Different methods of modelling intraoperative hypotension and their association with postoperative complications in patients undergoing non-cardiac surgery

Background: Associations between intraoperative hypotension (IOH) and postoperative complications have been reported. We examined whether using different methods to model IOH affected the association with postoperative myocardial injury (POMI) and acute kidney injury (AKI). Methods: This two-centre cohort study included 10 432 patients aged >= 50 yr undergoing non-cardiac surgery. Twelve different methods to statistically model IOH [representing presence, depth, duration, and area under the threshold (AUT)] were applied to examine the association with POMI and AKI using logistic regression analysis. To define IOH, eight predefined thresholds were chosen. Results: The incidences of POMI and AKI were 14.9% and 14.8%, respectively. Different methods to model IOH yielded effect estimates differing in size and statistical significance. Methods with the highest odds were absolute maximum decrease in blood pressure (BP) and mean episode AUT, odds ratio (OR) 1.43 [99% confidence interval (CI): 1.15-1.77] and OR 1.69 (99% CI: 0.99-2.88), respectively, for the absolute mean arterial pressure 50 mm Hg threshold. After standardisation, the highest standardised ORs were obtained for depth-related methods, OR 1.12 (99% CI: 1.05-1.20) for absolute and relative maximum decrease in BP. No single method always yielded the highest effect estimate in every setting. However, methods with the highest effect estimates remained consistent across different BP types, thresholds, outcomes, and centres. Conclusions: In studies on IOH, both the threshold to define hypotension and the method chosen to model IOH affects the association of IOH with outcome. This makes different studies on IOH less comparable and hampers clinical application of reported results.