Time of day of cardiac surgery and postoperative outcomes: a reply

We thank Davies and Szentgyorgyi for their correspondence 1 regarding our recent analysis of time of day and outcomes following cardiac surgery 2. We agree that interpretation of observational analyses in this area requires caution, and we welcome the opportunity to clarify the intent, scope and limitations of our findings. Knife-to-skin time is the culmination of a complex peri-operative sequence influenced by organisational, anaesthetic and surgical factors. We recognise the complex organisational and clinical considerations that underpin daily list planning and prioritisation. We selected this exposure pragmatically because it is recorded reliably at scale within national datasets and represents a consistent point following which cardiopulmonary bypass is initiated. We acknowledge fully that this measure cannot disentangle the multiple upstream contributors to surgical timing, and we agree that variability in operating theatre processes, staffing and critical care capacity is likely to contribute to observed patterns. Therefore, our analyses were framed explicitly as exploratory rather than causal. We agree that case mix is not distributed randomly across the operating day. As reported, patients undergoing late morning surgery had higher predicted surgical risk, longer cross-clamp times and a greater proportion of complex or redo procedures. These factors support the authors' point that anticipated complexity and risk influence scheduling decisions. Adjustments for EuroSCORE 2 and cross-clamp duration were therefore pre-planned and undertaken to help address confounding, while recognising that residual confounding is almost inevitable in secondary analyses of observational data. We agree that surgeon-specific listing preferences, cancellations and on-call commitments are important influences that cannot be captured adequately in national audit data. The choice of time epochs also warrants discussion. Our groupings were defined prospectively to reflect typical UK cardiac surgery lists, in which one case is usually performed in the morning and one in the afternoon. We accept that a later knife-to-skin time in the late morning may reflect appropriate planning rather than delay, particularly for complex all-day cases. Our intention was not to imply inefficiency or suboptimal care, but rather to identify whether outcomes differed across routinely used start times within daytime operating hours. We would also emphasise that the observed association between late morning surgery and cardiovascular mortality is modest and lies at the threshold of statistical significance. We highlighted that, after adjustment, no corresponding differences were observed in major adverse cardiovascular events, myocardial infarction, heart failure readmission or all-cause mortality. We agree that multiple outcomes were examined and that the possibility of a false-positive finding cannot be excluded. For this reason, our conclusions were cautious and framed as hypothesis-generating. We did not intend our findings to supersede earlier high-quality studies reporting neutral effects of daytime surgical timing, but rather to complement them by highlighting a specific signal that merits further investigation. Davies and Szentgyorgyi raise important statistical considerations, including potential collider bias and alternative modelling strategies. These are valid points, and future analyses could explore approaches such as negative-control outcomes, centre-level adjustment or first vs. second case comparisons. However, such analyses require additional assumptions and data granularity that are not available uniformly within linked national datasets. We recognise that the reciprocal relationship between anticipated postoperative risk and timing of surgery is central to interpretation. Indeed, one of the key messages of our study is that patients at higher risk tend to undergo surgery later in the morning. Whether this reflects optimal clinical judgement, organisational constraints or interactions with circadian or peri-operative physiological factors cannot be determined from retrospective data alone. Our findings do not justify changes to clinical scheduling at present but highlight an area where prospective, mechanistic and system-level research is needed to better understand how organisational and biological factors intersect to influence outcomes following cardiac surgery.