AcumenTM hypotension prediction index guidance for prevention and treatment of hypotension in noncardiac surgery
AcumenTM hypotension prediction index guidance for prevention and treatment of hypotension in noncardiac surgery
AcumenTM hypotension prediction index guidance for prevention and treatment of hypotension in noncardiac surgery
Abstract
Background Intraoperative hypotension is common during noncardiac surgery and is associated with postoperative
myocardial infarction, acute kidney injury, stroke, and severe infection. The Hypotension Prediction Index software
is an algorithm based on arterial waveform analysis that alerts clinicians of the patient’s likelihood of experiencing
a future hypotensive event, defined as mean arterial pressure < 65 mmHg for at least 1 min.
Methods Two analyses included (1) a prospective, single-arm trial, with continuous blood pressure measurements
from study monitors, compared to a historical comparison cohort. (2) A post hoc analysis of a subset of trial partici-
pants versus a propensity score-weighted contemporaneous comparison group, using external data from the Multi-
center Perioperative Outcomes Group (MPOG). The trial included 485 subjects in 11 sites; 406 were in the final effec-
tiveness analysis. The post hoc analysis included 457 trial participants and 15,796 comparison patients. Patients were
eligible if aged 18 years or older, American Society of Anesthesiologists (ASA) physical status 3 or 4, and scheduled
for moderate- to high-risk noncardiac surgery expected to last at least 3 h. Measurements: minutes of mean arterial
pressure (MAP) below 65 mmHg and area under MAP < 65 mmHg.
Results Analysis 1: Trial subjects (n = 406) experienced a mean of 9 ± 13 min of MAP below 65 mmHg, compared
with the MPOG historical control mean of 25 ± 41 min, a 65% reduction (p < 0.001). Subjects with at least one episode
of hypotension (n = 293) had a mean of 12 ± 14 min of MAP below 65 mmHg compared with the MPOG historical
control mean of 28 ± 43 min, a 58% reduction (p< 0.001). Analysis 2: In the post hoc inverse probability treatment
weighting model, patients in the trial demonstrated a 35% reduction in minutes of hypotension compared to a con-
temporaneous comparison group [exponentiated coefficient: − 0.35 (95%CI − 0.43, − 0.27); p < 0.001].
Conclusions The use of prediction software for blood pressure management was associated with a clinically mean-
ingful reduction in the duration of intraoperative hypotension. Further studies must investigate whether predictive
algorithms to prevent hypotension can reduce adverse outcomes.
*Correspondence:
Xiaodong Bao
[email protected]
Full list of author information is available at the end of the article
© The Author(s) 2024. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or
other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this
licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecom-
mons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Bao et al. Perioperative Medicine (2024) 13:13 Page 2 of 12
device or multiple vasoactive agents; sepsis; planned ven- the monitor to alert anesthesia providers. The care team
tilation with tidal volume below 8 ml/kg of ideal body could administer fluid and/or vasoactive agents using
weight; as well as those scheduled for burn surgery, neu- advanced hemodynamic data or could choose to ignore
rosurgical procedures, open-chest procedures or urgent/ alerts based on their clinical assessment after coaching
emergent surgery. from the research team about the software.
Trial protocol
Subjects were enrolled after signing informed consent Trial endpoints
and having an arterial catheter connected to FloTrac IQ The primary endpoint for the trial was a cumulative dura-
sensor and EV1000 platform containing the predictive tion of intraoperative hypotension. The secondary end-
software. The software was activated after confirming point was the area under MAP of 65 mmHg. Data for the
a good-quality arterial waveform signal using a square trial endpoints were downloaded from the EV1000 moni-
wave test. The predictive index, ranging from 0 to 100, tor, which records vitals and the predictive index every 20
was displayed on the monitor, indicating the likelihood of s during the monitoring time. An episode of intraopera-
patients having a hypotensive event. A secondary screen tive hypotension was defined as three or more consecu-
with quantitative hemodynamic parameters including tive 20-s observations of MAP < 65 mmHg. Duration of
cardiac output, stroke volume variation, change of pres- hypotension was calculated as the sum of episodes where
sure over the change of time, systemic vascular resist- MAP < 65 mmHg for at least 1 min during the monitor-
ance, and dynamic arterial elastance aided clinician ing period. The area under MAP of 65 mmHg was cal-
assessment of physiological conditions (Fig. 1). When culated using the trapezoidal rule to estimate the area of
the hypotension prediction index exceeded 85 for two pressure and time. The total area was obtained using the
consecutive 20-s updates, a popup alert appeared on formula below.
