Response to "Comment on The effect of perioperative ketamine and esketamine administration on postoperative nausea and vomiting in patients undergoing general anesthesia: a systematic review and meta-analysis"

Dear Editor,

We appreciate the thoughtful comments regarding our publication, “The effect of perioperative ketamine and esketamine administration on postoperative nausea and vomiting in patients undergoing general anesthesia: a systematic review and meta-analysis” [1]. Important methodological and clinical concerns were raised, and we welcome the opportunity to clarify our approach and interpretations.

First, we would like to clarify that the “pooled PONV” endpoint does not imply that postoperative nausea (PON) and postoperative vomiting (POV) are interchangeable. In PONV research, composite PONV generally indicates the occurrence of nausea, vomiting, or both. Therefore, when studies reported PON and POV separately, summing both outcomes could have led to double-counting patients who experienced both symptoms. Because patient-level overlap between PON and POV was generally unavailable, we selected PON as a pragmatic, non-duplicative representative component. This operational decision was intended to maximize the available PONV-related data while avoiding duplicate counting rather than to imply clinical equivalence between PON and POV. POV was analyzed separately to preserve its distinct objectivity and clinical significance. This operational rule was not prespecified in the PROSPERO protocol but was applied during data extraction after heterogeneous outcome reporting and unavailable overlap data were identified. As emphasized by Apfel et al. [2], nausea is a subjective conscious perception, whereas retching and vomiting are brainstem-mediated reflex responses. Thus, vomiting is not simply a more severe form of nausea, and these outcomes should be assessed independently.

With respect to outcome timing, the included studies reported PONV-related outcomes over heterogeneous periods, ranging from post-anesthesia care unit or other early postoperative assessments to 24 h, 48 h, or 3 days post-operation [3]. We selected the earliest available postoperative time point to capture the immediate pharmacological effects of intravenous ketamine or esketamine and reduce confounders from subsequent antiemetic administration, opioid exposure, or other postoperative management. This approach is consistent with that of a recent perioperative pharmacological meta-analysis of PONV-related outcomes [4]. However, when both early and late assessments are available, early- or short-acting effects may be reflected preferentially and delayed PONV may be underrepresented. Formal early/late subgroup analyses were unreliable because the assessment windows were sparsely and inconsistently distributed, and cumulative versus interval-specific measurements were often unclear. Thus, the pooled estimates should be interpreted as the earliest available PONV-related outcomes and not as standardized estimates of early, delayed, or the total 24 h/post-discharge PONV burden.

Second, we agree that the comparator-based subgroup analyses require cautious interpretation. The placebo subgroup included both true placebo-controlled trials and add-on designs, in which ketamine or esketamine was added to an otherwise comparable background regimen. To reduce confounding factors, we excluded comparisons in which an additional cointervention was administered only to the ketamine/esketamine arm. The non-opioid comparator group included patients receiving pharmacologically diverse agents [3]. Therefore, this subgroup should not be interpreted as evidence that ketamine or esketamine is inferior or superior to any individual non-opioid agent but rather as a contextual comparison with heterogeneous opioid-sparing or non-opioid perioperative strategies, some of which may have analgesic, anesthetic-sparing, opioid-sparing, or direct antiemetic effects. Thus, these subgroup findings are hypothesis generating and should not be overinterpreted as definitive comparative effectiveness estimates. Additional analyses restricted to placebo-controlled comparisons showed no significant dose-response relationship for either ketamine or esketamine, consistent with the primary findings [3]. Sex- or Apfel-score-based subgroup analyses could not be performed reliably because risk–score distributions and subgroup-specific outcome data were not consistently reported.

Third, we recognize the importance of appropriately handling multi-arm trials to avoid unit-of-analysis errors and the disproportionate weighting of shared comparator groups. In accordance with standard meta-analytical guidelines, we proportionally allocated shared comparator groups whenever feasible [5]. In a limited number of very small or sparse data comparisons, duplicate comparisons were retained to preserve available outcome data, which may raise residual unit-of-analysis concerns, particularly for small subgroup analyses. These cases were limited in number and transparently reported, and sensitivity analyses showed that they did not materially affect the overall conclusions.

Fourth, we agree that non-randomized studies may introduce additional confounders and reduce internal validity, particularly for PONV-related outcomes, which are influenced by anesthetic techniques, opioid exposure, antiemetic prophylaxis, and other perioperative factors. In our meta-analysis, only three of the 55 included studies were non-randomized. However, their contribution to the pooled estimates was limited: one non-randomized study was included in each of the PONV, PON, ketamine, and esketamine analyses, and two were included in the pooled PONV and POV analyses. Randomized studies were assessed using the Risk of Bias 2 tool and non-randomized studies were assessed using the Risk of Bias in Non-Randomized Studies of Interventions (ROBINS-I) tool. We also performed sensitivity analyses excluding nonrandomized studies for outcomes that included two or more such studies, and the pooled effect estimates and heterogeneity were not materially changed [3]. Thus, although the inclusion of nonrandomized studies is a limitation, it did not meaningfully affect the overall conclusions of our analysis.

Finally, we share the view that adverse effects are central to the clinical interpretation of ketamine and esketamine for PONV-related outcomes. In our analysis, hallucinations were significantly increased with both ketamine and esketamine. Drowsiness increased significantly in the overall analysis, whereas dizziness increased significantly only in the esketamine subgroup [3]. These findings were incorporated into our interpretation, which emphasized that any potential benefit of ketamine or esketamine in opioid-sparing or low opioid regimens should be weighed against neuropsychiatric, vestibular, and sedative risks. Accordingly, our conclusion does not support the routine use of ketamine or esketamine solely for PONV prevention but highlight the need for cautious, context-specific use and more standardized adverse-effect reporting in future studies.

Overall, these comments highlight key methodological issues in perioperative evidence synthesis. Our study should be interpreted as a synthesis of heterogeneous perioperative evidence rather than as support for routine ketamine or esketamine use in PONV prevention. Any potential benefit in opioid-sparing contexts should be balanced against the neuropsychiatric, vestibular, and sedative risks. Future studies using standardized PONV definitions, consistent assessment intervals, clearly defined comparator groups, appropriate multi-arm handling, and systematic adverse effect reporting are needed.