Procalcitonin‐guided or C‐reactive protein‐guided antibiotic discontinuation protocols for adult patients who are critically ill with sepsis: a reply

We appreciate the correspondence from Huang et al. [1] regarding our rapid systematic review and meta-analysis on the clinical effectiveness of procalcitonin (PCT) or C-reactive protein-guided antibiotic discontinuation protocols for adult patients who are critically ill with sepsis [2]. We would like to clarify several points to address their concerns, which might not have considered the limitations and judgements we detailed in the review. We agree with Huang et al. that the studies by Mahmutaj et al. and Kyriazopoulou et al. were not conducted in ICUs [3, 4]. In our review, our prespecified eligibility criteria noted that “Trials were eligible if they included patients aged ≥ 16 y with confirmed or highly suspected sepsis, requiring antibiotic and organ support therapy and receiving treatment in critical or intensive care settings. As the definition of sepsis can vary (both over time and geography), where the criteria of a study were ambiguous, we used a sequential organ failure assessment (SOFA) score of ≥ 2 and consulted with clinicians (PD and TF) to determine eligibility.” [2]. In the study by Kyriazopoulou et al., patients meeting Sepsis-3 criteria received advanced care in medical wards due to ICU bed shortages [4]. Similarly, patients in a surgical ward in the study by Mahmutaj et al. were deemed by experts to be clinically eligible for ICU care based on their baseline characteristics, as noted in the footnotes to our online Supporting Information Table S2 [2, 3]. We screened and decided to include both studies based on discussions with our team members, who are experienced critical care clinicians, and their inclusion aligned with our eligibility criteria and research question. This also reflects real-world practice, where critically ill patients may receive critical care outside ICUs due to bed shortages but still receive critical-level care. The inclusion of the study by Bloos et al. was in line with our criteria for including protocols for discontinuing antibiotics [5]. Our prespecified criteria did not include studies that used PCT or C-reactive protein to both initiate and discontinue antibiotics [2]. The protocol in Bloos et al. provided recommendations for discontinuing antibiotic therapy and for changing or optimising source control, as summarised in our online Supporting Information Table S3 [2, 5]. We considered the primary function of this protocol to be providing guidance on antibiotic discontinuation based on specified absolute and relative stopping thresholds. Therefore, the study was included as the protocol focused on discontinuation, not initiation, of antibiotics. Regarding the concerns from Huang et al. about the inclusion of lower severity studies that may amplify the antibiotic-reducing effect of PCT, we agree that PCT-guided discontinuation may have less impact on reducing antibiotic therapy duration in patients who are more severely ill (when judged clinically), as indicated by our meta-regression [2]. In the discussion, we interpreted this effect as perhaps reflecting differences in clinical decision-making in higher acuity patients. Furthermore, we noted caution when interpreting observational findings from meta-regression results. We would also not recommend post hoc exclusion of studies based on illness severity after analysis, as this would likely increase the risk of bias. It is also important to note that we used SOFA scores as an accepted proxy measure for illness severity; in practice, multiple clinical variables contribute to assessing illness severity, which was beyond the scope of the review. We appreciate Huang et al. highlighting the event rate for long-term mortality in online Supporting Information Figure S22 [2]. We have corrected the long-term mortality event rate, which has been amended in the published supporting figure and a Corrigendum will also be published to detail the amendments to the main article. Importantly, this correction very slightly changed the pooled risk ratio and does not affect the overall conclusions of the long-term mortality findings nor the review.