k l
pij + p(i−1)j sites from May 2019 to October 2020 and who did not
Total AUC = tij − t(i−1)j ∗ 65 −
2 participate in the trial (“Comparison”). Data were not
1 1
available from MPOG for the 11th trial site for that time
where tij is the measurement time of the ith hypoten- period. To maintain the inclusion and exclusion crite-
sion increment of the jth hypotension episode for the ria from the trial, similar eligibility criteria were applied
patient and p
ij is the mean pressure in mmHg for the ith using definitions that could be applied to retrospective
hypotension increment of the jth hypotension episode. data, as specified in Appendix 2.
The episode for each patient begins, t0j, with the first This post hoc analysis was conducted in collaboration
of two successive pressure measurements below MAP with the MPOG consortium. The MPOG Site Primary
65 mmHg and continues until the MAP raises to 65 Investigators at the HPI participating sites approved
mmHg or above. The trapezoidal rule sums the average the use of the dataset for this project, and the analytic
decreases in pressure from 65 mmHg between two meas- plan was presented at the MPOG Perioperative Clinical
urement times and multiplies that by the difference of the Research Committee. As has been previously described,
time increment between. Then, the areas per episode are the MPOG consortium (see www.mpog.org) maintains a
summed across the total number of episodes. detailed clinical and administrative data repository from
participating hospitals across the United States. MPOG
Statistical analysis data include automated extraction of both device-cap-
Single‑arm trial with historical comparison group tured and manually entered Electronic Health Record
A statistical analysis plan for the single-arm trial was (EHR) data, including patient and procedural character-
written, date-stamped, and recorded in the investigators’ istics, anesthetic medications, physiologic parameters,
files before data were accessed (Appendix 1). The analy- and key surgical events for patients undergoing anesthe-
sis excluded subjects in the pilot cohort. Mean duration sia care at contributing institutions. Monthly site-specific
of hypotension was calculated with a weighted average case validation for a random sample of submitted data
of site means and standard deviations as described in by subject-matter experts is required of all contribut-
Appendix 1. The standard deviation of the duration of ing sites, and additional quality checks are conducted
hypotension is the square root of the pooled variance at the coordinating center to monitor each center’s data
with each study site’s hypotension variance. The trial uploads and to remove artifacts from machine-captured
participants were compared to a historical comparison variables.
cohort identified in registry data using t-tests. The Multi- Data elements from MPOG used in this analysis
center Perioperative Outcomes Group (MPOG) provided included patient demographics (age, sex, BMI, ASA sta-
summary statistics on 22,109 adult patients with ASA 3 tus), clinical characteristics (Elixhauser comorbidities),
and 4 physical status, undergoing surgeries ≥ 180 min, procedural characteristics (procedure codes, blood pres-
with arterial line monitoring, treated between January sure observations, vasopressor use, estimated blood loss),
1, 2017 and December 31, 2017, at the same 11 hospitals and patient outcomes (AKI).
that participated in the HPI effectiveness study (Shah The main outcome of the post hoc analysis was the
et al., 2020). This information was used as a historical duration of hypotension in minutes (defined as MAP
control to define the retrospective amount of IOH, which < 65 mmHg for at least 1 min). The secondary outcome
was compared to that found in this prospective HPI mul- was AKI (using the Kidney Disease–Improving Global
ticenter study. All trial analyses were completed with SAS Outcomes definition of Stage 1 or greater, as an increase
version 9.4. of serum creatinine more than 0.3 mg/dl above baseline
within 48 h of anesthesia end time or more than 50% ele-
Post‑hoc analysis with propensity score‑weighted vation within 7 postoperative days) (Disease, 2012).
