Does Hypotension Prediction Index limit the occurrence of postoperative complications associated with intraoperative hypotension? A systematic review and model-averaged Bayesian meta-analysis of statistically sound studies.

Intraoperative hypotension (IOH) is considered a potential contributing factor to postoperative complications. In 2018, a machine-learning algorithm to predict hypotension has been included in the Hemosphere™ platform (Edwards Lifescience, USA), of which the Hypotension Prediction Index (HPI) is the marker. We conducted a systematic review and Bayesian model-averaged meta-analysis to evaluate whether the limitation of IOH had a limiting effect on postoperative cardiac, pulmonary and renal complications and mortality. Nine randomized controlled trials and one propensity score analysis were included. Out of 978 patients, 488 received HPI monitoring, and 490 received routine monitoring. The Bayesian meta-analysis demonstrated that, in the HPI group, there is overwhelming evidence of a lower time-weighted average (TWA) for mean arterial pressure (MAP) < 65 mmHg and a shorter duration of hypotension. However, HPI-guided management did not demonstrate a convincing reduction in postoperative cardiac, pulmonary or renal complications, nor in mortality, and the certainty of evidence for these outcomes was low to very low. High heterogeneity and low certainty of evidence limit the strength of conclusions for postoperative outcomes. In this model-averaged Bayesian meta-analysis, HPI-guided management consistently reduced intraoperative hypotension, but did not translate into a demonstrable reduction in postoperative complications or mortality.