Intravitreal injection of anti-vascular endothelial growth factor (anti-VEGF) is currently the best treatment for neovascular age-related macular degeneration (nvAMD). Anti-VEGF use is associated with concerns about adverse events in the elderly. Some adverse events such as arterial thromboembolic (AT) events, though rare, are potentially fatal. Investigations of AT event risk are typically a secondary objective in clinical trials of anti-VEGF agents. Some studies show increased risk of AT adverse events, while other studies show no appreciable difference in risk.
The MarketScan® insurance claims database provided 90% power to detect differences as low as 5%, and encompassed a nationally representative population, to enable the investigation of rare AT adverse events. I analyzed insurance claims data for 153,019 newly-diagnosed nvAMD patients aged 50 years or older continuously enrolled between 2006-2012. Of the study participants 41,336 (27.0%) had AT events during the study period, and 76,014 (49.7%) of the study participants received anti-VEGF injections.
I used Cox regression models, with adjustments for selected covariates and propensity score (PS) weighting, to assess the risk factors for AT adverse events associated with anti-VEGF therapy. I also used logistic regression methods to investigate the odds of AT events associated with any anti-VEGF therapy, and used Cox proportional hazards to assess AT event risk for patients in treatment subgroups for each of the three currently used drugs, Bevacizumab (Avastin), Ranibizumab (Lucentis) and Aflibercept (Eylea).
I found a three-fold increase of the risk of new AT adverse events in patients with a history of AT events prior to their nvAMD diagnosis (Hazard ratio [HR] 3.28, 95% confidence interval [CI] 3.21, 3.38). Other covariates associated with higher AT adverse event risk were the medication Plavix (HR 1.59), peripheral arterial disease (HR 1.39), diabetes (HR 1.27), renal disease (HR 1.17), dementia (HR 1.08), Preferred Provider health insurance subscription (HR 1.06), and age (HR 1.17).
The marginal odds ratio I observed for AT events in association with any anti-VEGF therapy was 20% lower (Odds ratio 0.79, 95%CI 0.77, 0.81) than without therapy, in a logistic regression model adjusted for potential confounders. When I included patient time and propensity score weights in the logistic regression, the estimated odds was 1.18 (95% CI 2.2, 1.2), indicating almost 20% marginal increased odds of AT adverse event in association with anti-VEGF, and suggesting that the time element and confounding by indication, are important in consideration of anti-VEGF-associated AT adverse event risk.
Cox regression showed that AT events are up to 20% less likely to occur within the first 30 days, and 10% less likely in days 31-60 after nvAMD diagnosis in patients who received any anti-VEGF, compared to those who did not. AT events were more likely to occur when follow-up was longer, and my Kaplan-Meier curves portrayed a switch from negative to null risk after 90 days, with no further effect of anti-VEGF after approximately 90 days.
There were no marked differences in AT adverse event risk associated with the individual anti-VEGF medications. Avastin and Lucentis lowered AT event risk in the first 90 days after nvAMD diagnosis by about 10%, and there was no difference in AT risk over the same time period between those who did, or did not receive Eylea.
In conclusion, I identified that a history of AT events prior to the use of anti-VEGF, use of Plavix, peripheral arterial disease, diabetes, renal disease, dementia, Preferred Provider health insurance subscription, and age, were associated with a higher risk of AT events. My observation of change in direction of effect estimates from odds ratio 0.77 to odds ratio 1.18 before and after accounting for patient time on the study, and for confounding by clinical indication (using propensity scores), illustrated the importance of these factors in interpreting any assessment of AT adverse events from insurance claims data. The data in fact indicated a negative risk of AT events associated with anti-VEGF in the first 90 days that switched to a positive risk thereafter - possibly due to increased recognition and preventive treatment for patients on anti-VEGF therapy, compared to patients who were not on treatment.