Concerns about “Complicated Skin and Skin‐Structure Infections and Catheter‐Related Bloodstream Infections: Noninferiority of Linezolid in a Phase 3 Study”

To the Editor—We have several concerns regarding the recent article by Wilcox et al. [1]. The authors state that “[t]he frequency and severity of adverse events were similar between groups” [1, p. 210]. This statement, however, is not consistent with their data. For 48 (13%) of 363 patients in the linezolid arm and for 25 (7%) of 363 patients in the control arm, treatment was discontinued secondary to adverse events. Although the authors do not present a statistical comparison, this difference is statistically significant (P =.006 by use of Fisher’s exact test). Thus, we cannot conclude that the frequency and severity of adverse events were similar between groups. In their table 5, the authors present data showing that the number of drug-related adverse events leading to discontinuation of the study was similar between the 2 arms (6 adverse events in the linezolid arm vs. 4 adverse events in the control arm). However, it is unclear what is meant by drug-related versus non-drug-related adverse events leading to discontinuation of the study. It would have been helpful if the authors explained the dCfference; presumably 42 individuals in the linezolid arm and 21 individuals in the control arm discontinued treatment as a result of having nondrug-related adverse events. With respect to the increased mortality seen in patients without gram-positive bacteremia, we disagree with the authors’ statement that “the post-hoc nature of these analyses and the size of these subsets of the primary analysis populations limit certainty” [1, p. 210]. The authors do not present a statistical analysis of the interaction between treatment group and the presence or absence of gram-positive bacteremia. On the basis of the frequencies presented in the article, we calculated that the presence of gram-positive bacteremia was a significant modifier of the association of the odds of mortality with the treatment arm (P = .058 by use of the Mantel-Haesnzel test for homogeneity, with χ2 = 3.6). Given that there is no reason to expect that linezolid therapy would be efficacious for patients without gram-positive bacteremia, we feel this interaction is compelling. The effect of linezolid therapy on mortality must be viewed in terms of the type of bacteremia present. When the patients without gram-positive bacteremia who received linezolid were evaluated (using the frequencies presented in Wilcox et al. [1]), we calculated an odds ratio of 2.3 (95% confidence interval, 1.3–4.4) for mortality during the follow-up period. Thus, patients who received linezolid without a gram-positive bacteremia had increased mortality during the course of the study. In summary, on the basis of these data, we feel that linezolid should not be given as empirical therapy to patients with catheter-related bloodstream infections before the isolation and identification of a pathogenic organism. It would be better to start with vancomycin. Conversely, once a gram-positive organism has been isolated and identified, the data presented by Wilcox et al. [1] suggest that, depending on the clinical circumstances, linezolid could be used.