzed to assess the risk of DES attributable to glaucoma medications. Life table analysis was used to calculate the background incidence of DES. Time-dependent Cox proportional hazards regression was performed to estimate the risk of exposure to glaucoma medications on DES incidence while accounting for important covariates. The main outcome measure was a risk ratio (RR) of incident DES adjusted for age and gender.
Results
55,534 patients were included in the analysis of which 49,958 (90%) were never exposed to glaucoma medications. Among these, the cumulative proportions with DES were 3.8% at 12 months; 6.3% at 24 months; 9.6% at 36 months; 16.4% at 5 years; and 44.9% at 10 years. Exposure to one or more glaucoma medicines significantly increased the risk of developing DES (RR 1.24, CI 1.2-1.3, p<0.001). Individual glaucoma medications conferring significantly increased risk of developing DES (p<0.008) included: topical carbonic anhydrase inhibitors (RR=1.21), timolol/dorzolamide (RR=1.23), travoprost (RR=1.24), brimonidine (RR=1.31); travoprost Z (RR=1.47); and systemic carbonic anhydrase inhibitors (RR=1.56).
Conclusion
Exposure to one or more glaucoma medication as well as 6 out of 8 individual medications demonstrated a significant effect on the incidence of DES, with RRs ranging from 1.21-1.56. Because glaucoma medications were not randomly assigned, we cannot exclude the possibility that mediating and confounding variables may, in part, account for the results.