Reply to Schuetz and Wahl

To The Editor— We appreciate Drs. Schuetz’s and Wahl’s concerns regarding residual confounding as a possible explanation for the association between procalcitonin orders and increased antibiotic days and Clostridium difficile infection. In order to attenuate residual confounding by indication, we had performed 2 sensitivity analyses using complementary methods at low risk for confounding by indication: (1) an analysis using hospital-level procalcitonin rates as the “exposure” of interest [1, 2] and (2) a hospital difference-in-difference analysis [3]. In both analyses, we did not identify an association between procalcitonin (PCT) orders and patient outcomes. Although it is possible that increased risks associated with PCT testing observed in patient-level analysis may be partly explained by residual confounding by indication, the more important finding is that we did not observe lower antibiotic utilization in any analysis, including analyses using methods designed to attenuate confounding by indication. Thus, our findings did not support successful translation of PCT testing from clinical trials into real world clinical settings.

With respect to the comparison between propensity score-based analyses and multivariable-adjusted models, studies have shown little difference in effect estimates when comparing these 2 approaches directly [4].

With respect to the recently published study by Balk et al. investigating outcomes associated with PCT use during suspected sepsis, our analysis and results differed in several ways. Our study focused on the translation of results from randomized clinical trials to clinical practice. Clinical trial evidence supports procalcitonin-guided therapy as a practice that reduces antibiotic duration in septic patients in the intensive care unit (ICU) but has not yet shown that procalcitonin can be safely used in decisions to initiate antibiotics for critically ill patients with suspected sepsis [5, 6]. As such, current guidelines [7] do not recommend use of PCT levels to assist in decisions regarding initiation of antibiotics. Therefore, our study evaluated antibiotic duration among ICU patients with a diagnosis of sepsis who received at least 1 dose of antibiotic, with an outcome of antibiotic duration. In contrast, Balk et al. evaluated the association of PCT orders with “total number of antibiotics given” (rather than duration) for patients with a sepsis diagnosis admitted to an ICU [8], including patients who never received an antibiotic. It is worrisome that the PCT strategy evaluated by Balk et al. was associated with slightly higher hospital mortality rates in 2 of 3 analyses. We would submit that the identified absolute difference of 0.7 (95% confidence interval 0.4–0.9) antibiotic-days is unlikely a clinically significant difference and we would argue that the findings of Balk et al., similar to our findings, show that use of PCT testing outside of a clinical trial setting does not currently approach the 2–3 day reduction in antibiotic duration that has been observed in clinical trials [5, 9, 10]. Taken together, studies of procalcitonin use during sepsis suggest that more work needs to be done to effectively implement procalcitonin-guided therapy in real-world settings.

Notes

Financial support. The work was supported by the National Heart, Lung and Blood Institute at the National Institutes of Health [grant K01HL116768 to A. J. W.].

Potential conflicts of interest. Both authors: No reported conflicts of interest. Both authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References