comparison group A statistical analysis plan for the post hoc analy-
To supplement the trial analysis that was designed a sis, specifying the outcomes and methods, was drafted
priori, we conducted a post hoc analysis that compared after the completion of the trial but before the post hoc
the duration of hypotension and the secondary outcome analyses began (Appendix 2). Descriptive statistics for
(AKI) among a subset of patients in the trial treatment continuous data were reported as mean ± standard devia-
group with a propensity-score weighted contemporane- tion or median (interquartile range) depending on the
ous control group, using data reported from MPOG for distribution of the data. Categorical data is presented
hospitals that participated in the trial. Trial subjects, as frequency counts and proportions. Weighted and
including both pilot and non-pilot subjects, from 10 of unweighted standardized mean differences were calcu-
the 11 sites (“Trial”) were compared with a contempora- lated and reported to compare trial participants to the
neous cohort of patients who had surgery in the same 10 comparison group.
Bao et al. Perioperative Medicine (2024) 13:13 Page 5 of 12
The post hoc analysis evaluated the association Austria) and RStudio (RStudio PBC, Boston, MA, USA),
between the prediction software and the incidence of with two-sided p values < 0.05 considered statistically
intraoperative hypotension, as defined above. To evaluate significant.
the difference in hypotension duration conditional on the
use of the prediction software, a generalized linear model Trial sample size estimate
was conducted regressing the duration of hypotension on Recent analyses demonstrated a mean duration of hypo-
the fixed effects for software presence or absence. Given tension of 29.27 min with a standard deviation 43.44
the skewed distribution for the duration of hypotension, (Shah et al., 2020). We anticipated that the use of the
a log link was specified to accommodate the distribution predictive algorithm would reduce that duration by
under study. Because it was anticipated that patients with 25%, based on based on an advisory panel expert opin-
longer surgical cases may have an increased period at ion and review of recent literature (van Waes et al., 2016;
risk of developing hypotension, the model was adjusted Maheshwari et al., 2018; Stapelfeldt et al., 2017). The ratio
a priori for the duration of intraoperative blood pressure of the standard deviation to the mean in the previous
measurement (i.e., time at risk). data was 1.44; however, the ratio from other publications
It was also anticipated that patients may have a differ- on hypotension and data gathered by the sponsor varied
ent probability of participating in the trial, therefore the between 0.88 to 1.76. To protect against the uncertainty
analysis employed the use of propensity score analyses. of the ratio underpowering the study, additional compu-
Specifically, individuals who elected to participate in tations were done to make the estimate more conserva-
the single-arm trial might possess characteristics that tive with a standard deviation to mean ratio of 1.65. The
differentiate them from eligible individuals who either minimum recommended required sample size, using Pass
declined to participate or were not offered participation. 14, is 380 completed subjects for 90% power for a one-
To address this selection mechanism, a logistic regression sided alpha = 0.025 test. Assuming 10% attrition for a less
model was first developed that predicts trial participation than 3-h surgery and a 5% loss to follow-up, the recruited
(i.e., yes vs no) conditional on demographic and disease sample size was estimated as a minimum of 380/0.85 ≈
characteristics. The predicted probability of participa- 448 for 90% power in the non-pilot cohort. Therefore, the
tion was then used as an inverse probability of treatment study sample size was capped at 485, including up to 33
weight (IPTW) in a second and final model (the primary pilot subjects and up to 452 non-pilot subjects, for a min-
analysis) that examines each outcome conditional on imum of 90% power.
