LETTER
We read the recent paper by Cheng et al. with interest and commend the authors for assessing the clinically important question of whether piperacillin-tazobactam can be used as definitive therapy for bacteremia caused by Enterobacteriaceae harboring inducible AmpC β-lactamases (1). However, we disagree with the main conclusions of the paper, namely, that definitive therapy with piperacillin-tazobactam is no different than definitive therapy with either cefepime or meropenem.
In the propensity score-matched cohort, both 30-day mortality and persistent bacteremia were numerically doubled in patients receiving piperacillin-tazobactam versus those receiving cefepime or meropenem, although the reported odds ratio (OR) for 30-day mortality is 0.5 (Table 3 in reference 1). Similarly, unadjusted 30-day mortality was lower for recipients of piperacillin-tazobactam in the overall cohort, but the reported odds ratio for 30-day mortality is 1.16. Could the control group or mortality have been misclassified in these analyses? Although the lack of statistical significance for the comparison does not change, the qualitative interpretation of the findings is fundamentally reversed.
An additional consideration lies in the use of propensity score-based modeling. Although propensity scores may increase precision and reduce bias when assessing relatively uncommon events, standard multivariable logistic regression models may outperform propensity score-based models when a sufficient number of events are observed (2). In this case, the more limited propensity score-matched cohort included only 82 of the original 165 patients and reduced the observed mortality events from 18 to 9. Under these conditions, the use of propensity score matching may have increased model bias and, through reduction of the number of patients, led to a loss of power and a possible type 2 error. Was a standard multivariable logistic regression model that adjusted for severity of illness considered?
The numerical increase in mortality and persistent bacteremia are consistent with the results of a recent study performed by Harris and colleagues addressing risk factors for recurrence in patients with monomicrobial Enterobacter species bacteremia (3). In this study, the univariate odds of recurrence were nearly doubled for those who received piperacillin-tazobactam or ticarcillin-clavulanate compared to those who received carbapenems (OR, 1.83; 95% confidence interval [CI], 0.64 to 5.21; P = 0.26). Although only 21 patients received definitive therapy with a β-lactam or β-lactamase inhibitor and the finding was not statistically significant, taken together, these two studies lead to significant concerns regarding the use of piperacillin-tazobactam as definitive therapy for bacteremia caused by Enterobacteriaceae with inducible AmpC β-lactamases. Further research is needed to define the role of piperacillin-tazobactam as a treatment for these challenging infections.
FOOTNOTES
For the author reply, see https://doi.org/10.1128/AAC.01163-17.
- Copyright © 2017 American Society for Microbiology.