This randomized controlled study was conducted from January 2020 to September 2021 at KK Women’s and Children’s Hospital, Singapore, after approval by the SingHealth Centralized Institutional Review Board (Ref: 2018/2213) on 06/04/2018 and registration on clinicaltrials.gov (NCT03620942) on 08/08/2018. Written informed consent was obtained from all participants in accordance with the Declaration of Helsinki, and this manuscript adhered to the relevant Consolidated Standards of Reporting Trials (CONSORT) guidelines.

We included parturients aged 21 to 45 years old, weighing 40 to 100 kg, 145 to 170 cm in height, American Society of Anesthesiologists (ASA) physical status 2, with singleton full term pregnancies undergoing spinal anesthesia for elective cesarean delivery. Parturients with contraindication to spinal anesthesia, hypertensive disorders requiring medication, premature rupture of amniotic membranes for > 48 h, diabetes requiring insulin, and uncontrolled medical conditions such as cardiac disease were excluded.

Parturients were randomized (1:1 ratio) to receive ADIVA or DIVA using a computer-generated random number generator. The allocation sequence was created by the study statistician and concealed using sequentially numbered opaque sealed envelopes. Prior to spinal anesthesia, the study investigator opened the envelope containing the group allocation. Parturients, obstetricians, nurses, and anesthesiologists involved in anesthesia management and data collection were blinded to the group allocation.

Baseline SBP was measured in the ward as the mean of three consecutive readings taken at one-minute intervals using an oscillometric device on the right arm with the patient supine and left uterine displacement. Intravenous access was obtained using an 18-gauge cannula, pulse oximeter and electrocardiogram were applied, and the Nexfin finger cuff was attached to the right second or middle finger. Spinal anesthesia was performed in the sitting flexed position with a 27-gauge Whitacre needle (BD Medical, New Jersey, USA), 11 mg hyperbaric bupivacaine 0.5%, 15 μg fentanyl, and 100 μg morphine. The parturient was then positioned supine with left uterine displacement, free flowing infusion of lactated ringers solution was commenced, and the ADIVA system was initiated by the study investigator.

SBP and heart rate measurements by Nexfin were uploaded continuously to a laptop computer which integrated the data every 10 s to form a moving average value (LabVIEW running on Windows XP, Microsoft Corporation, Washington, USA). The drug delivery system consisted of two syringe driver pumps (B.Braun, Melsungen, Germany) with one 50 ml syringe filled with 100 μg.ml−1 phenylephrine and another 50 ml syringe containing 8 mg.ml−1 ephedrine, both connected to the intravenous cannula by three-way stopcocks. Parturients randomized to the ADIVA system (Fig. 1A) received 25 μg phenylephrine (heart rate > 60 beats.min−1) or 2 mg ephedrine (heart rate < 60 beats.min−1) if SBP fell between 90 to 110% of baseline, 50 μg phenylephrine or 4 mg ephedrine if SBP was between 80 to 90% of baseline, and 75 μg phenylephrine or 6 mg ephedrine if SBP dropped below 80% of baseline. ADIVA calculated the moving average SBP every ten seconds, and the gradient of the slope of SBP changes was calculated using the least squares method. Vasopressors were administered rapidly (over 4.5 to 13.5 s depending on vasopressor dose) if SBP trended downward in the ten-second interval preceding vasopressor delivery, while the same vasopressor dose was delivered over 30 s if SBP trended upward [8]. Conversely, parturients randomized to the DIVA system received 25 μg phenylephrine (heart rate > 60 beats.min−1) or 2 mg ephedrine (heart rate < 60 beats.min−1) if SBP fell between 90 to 100% of baseline, and 50 μg phenylephrine or 4 mg ephedrine if SBP dropped below 90% of baseline (Fig. 1B). A lockout period of 10 s occurred after each vasopressor bolus in both ADIVA and DIVA systems. Both systems were monitored by an investigator who could manually administer atropine and/or vasopressors in the event of bradycardia or if SBP remained < 70% of baseline for more than three minutes.

Schematic diagram of the algorithm used in (A) ADIVA, and (B) DIVA. ADIVA: Advanced double intravenous vasopressor automated; DIVA: double intravenous vasopressor automated

Five minutes after spinal anesthesia, sensory block was assessed using loss of sensation to cold while motor block was measured using the modified Bromage scale [11]. Patient and clinical characteristics, SBP, heart rate, duration of spinal anesthesia to fetal delivery, vasopressor doses, incidence of nausea/vomiting, umbilical cord pH, and Apgar scores were recorded. We evaluated ADIVA and DIVA using performance measures previously employed on similar closed-loop systems, adjusted with a pooled-data approach that provides consideration for the wide variation in number of measurements taken for each parturient, described below [12,13,14].

Percentage performance error was defined as the percentage difference between each SBP value from the baseline, and is calculated for the ith parturient at the jth second as follows:

MDAPE indicates the absolute magnitude of differences between measured and baseline SBP, and is a measure of inaccuracy. It is defined as the median of absolute PE (|PE|) values, calculated as follows where Ni is the number of values of |PE| for the ith parturient and M is the number of parturients in the study:

$$MDAPE=\frac1^MNi}X\sum_^M(Ni\;X\;MDAPEi)$$

MDPE is a measure of bias, and indicates whether SBP was systematically above or below the baseline, calculated as follows:

$$MDPE=\frac1^MNi}\;X\;\sum_^M(NiXMDPEi)$$

Wobble measures how much PE fluctuates around the MDPE with time for each parturient, calculated as follows:

$$WOBBLE=\frac1^MNi}\;X\overset M}(Ni\;X\;WOBBLEi)$$

Divergence is the slope obtained from linear regression of each parturient’s |PE| with time, and assesses the trend of |PE| change over time, thereby indicating if the system accuracy is improving (negative divergence) or worsening (positive divergence) with time. Divergence (per minute) was calculated as follows where tij is the time of ith individual measurement in minutes:

$$DIVERGENCEi=\frac^\left|PEij\right|\;X\;tij-(\sum_^PEij)X\left(\sum_^tij\right)/Ni}^\left(tij\right)}^2=^tij)}^2/Ni}$$

$$DIVERGENCE=\frac1^MNi}\;X\;\sum_^M(Ni\;X\;DIVERGENCEi)$$

Continuous and categorical variables were summarized as mean ± standard deviation (SD) or median (interquartile range (IQR) [range]) as appropriate, or number (proportion) respectively. Categorical and continuous variables were compared using the X2 test, two-sample t-test, or Mann–Whitney U-test as appropriate. All analyses were performed using SAS version 9.4 (SAS Institute, North Carolina, USA).

A sample size of 92 parturients (46 in each group) is required to detect 25.5% absolute difference in the incidence of hypotension, based on the following assumptions: incidence of hypotension in ADIVA group of 13.5%, incidence of hypotension in DIVA group of 39.0% based on previous DIVA studies on reported incidence and difference in incidence [10, 14], 1:1 allocation ratio, alpha of 5%, power of 80%, and 5% loss due to failure of spinal anesthesia.