This single-center randomized, controlled trial was conducted at a tertiary referral hospital from July 22, 2021 to January 26, 2022.

Inclusion and exclusion criteria

We recruited adult patients between 20 and 80 years of age, who were planning to receive elective non-cardiac surgery with general anesthesia, which required continuous invasive blood pressure monitoring via an arterial catheter. The target MAP for all patients was at least 65 mmHg. We excluded patients 1) who had undergone emergency procedures; 2) with known clinically important cardiac disease, such as moderate to severe valvular disease; 3) with a need for a tidal volume < 8 mL/kg of ideal body weight during surgery; and 4) with current persistent atrial fibrillation (Maheshwari et al.2020). After initially evaluating eligible patients during their preoperative visit and obtaining their written informed consent, we enrolled patients who were undergoing general anesthesia that was expected to last for > 4 h.

Randomization, allocation and blinding

Before surgery, the eligible participants were randomized into two groups: one undergoing hemodynamic management with HPI guidance (intervention group) and one being treating according to the standard of care (control group) (Fig. 1A). Computer-generated permutated block randomization (concealed and with varying permutated block sizes of four patients) was applied at a 1:1 allocation ratio, using SAS 9.4 (SAS Institute, Cary, NC, USA). Group allocations were kept in sequentially numbered, opaque, sealed envelopes and were only disclosed to a research assistant not directly involved in the clinical management of the participants or the data collection. To eliminate potential bias, we adopted a single-blinded study design and ensured that the participants were unaware of their group assignment.

Fig. 1

Diagnostic guidance and treatment protocol. A Treatment algorithm of intraoperative hypotension management. HPI: Hypotension Prediction Index; MAP: mean arterial pressure; Intervention: HPI guidance; control: no HPI guidance; SVR: systemic vascular resistance; Eadyn: dynamic arterial elastance; dP/dtmax: systolic slope; B Calculation of relevant duration in the intervention and control groups. AT: alarm time; FTT: time of the first intraoperative hypotension treatment; SAT: silent alarm time; HT: hypotension time

Anesthetic management

The patients underwent standard anesthetic monitoring, which included electrocardiography, pulse oximetry, upper-arm cuff oscillometry, and Bispectral Index (BIS) monitoring, after entering the operating room. General anesthesia was induced with propofol (1.5 -2 mg/kg) and fentanyl (2 μg/kg). Rocuronium was administered at a dosage of 0.7–1.0 mg/kg to assist with tracheal intubation, with additional 10 mg bolus doses administered as needed to sustain muscle relaxation throughout the surgery based on the train of four monitoring. Desflurane was used for maintenance, to achieve a target BIS value between 40 and 60. Mechanical ventilation was administered using a tidal volume of 8 ml per kilogram of predicted body weight, along with a positive end-expiratory pressure of 5 cmH2O, and an inspired oxygen fraction adjusted to maintain oxygen saturation at or above 96%. No epidural anesthesia or other regional anesthesia has been used. All patients have been intubated, with each patient equipped with a central venous catheter and an arterial catheter featuring an Acumen IQ sensor (Edwards Lifesciences). All patients were transferred to the intensive care unit (ICU) postoperatively. Extubation was performed base on the same criteria (Popat et al.2012) either before leaving the operation room or in the ICU.

Hypotension prediction index guidance (Intervention) group

All patients were fitted with arterial catheters in the radial artery and connected to the Acumen IQ sensor with the HPI early warning system software (Edwards Lifesciences). Arterial pressure waveform was measured continuously with a sampling frequency of 100 Hz. The HPI early warning system software is installed in a HemoSphere monitor (Edwards Lifesciences), which displays hemodynamic parameters calculated from the arterial waveform every 20 intervals. These values were updated every 20 s. The HPI early warning system generates a visible and audible alarm when the HPI value exceeds 85 (Hatib et al.2018), indicating an 85% likelihood that a hypotensive event will occur within the next 15 min (Davies et al.2020). The system can also detect the physiological mechanisms that lead to hypotension, including stroke volume variation, dynamic arterial elastance, and systemic vascular resistance (Wijnberge et al.2020).

The Acumen IQ sensor pressure transducer was connected to the HemoSphere monitor. The electrical signal generated was also transmitted to the standard monitor in the operating room (Phillips Healthcare, Best, Netherlands), which displayed MAP, systole, diastole, and pulse pressure variation. After placing the arterial catheter, the anesthesiologists visually inspected the arterial waveform signal to detect damping. All anesthesiologists and nurses were informed of the study protocol. A research assistant recorded the medical records related to the surgery and anesthesia procedures. The anesthesiologist and nurse were instructed to take actions within 2 min of an HPI alarm to prevent the occurrence of hypotension. The HemoSphere monitor shows several parameters (including stroke volume variation, dynamic arterial elastance, and systemic vascular resistance) that aided the anesthesiologists in making a differential diagnosis of the underlying cause of the predicted hypotension. The anesthesiologists treated intraoperative hypotension by following one of six treatment options: fluid, vasopressor, inotrope, fluid plus vasopressor, fluid plus inotrope, or observation. The underlying hemodynamic diagnostic guidance and treatment protocol (Fig. 1A) was designed based on a previous published protocol (Maheshwari et al.2020).

