We appreciate the interest and comments by Dr Eggenberger in regard to our recent article examining the incidence of optic neuritis in patients using anti–tumor necrosis factor therapies. His letter highlights many important aspects of our paper and the methodologies inherent to observational study designs. We agree with Dr Eggenberger’s thoughts regarding the importance of case ascertainment. As we pointed out in our manuscript, we were not able to review the medical records of the identified cases, and we accordingly relied upon a validated algorithm that consisted of multiple diagnostic codes consistent with acute optic neuritis. When performing our validation exercise in developing this algorithm, we indeed did use the case definition described within his letter as our gold standard (Optic Neuritis Treatment Trial definition ). To his point that some of the cases identified in our study might not be true cases (eg, the 85-year-old patient cited within his letter), we agree this is possible. Although this algorithm provided a positive predictive value of 100% in our validation study, we recognize that there exists a 95% confidence interval around this estimate such that the true positive predictive value could be lower. We also do not know if the true positive predictive value could be lower in other healthcare systems outside of where our validation exercise was performed. However, it is reassuring that if our algorithm was overly sensitive and lacked specificity (and thus identified false-positive cases), then the true incidence rates of optic neuritis would likely be even lower than those estimated by our study.
Dr Eggenberger’s letter also correctly reiterates some of the important limitations of observational studies, including the inability of such studies to fully prove or disprove a causal link between an exposure and an event. We do agree that evidence exists suggesting a possible link between anti-TNF therapy and optic neuritis (hence the rationale for our study), although we believe our study supports the idea that if such events do occur, at least in the patient population that we studied, they are quite rare.