Documented preoperative statin exposure and early mortality after aortic valve replacement diverged across cohorts
A multicohort observational analysis has raised a cautionary signal around preoperative statin exposure before aortic valve replacement. Across 5,291 records drawn from SNUH, MIMIC-IV, and eICU, patients without documented statin exposure carried consistently higher adjusted short-term mortality estimates than those with documented exposure, although the primary comparison stopped short of conventional statistical significance.
The central question was straightforward: does documented statin use before valve surgery track with lower early death after AVR, or do the real-world data tell a more complicated story? Here, the answer was complicated. In the primary analysis, overlap-weighted 30-day mortality was 1.29% versus 2.55% in SNUH and 1.18% versus 2.36% in MIMIC-IV for those with versus without documented exposure. When the two cohorts were pooled with fixed effects, the hazard ratio reached 2.01, with a 95% confidence interval of 0.98 to 4.11 and a P value of 0.056.
That result sits just beyond the usual threshold for statistical significance, but the direction of effect was not ambiguous. Mortality estimates were higher when statin exposure was absent from the record, and the same pattern persisted through several sensitivity analyses. The study did not provide a clean causal answer; it provided a warning about how exposure documentation, patient selection, and residual confounding can reshape a clinical association.
Why a preoperative lipid drug might appear tied to postoperative survival
Statins are not simply cholesterol-lowering agents. In cardiovascular surgery, they are often discussed for pleiotropic effects on endothelial function, inflammatory tone, and plaque stability, all of which could influence perioperative risk. In a patient heading into AVR, a drug class that tempers vascular inflammation might plausibly reduce vulnerability to ischemic complications, hemodynamic instability, or downstream organ injury during the surgical stress response. The biologic argument is attractive. The epidemiology here, however, remains uneven.
The investigators tested whether the observed association held under different ways of defining exposure and follow-up. A 1-day MIMIC-IV exposure window reclassified seven records and, when pooled with the SNUH 24-hour definition, produced a 30-day hazard ratio of 2.29 (95% CI, 1.14 to 4.61; P = 0.020). A recorded 365-day analysis also pointed in the same direction, with a hazard ratio of 1.76 (95% CI, 1.00 to 3.09; P = 0.050). Thirty-day inverse probability of treatment weighting estimates and the SNUH available-follow-up analysis were significant as well, but both remained sensitivity analyses rather than the primary result.
Exposure misclassification may be doing much of the work
A one-day change in the exposure window altered classification enough to reassign seven records in MIMIC-IV. In a clinical dataset built on documentation rather than pharmacy reconciliation, even modest reclassification can shift the apparent treatment effect. That is especially relevant when the variable of interest is not drug initiation in the operating room, but whether a statin was actually recorded in the preoperative window.
The study design makes the distinction unavoidable. Documented exposure is not the same thing as biologic exposure. Some patients with true statin use may not have been captured in the record, while others may have had incomplete medication histories transferred across systems. When exposure is misclassified, the apparent benefit or harm can move in either direction. Here, the consistent signal favored better survival among those with documented exposure, but the confidence intervals and analytic sensitivity keep the interpretation guarded.
The numbers point in one direction, but the study stops short of causality
The primary analysis was inconclusive. That is the correct reading, even with the repeated trend toward lower mortality among patients with documented statin exposure. The pooled estimate missed statistical significance by a narrow margin, and the authors explicitly identified residual confounding, exposure misclassification, and incomplete death ascertainment as limitations on causal interpretation.
Those limitations are not academic footnotes. In an observational perioperative study, the decision to prescribe or continue a statin may itself mark a healthier, more engaged, or more intensively managed patient. Conversely, patients with the sickest perioperative profiles may be less likely to receive or continue oral medications before surgery. Either direction can distort the apparent relationship between medication exposure and postoperative mortality. The study design cannot fully separate a true pharmacologic effect from the clinical circumstances that determine whether the drug appears in the chart.
What this multicohort study adds to AVR risk discussions
Even without a definitive causal claim, the analysis adds a useful signal to preoperative cardiovascular care. It suggests that documented statin exposure deserves attention in the risk profile of patients undergoing AVR and that medication histories should be captured with more precision than a routine chart abstraction often allows. For a procedure where early postoperative mortality remains clinically meaningful, even small documentation gaps can change the interpretation of outcomes research.
The data also show how brittle observational estimates can be when follow-up ascertainment differs across systems. SNUH provided available-follow-up data, while MIMIC-IV and eICU contributed deidentified critical care records under data-use agreements. Those strengths give the study breadth, but they also mean the mortality endpoint is not anchored by a single uniform ascertainment process. The result is a layered evidence base rather than a settled answer.
The most defensible conclusion is narrow: absence of documented statin exposure tracked with higher adjusted early mortality estimates after aortic valve replacement, but the primary comparison did not establish a statistically secure association. Further work with cleaner medication reconciliation and more complete outcome capture will be needed before statin exposure can be treated as a reliable perioperative prognostic marker in this setting.