Response to: Infection indication and severity

We thank Hsu and Lai for their interest in our study.¹ They suggest that confounding by infection indication and severity may have influenced the results.² We agree that these are important considerations and, to the extent possible, we controlled for these factors. First, recognizing that azithromycin, levofloxacin, and amoxicillin-based antibiotics are used to treat non-respiratory infections, we included other infection types (i.e., skin/soft tissue, genitourinary) in our propensity score models for confounding control. The authors are correct that we did not control for recent ear, eye, or sexually transmitted infections. However, we excluded otic and ophthalmic antibiotics. Second, to minimize confounding from illness severity, we studied oral, outpatient antibiotics and excluded patients who were hospitalized or in a skilled nursing facility during the 30 days before study antibiotic initiation. We also excluded antibiotic prescriptions with therapeutic durations typical of more severe infections (i.e., >5 days for azithromycin and >10 days for comparator antibiotics). Finally, our negative control outcome analyses produced null results, suggesting minimal bias.³

While we made concerted effort to minimize bias, residual confounding is an inherent limitation of observational studies, and, as we acknowledged, the possibility of residual confounding remains and should be considered when interpreting the results.¹ However, randomized controlled trials in the setting of rare events, like sudden cardiac death, are difficult, if not impossible, to conduct. As such, rigorous pharmacoepidemiology studies have an important role in investigating medication safety.⁴ Our study provides population-specific safety information that clinicians can consider when prescribing azithromycin to hemodialysis patients.