Effect of intraoperative dexmedetomidine on prognosis in patients with cancer undergoing surgical procedures. Responses to recent commentary

Editor

We appreciate the insightful comments provided by Xu and colleagues,¹ Tian and Jin,² and Wang and colleagues³ regarding our meta-analysis,⁴ which highlighted several important methodological and interpretative issues. Our analysis of 12 studies revealed inconsistencies between study types: RCTs suggest a potential benefit in recurrence-free survival, whereas retrospective studies have reported a possible reduction in overall survival.

Xu and colleagues¹ and Tian and Jin² correctly pointed out that pooling survival outcomes using odds ratios (ORs) can be suboptimal for time-to-event data, as hazard ratios (HRs) better account for censoring and the temporal nature of survival analysis. We fully acknowledge the advantages of HRs and appreciate this important methodological observation. However, in this study, ORs were selected for a number of reasons. Firstly, event rates in several included studies (e.g. Cho 2021, Luo 2023, Zhou 2017) were relatively low (<10%), under which conditions ORs can reasonably approximate HRs. Secondly, only ∼50% of the eligible studies reported extractable HRs, while ORs or equivalent risk estimates were consistently available across all studies. To maximise the inclusion of relevant literature and avoid bias attributable to data exclusion, we opted to use ORs in this analysis. Furthermore, methodological literature^5,6 supports the acceptability of ORs as surrogate measures in such situations, particularly when HRs are not reported. We agree that future updates of this meta-analysis should prioritise HR extraction as reporting standards evolve and more time-to-event data become available.

The issue of dosing heterogeneity across studies is also an important one. According to current clinical guidelines and practice, intraoperative dexmedetomidine is typically administered as a loading dose of 0.5–1.0 μg kg⁻¹, followed by a maintenance infusion of 0.1–0.5 μg kg⁻¹ h⁻¹. Most RCTs included in our analysis used doses within this range. Although some retrospective studies reported less standardised dosing (e.g. median doses without weight adjustment), the overall regimens were consistent with routine clinical practice. While such pharmacological variability might have influenced the pooled estimates, there is currently no consensus on how different dexmedetomidine doses affect tumour biology in humans. Limited preclinical evidence suggests that dexmedetomidine can exert both protective and deleterious effects on tumour progression in a dose- and context-dependent manner.7, 8, 9 However, these findings are mostly based on animal models, and the clinical relevance remains uncertain. Therefore, it is premature to assume that lower or higher doses yield distinct oncological outcomes. This limitation was addressed in our discussion, where we also emphasised the need for future studies using standardised dosing protocols to better assess the impact of dexmedetomidine on cancer-related outcomes.

We have already discussed the discrepancy between findings from RCTs and retrospective studies in our discussion section, where we analysed potential contributing factors such as study design, selection bias, and confounding. This inconsistency further underscores the need for additional high-quality, prospective RCTs to clarify the role of dexmedetomidine in cancer outcomes.

Xu and colleagues¹ and Wang and colleagues³ also noted that pooling data from different cancer types without stratification could be inappropriate. We acknowledged this limitation in our paper, recognising that combining diverse tumour types introduces clinical heterogeneity that might compromise generalisability of our findings. Although subgroup analyses by tumour type would be ideal, the current number of available studies is insufficient to allow for such stratification. Future research should aim to investigate tumour-specific effects of dexmedetomidine in larger and more homogeneous cohorts.

Wang and colleagues³ further suggested that we account for other perioperative medications, as this might help control for potential confounding and reduce bias. We agree that this is an important consideration. However, because of the limited number of studies and the lack of detailed reporting on concomitant medications, we were unable to perform relevant subgroup or sensitivity analyses. We hope that future studies will provide more comprehensive perioperative data to allow for such analyses.

In response to Tian and Jin’s² comment regarding the inclusion of Chinese-language literature, we emphasise that these studies represent an important and valid source of data and have been widely included in many high-quality meta-analyses. For example, studies by Chen and colleagues¹⁰ and Zhang and colleagues¹¹ incorporated Chinese-language publications to achieve more comprehensive and regionally relevant syntheses. Excluding such studies outright would introduce language bias and reduce the representativeness of the evidence. Therefore, we believe that including Chinese-language studies is both reasonable and consistent with established meta-analytic practices.

We also appreciate Tian and Jin’s² suggestion regarding the use of the Hartung–Knapp–Sidik–Jonkman (HKSJ) method. Several methodological reviews suggest that HKSJ provides more conservative and robust estimates, especially when the number of included studies is small. Nonetheless, it is not the only statistically acceptable approach. For instance, some meta-analyses with a limited number of studies have used the Mantel–Haenszel method under a fixed-effect model (e.g. Hsiao and colleagues¹²), while others have applied a random-effects model without HKSJ correction (e.g. Sánchez-de-la-Torre and colleagues¹³). We thank the reviewers for this thoughtful recommendation and will consider applying the HKSJ or other alternative approaches in future work to further enhance statistical rigour. The limitation of small sample size was also explicitly discussed in the paper.

We thank Xu and colleagues,¹ Tian and Jin,² and Wang and colleagues³ for their thoughtful and constructive comments. Their insights provide valuable perspectives on our work and help highlight important methodological and clinical considerations. We hope our responses have addressed the concerns raised, and we welcome continued discussion to further advance progress in this area.

Declaration of interest

JT is a member of the associate editorial board of the British Journal of Anaesthesia. The other authors declare that they have no conflicts of interest.

Handling Editor: Hugh C Hemmings Jr