Findings of an observational study of neuraminidase inhibitors highly sensitive to decision to exclude 1652 treated patients

[Extract] To the Editor—Venkatesan and colleagues [1] reported that outpatient neuraminidase inhibitor (NAI) treatment of influenza A(H1N1)pdm09 reduced the odds of hospitalization (unadjusted odds ratio, 0.23; 95% confidence interval [CI], .19–.28). This association, however, was driven entirely by the authors’ decision to remove patients who received NAI treatment in the community on the day of their hospital admission from their analysis. There were 1652 such patients.

Although this seems reasonable (arguing that a drug taken for no more than 1 day would be insufficient to reduce risk of hospitalization), this introduces a large bias because the dropping of these 1652 patients could not be matched for the comparison group (which, in this observational cohort study, is defined by nonreceipt of the intervention, and so no specific date can be established for nonreceipt). Restoring those 1652 patients into the analysis completely alters the direction of the effect: The crude odds ratio of hospitalization changes to 1.97 for those treated with NAIs being hospitalized compared to those who were not (Table 1).