prediction software use (i.e., yes or no) and duration of
blood pressure measurement. Variables in this IPTW Results
included age, sex, race, ethnicity, ASA physical status, Single‑arm trial with historical comparison group
Vanwalraven Elixhauser Comorbidity Index, procedure A total of 778 patients were screened for trial eligibility
timing (afternoon or morning), and procedural service. (Supplementary Figure S1). Among them, 293 failed to
Several versions of the inverse-probability of treatment meet inclusion criteria. Four hundred eighty-five subjects
weights were considered but non-truncated and non- have consented to the study, 425 of whom ultimately had
stabilized weights were chosen based on the distribution a surgery length of 3 h or longer (19 in the pilot cohort
of estimated propensity scores. Results are presented as and 406 in the non-pilot cohort). The analysis focused on
exponentiated coefficients that yield a percent difference the 406 subjects in the non-pilot cohort and 293 of those
(i.e., a ratio of geometric means) in the duration of hypo- who had at least 1 min of MAP below 65 mmHg. Table 1
tension between the trial and comparison groups and shows the primary and secondary endpoints from the
their associated 95% confidence intervals (CI). single-arm trial. During the study period, participants (n
The secondary outcome was evaluated using a simi- = 406) experienced a mean of 9 min (SD 13) of MAP <
lar approach but with a generalized linear model that 65 mmHg, as reported via the EV 1000 monitors, and a
included a binomial distribution and logit link function. mean area under MAP < 65 mmHg of 47 mmHg × min-
Results of the secondary outcome are thus presented as utes (SD 85). A historical comparison group from the
an odds ratio and 95% confidence interval (CI). A sensi- same set of hospitals (n = 22,109) experienced a mean of
tivity analysis was performed for both the primary and 25 min (SD 41) of MAP < 65 mmHg, based on data from
secondary outcomes using multiple imputation with the MPOG registry, representing 65% fewer minutes of
chained equations (m = 40 imputations) which were hypotension than trial participants (p < 0.0001). Subjects
derived from the preoperative clinical characteristics with at least one episode of hypotension (n = 293) had
used in the propensity model. a mean of 12 ± 14 min of MAP below 65 mmHg com-
All post hoc analyses were performed using R version pared with the MPOG historical control mean of 28 ± 43
4.1.2 (R Foundation for Statistical Computing, Vienna, min, a 58% reduction (p < 0.001). Trial sites reported a
Bao et al. Perioperative Medicine (2024) 13:13 Page 6 of 12
total of 21 postoperative safety events, including 17 inci- associated with a 35% reduction in the duration of hypo-
dents of AKI, 3 instances of myocardial injury, and 1 non- tension. There was no statistically significant difference in
fatal cardiac arrest. No strokes or in-hospital deaths were AKI (p = 0.637) between the two groups; 13.8% of trial
observed. participants and 15.8% of comparison patients had AKI.
All inferences were consistent across a series of sensi-
Post hoc analysis with propensity score‑weighted tivity analyses that used different modeling techniques,
comparison group including propensity adjustment, multiple imputation
The post hoc analysis focused on 457 subjects (pilot and propensity adjustment, and a post hoc sensitivity analysis
non-pilot) from 10 of the 11 sites with the intention to adjusted for patient sex (Table 4).
treat the analysis for which data were available. MPOG
identified in their registry 177,519 surgical cases from Discussion
10 trial sites that occurred during the trial recruitment The goal of predictive technology is to warn clinicians
period or within approximately 7 months thereafter. and prevent untoward events by timely intervention. We
These were then limited to the 16,253 cases that matched report that the use of the hypotension prediction algo-
the trial inclusion/exclusion criteria as closely as pos- rithm was associated with a 35% reduction in the dura-
sible: noncardiac inpatient surgeries that lasted at least tion of intraoperative hypotension versus comparison
3 hours and had hemodynamic monitoring through an patients in a propensity-weighed model. Our results align
arterial line (Supplementary Figure S2). Of those, 457 with several single-center trials that have tested this soft-
were participants enrolled in the trial, and the remaining ware device. In a small, randomized controlled trial in the
15,796 were designated as the comparison group. Table 2 Netherlands, Wijnberge et al. (2020) observed a statisti-
shows the patient characteristics and surgical charac- cally significant reduction in time-weighted average MAP
teristics and the standardized mean difference between < 65 mmHg (Wijnberge et al., 2020). Their secondary
groups. Figure 2 shows a balance plot for the group char- outcome (median minutes of MAP < 65 mmHg) was 8.0
