In Reply We appreciate the letters regarding our article by Dorney et al1 in which messenger RNA (mRNA) COVID-19 vaccinations were not associated with retinal vascular occlusions (RVO). To summarize, the concerns with our study are in relation to Li and colleagues’2 comparable article, our choice of a 21-day time frame and the association between first and second vaccination, evaluation of confounder variables, and control group selection.

Li et al used the same database, TriNetX, that our study used but reported an increased association with vaccinations and RVO development. Li and colleagues’ conclusion is based on all types of COVID-19 vaccinations, not specifically mRNA vaccinations. However, they did report the risk of RVO, specifically with mRNA vaccinations, and found no association between vaccination and RVO development at 12 weeks, which agreed with our findings.2

For our study,1 a 21-day time frame was used to evaluate RVO development for several reasons. First, this allowed us to evaluate both the first and second vaccination as 21 days, which was the shortest time frame before receiving a second dose. Second, to our knowledge, the largest case series evaluating RVO cases postvaccination reported that RVO presented, on average, 14 days after vaccination with a maximum time frame of 42 days.3 Although cases have been reported after this time frame, this would have limited our ability to evaluate both vaccines separately. However, Li et al2 analyzed mRNA vaccinations until 12 weeks and found no association with RVO, further emphasizing that extending the observation time point for RVO development did not increase the risk of RVO. Further extending the timeline would also increase the risk that patients contract COVID-19 infection, which could itself increase the risk of RVO,4 confounding results.

Our study1 specifically chose diabetes, hypertension, and hyperlipidemia to match on due to the known association they has with RVO. Multiple confounders, including cardiovascular diseases and cerebrovascular accidents, were not controlled for in this analysis and would be an appropriate next step for further investigating this relationship. In addition, our study did not exclude patients with a prior history of a positive COVID-19 infection result, which may have affected the results. The US Centers for Disease Control and Prevention reported 43.9% of individuals in the US were estimated to have a COVID-19 infection.5 Effectively controlling for COVID-19 infection was not feasible for several reasons; TriNetX only captures information at specific affiliated institutions, missing patients with a COVID-19–positive result who took tests at unaffiliated institutions or at-home tests and asymptomatic patients without a test.

Lastly, several comments addressed choice of control group. We chose a historical control of vaccinated patients to ensure that neither COVID-19 infection nor vaccine confounded the results. This also reduced confounding from fundamental differences between unvaccinated and vaccinated patients who differ in access to care and health behaviors.6 In summary, our results agree with Li and colleagues’ study,2 used a time point to compare first and second vaccine dose, which is in line with all known reports of postvaccine RVO, and controlled for some but not all potential confounders.