Abstract
We thank Berenholz and colleagues for their interest in our study. We also appreciate the authors’ previous efforts to assess the neomycin ototoxicity in the presence of nonintact tympanic membranes (NITM).
The authors begin with a statement that no ototoxic effect of neomycin in NITM has been noted in clinical practice. The literature is replete with examples of uncommon safety issues not discovered in clinical practice and only appreciated through large population-based studies.
The authors state concerns about confounding or bias. As no concern about confounding is specified, we assume these terms were used interchangeably. Following are our thoughts about the three concerns about bias the authors specify.
Indeed the use of physician-assigned ICD9 coding to measure sensorineural hearing loss has not been validated, but several factors point to the validity of this approach. First, we did require charges for audiometric testing to assure appropriate diagnostic work up. Second, compromised sensitivity or specificity of measurement would have biased this study toward the null hypothesis (no ototoxic effect) and thus against our findings. Third, increased vigilance in neomycin users should have introduced bias in the analysis of first and not only after repeated doses. On the same note, we apologize for the lack of clarity in our analytic design. While we used all fluoroquinolone users as comparator, we adjusted the analysis for the number of repeat prescriptions, resulting essentially in a comparison of 1 dose neomycin : 1 dose fluoroquinolone, 2 doses : 2 doses, and so forth. A sensitivity analysis where we restricted the analysis to repeat users showed near identical results but wider confidence intervals. An analysis of all neomycin users versus all fluoroquinolone users as the authors suggest would lack the ability to find a cumulative effect and thus add no value to our research question. Note that repeat prescriptions was a statistically significant determinant of hearing loss, supporting the suggested association (as well as the importance for multivariate adjustment, as done in our analysis).
Finally, the authors suggest a secular effect, such that newborn hearing screening might have reduced the incidence of sensorineural hearing loss post-tympanic membrane perforation in later study years, which in turn were more often attributed to fluoroquinolone users. As shown in Table 2 we did adjust for study year, which in fact was not associated with risk for hearing loss.
We conclude that the observed results, even though pointing to only a small increase in risk, along with solid biological plausibility and the availability of alternative treatment options support the existing labeling to avoid the use of neomycin in patients with NITM.
Disclosures
Footnotes
Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article.