Carvalho et al. aimed to address whether the routine use of intra-operative opioids remains necessary in contemporary peri-operative practice 1. Our primary concern is whether the study really compared the use of intra-operative opioids or instead compared broader combinations of anaesthetic regimens. The authors classified interventions as opioid-free, remifentanil-only or other opioid-inclusive strategies and further stratified them according to the use of local/regional anaesthesia. However, in the included randomised trials, these categories also encompassed different maintenance techniques, adjunctive infusion drugs and different local/regional analgesic approaches. Thus, many comparisons within the network did not isolate the opioid variable alone but rather compared two peri-operative pathways. The baseline table shows that inhalational anaesthesia accounted for 65% of the included studies, total intravenous anaesthesia for only 16% and mixed or unclear techniques for 19%. It is therefore difficult to conclude that avoiding intra-operative opioids may yield higher quality recovery. This concern relates directly to a fundamental assumption of network meta-analysis, which is that key effect modifiers should be distributed similarly across the different sources of evidence. This concern is further compounded by the authors' treatment of regional anaesthesia. Although the methods section refers to local/regional anaesthesia, the figures in the main text collapse these interventions into a single regional anaesthesia node. As a result, techniques with substantially different durations and analgesic intensities, such as local infiltration, neuraxial anaesthesia, peripheral nerve blocks, single-injection blocks and continuous catheter techniques, are treated as a single intervention. This is particularly problematic for time-dependent outcomes assessed at 2 h, 12 h, 24 h and 48 h because short-acting local anaesthetic infiltration and continuous regional anaesthesia are not interchangeable 2, 3. Under these circumstances, imbalances in key effect modifiers across the network may compromise the validity of indirect comparisons and thereby threaten the assumption of transitivity. Methodologically, such highly heterogeneous, multi-component interventions are more appropriately regarded as combination strategies. From a statistical perspective, although the authors state they extracted or derived per-protocol data, they did not report a corresponding intention-to-treat sensitivity analysis. Previous methodological work has shown that reliance solely on per-protocol analyses may introduce bias, because treatment adherence is often associated with prognostic factors 2. Although the main text describes the network as ‘overall consistent’, significant signals of local inconsistency remain evident in comparisons involving early postoperative opioid consumption, as well as the incidence of postoperative nausea and vomiting and pruritus. Several comparisons for early opioid consumption yielded multiple extremely low p values. These findings suggest that interpretation of the relevant outcomes still requires considerable caution. We believe that the conclusion supported most firmly by this study is that local/regional analgesic techniques themselves are associated with improved postoperative outcomes. The network may be more limited in its ability to disentangle the independent effects of individual components within complex anaesthetic pathways.
Liu et al. (Thu,) studied this question.