Comparison of a mean arterial pressure alarm to the hypotension prediction index in preventing intraoperative hypotension in elective moderate- to high-risk non-cardiac surgical patients: a study protocol for a blinded, parallel, randomized controlled trial with a non-inferiority framework

Background: Intraoperative hypotension is common in surgical patients and associated with an increased risk for acute kidney injury, myocardial injury, and 30-day mortality. To predict hypotension and thereby promote proactive blood pressure management, a machine learning algorithm was developed: the Hypotension Prediction Index (HPI). In clinical trials, it proved to prevent hypotension. However, a recent observational study found that the performance of the HPI in predicting hypotension up to 15 min in advance was virtually identical to that of the concurrent mean arterial pressure (MAP). This indicates that the added value of the HPI above continuous MAP monitoring may be limited. It is hypothesized that the effect of preventing intraoperative hypotension triggered by a MAP threshold of 70–75 mmHg is comparable to clinical care triggered by an HPI alarm. Methods: This is a monocentral, blinded, randomized controlled clinical trial, conducted in the Netherlands. Adult patients undergoing moderate- or high-risk elective non-cardiac surgery requiring continuous blood pressure monitoring aiming for a MAP of ≥65 mmHg are eligible for inclusion. A total of 142 participants is randomized into two groups: in the HPI group, the monitor alarms when the HPI value exceeds 85, whereas in the MAP group, it alarms when the MAP value drops below 72 mmHg. The monitor screen is partly covered to blind treating clinicians for the study group allocation. The primary outcome measure is the area under the threshold of hypotension (defined as MAP < 65 mmHg for at least 1 min) during surgery. Secondary objectives are the incidence and severity of intraoperative hypotension and hypertension, cumulative doses of administered medication and fluids, incidence of acute kidney injury and myocardial injury, length of hospital stay, and 30-day mortality after surgery. Discussion: The decision to initiate treatment upon an alarm is left at the discretion of the attending clinicians; if deemed appropriate, they can also choose to ignore the alarms. Consequently, clinical interventions could vary between participants and groups. Contrarily, early treatment could lead to overtreatment. Trial registration: Overview of Medical research in the Netherlands (OMON), NL-OMON53215, registered on 20-09-2023, last updated on 30-01-2025, https://onderzoekmetmensen.nl/nl/trial/53215.