Standard-of-care (control) group

In the control group, all patients had arterial catheters inserted in the radial artery and were connected to the Acumen IQ sensor with the HPI early warning system software. The alarm had been activated, but it was muted (silenced), and the relevant visual indicators and screen information were obscured with a cloth covering. All patients were informed that intraoperative hypotension treatment would be initiated at MAP < 65 mmHg and that they would receive the same hemodynamic management to treat the underlying cause of hypotension as the intervention group (Fig. 1A).

Time-weighted average MAP < 65 mmHg, and other blood pressure related outcomes

The acquisition system of the HPI early warning system software allowed the raw data (HPI and MAP measurements) to be exported to a spreadsheet (Microsoft Excel; Redmond, WA, USA). The data were analyzed in MATLAB 2019b (MathWorks, Natick, MA, USA) (L. Frassanito et al. 2022).

The primary outcome was defined as time weighted average (TWA) MAP < 65 mmHg (Additional File 1, Supplemental Fig. 1). The area under the curve (AUC) of MAP < 65 mmHg was calculated as the depth of hypotension (in mmHg) below a mean arterial pressure of 65 mmHg multiplied by the time in minutes spent below a MAP of 65 mm Hg (mmHg × min). The value of TWA MAP < 65 mmHg was calculated from the value of AUC MAP < 65 mmHg divided by total surgical duration. The formula is as follows:

TWA MAP < 65 mmHg (mmHg) = AUC MAP < 65 mmHg (mm Hg × min) ÷ total surgical duration (min).

We presented an example from one of our patients in Additional File 1, Supplementary Fig. 1–1. The pink area represented the AUC MAP < 65 mmHg, at 852.8 mmHg × min, divided by a total surgical duration of 538 min, yielding a result of 1.59 mmHg. The total duration of MAP < 65 mmHg was 128 min. Other blood pressure related outcomes related to AUC MAP and TWA MAP below the other thresholds (60 mmHg and 55 mmHg) were similarly calculated. The average MAP during surgery and the basal hemodynamic parameters (the first systolic and diastolic blood pressures, MAP, and heart rate in the operation room) were also recorded.

To assess the risk of overtreatment in the intervention group, we examined the occurrence of severe hypertension (MAP > 130 mmHg), reflected as a TWA-MAP above the threshold of 130 mmHg throughout the monitoring period (Schneck et al.2020).

Serious adverse postoperative clinical outcomes

The occurrence of serious adverse postoperative clinical outcomes was recorded based on the Postoperative Morbidity Survey within 30 postoperative days (Maheshwari et al.2020). Details of the postoperative complications, based on the Postoperative Morbidity Survey definition, are documented in Supplementary Table 1 of Additional File 1 (Grocott et al.2007; Maheshwari et al.2020). Death within 30 days of surgery was also recorded.

Intraoperative outcome recording and clinicians’ actions

We assessed the actions of anesthesiologists in the management of both groups. We recorded (1) treatment options (e.g., vasopressors, inotropic agents, and fluid challenge), (2) cumulative doses of intraoperative medication, (3) the time from the first HPI alarm to the start of hypotension treatment in the intervention group and from the onset of MAP < 65 mmHg to the start of hypotension treatment in the control group (Fig. 1B), and (4) the number of intraoperative hypotension treatments per patient. For (3), in the intervention group, following the previous published method (Maheshwari et al.2020), if multiple HPI alarms occurred within 15 min of the initial one, we regarded them as a single event. We would apply the intraoperative hypotension management protocol in our study, using the hemodynamic parameters recorded during the first HPI alarm. If multiple HPI alarms occur within 15 min after the first HPI alarm, we calculated the time difference between the first HPI alarm and the first treatment for hypotension. In the control group, if there were repeated occurrences of MAP < 65 mmHg (hypotension episodes) within 15 min after the initial occurrence, we considered them as one event. We applied the hypotension management protocol, using the hemodynamic parameters recorded during the first hypotension episode. If multiple episodes of hypotension occur within 15 min after the first hypotension episode, we calculated the time difference between the first hypotension episode and the first treatment for hypotension. The time point of the silent alarm was determined post hoc by retrospectively examining the time when MAP < 65 mmHg and calculating the first occurrence of the HPI alarm within fifteen minutes. The time difference between the silence alarm and the treatment of hypotension was obtained. (Fig. 1B) (Wijnberge et al.2020).

Sample size calculation

Following the methodology of a prior study (Wijnberge et al.2020), the sample size was determined based on the primary outcome. The primary outcome values for the intervention and control groups were 0.15 and 0.40, respectively, with a standard deviation of 0.32. By setting α and 1-β values at 0.05 and 0.8, the estimated required sample size was calculated as 54 using G Power 3.0. Accounting for a 10% dropout rate, a total of 60 patients were recruited to participate in the study.

Statistical analysis

For descriptive statistics, we presented continuous data as mean and standard deviation. For highly skewed data, we used median and interquartile range (IQR) instead. We presented categorical data using a contingency table and proportions. For statistical inference, we used the independent T-test and the Mann–Whitney U test to compare the means and medians, respectively. We calculated confidence intervals using the large-sample method for the differences between means and the Hodges-Lehmann method for the differences between medians. We used the chi-squared and Fisher's exact tests to compare the proportions between the groups. A P value < 0.05 was considered statistically significant. All analyses followed the intention-to-treat principle and were performed using MATLAB version R2019b (MathWorks) and SPSS version 25 (IBM, Armonk, NY).