Neurocognitive outcomes: methodological clarifications

We thank Mistry and Nair 1 for their comments on our umbrella review of 114 systematic reviews and meta-analyses of interventions aimed at reducing the incidence of peri-operative neurocognitive complications 2. We appreciate their observations but we believe some of the assumptions underlying their comments warrant clarification. They note that when inconsistency reaches extreme levels, any overarching pooled estimate becomes difficult to interpret and may not reflect real-world practice. However, the primary aim of an umbrella review is not to pool intervention effects, but to summarise the entirety of the evidence produced by multiple systematic reviews and meta-analyses, and to evaluate their methodological quality 3. As in other umbrella reviews, our second-level meta-analysis was not intended to provide a definitive clinical effect size. Rather, it offers an additional metric reflecting the average of published meta-analytical estimates and serves mainly as an indicator of coherence among reviews and of the overall robustness of the evidence. Mistry and Nair emphasise correctly that umbrella reviews are most informative when they synthesise high-quality meta-analyses. We agree, but umbrella reviews are the instrument through which such methodological quality is assessed 4, 5. In fact, our central finding is that most meta-analyses on postoperative neurocognitive complications are of low or very low quality, a point that the authors highlight and that we agree with. The concern regarding the aggregation of conceptually distinct interventions is understandable. In our review, however, all analyses were separated, with only one notable exception (melatonin), where other sleep-related interventions were grouped. We now report a subgroup analysis separating these interventions that confirm that the heterogeneity estimates, effect sizes, confidence intervals and significance levels remained unchanged, supporting the robustness of this merge (Fig. 1). They note that the absence of stratification by surgical specialty and anaesthetic technique may limit clinical applicability. However, such granularity was not attainable within the evidence base we analysed. The meta-analyses included in our umbrella review did not provide study-level distinctions beyond the broad categories of cardiac, non-cardiac and orthopaedic surgery. Accordingly, these were the only subgroups that could be constructed reliably, and our stratification reflects the maximum level of detail permitted by the available data. We agree fully that evolving definitions of peri-operative neurocognitive disorders warrant explicit consideration which is why we described our terminology across the included studies carefully. Because our objective was to synthesise all published meta-analyses on this topic, it was not feasible to conduct definition-specific subgroup analyses, especially given that most meta-analyses do not report distinctions at the study level and refer generically to delirium or postoperative cognitive decline. We thank Mistry and Nair again for their constructive comments 1. Their letter allowed us to clarify aspects that may not have been evident to all readers, and we believe this exchange helps refine the interpretation of umbrella reviews.