acteristics, unweighted and weighted; it indicates that all min in the intervention group vs 32.7 min in the control
but one of the weighted standardized mean differences group. This is similar to the median of 9 min of hypoten-
between the trial and comparison groups are less than sion we observed in the post hoc analysis of our multi-
0.1, suggesting that the groups are comparable on meas- center trial group; however, our U.S. comparison group
ured characteristics. had a much lower duration of hypotension (median 15
Patients in the study group experienced fewer min- min) than this European comparison group. A trial in
utes of intraoperative hypotension than the comparison Greece showed a 28% reduction in time-weighted aver-
group [median 9 (3, 20) vs. 15 (5, 39) min, p < 0.001], as age MAP < 65 mmHg in the intervention group, although
determined using EHR data from the MPOG registry they also observed an increase in hypertension and
(Table 3). In the inverse probability treatment weigh- increased use of phenylephrine in the intervention group
ing propensity model, patients in the trial had reduced (Tsoumpa et al., 2021). Five other randomized trials
total time spent in hypotension [exponentiated coeffi- observed that the use of the predictive algorithm reduced
cient − 0.35 (95%CI − 0.43, − 0.27); p < 0.001]. The use intraoperative hypotension (Grundmann et al., 2021;
of predictive software in hemodynamic management was Schneck et al., 2020; Murabito et al., 2022; Šribar et al.,
Bao et al. Perioperative Medicine (2024) 13:13 Page 7 of 12
Table 2 Patient and surgical characteristics in the single-arm trial and contemporaneous comparison group
Contemporaneous Trial Standardized
comparison N = 457 mean
N = 15,796 difference
Table 2 (continued)
Contemporaneous Trial Standardized
comparison N = 457 mean
N = 15,796 difference
Table 3 Comparison of intraoperative hypotension and acute kidney injury in the single-arm trial versus contemporaneous
comparison group
Contemporaneous Trial Crude modelb Inverse probability of treatment
comparison N = 457 weighing propensity model
N = 15,796
Effect Estimatec P-Value Effect estimatec P value
(95% CI) (95% CI)
Duration of Hypotension 15 [5, 39] 9 [3, 20] − 0.31 (− 0.39 to − 0.22) < 0.001 − 0.35 (− 0.43 to − 0.27) < 0.001
(MAP < 65)a
Acute kidney injury 1898/12,421 (15.3%) 45/325 (13.8%) 0.93 (0.67 to 1.27) 0.674 0.91 (0.63 to 1.33) 0.637
CI confidence interval: MAP mean arterial blood pressure)
Data is presented as median [quartile 1, quartile 3] in minutes, or n (%) depending on descriptive statistics
a
Data is only available for 15,749 observations
b
All models are adjusting for time in which blood pressure is measured (i.e., time at risk)
c
Effect estimates for hypotension are reported as exponentiated beta coefficients, whereas effect estimates for acute kidney injury are reported as odds ratios
2023; Lorente et al., 2023), although they found no differ- followed the study protocol. In addition, the comparison
ence in lab values, clinical outcomes (Šribar et al., 2023), group had a low incidence of hypotension compared to
or tissue oxygenation (Lorente et al., 2023). An observa- the European trials (about one-third of the time-weighted
tional study of the software device found that those in average in the Wijnberge et al trial), possibly indicating
the HPI group had less hypotension, fewer postoperative the practice difference between the USA and Europe.
complications, and lower length of stay (Solares et al., Although the use of the predictive algorithm was asso-
2023), and a second observational study saw shorter ICU ciated with a reduction in the duration of intraoperative
ventilation time among the HPI group, although no dif- hypotension, we did not observe a statistically signifi-
ference in AKI (Reddy et al., 2023). cant reduction in AKI, our secondary outcome for the
Of the known trials of this predictive algorithm, only post hoc analysis. The incidence of AKI was 13.8% in the
one U.S.-based pilot study showed no difference in hypo- trial participants versus 15.3% in comparison patients.
tension in the intervention and control groups, with This study was not powered to detect a reduction in AKI;
both groups having a time-weighted average MAP < 65 rather, AKI was one component of the composite safety
mmHg of 0.14 mmHg (Maheshwari et al., 2020). A sub- outcome in the single-arm trial, and it was a secondary
group analysis by the authors demonstrated that the outcome for the post hoc comparison group analysis.
time-weighted average MAP < 65 mmHg reduced to 0.06 The effect of hypotension on AKI could be influenced
mmHg in the subset of alerts where anesthesia providers by patient baseline comorbidities and procedure risks.
Bao et al. Perioperative Medicine (2024) 13:13 Page 10 of 12
Mathis et al. (2020) reported no association of increased in methodology to calculate hypotension time. These
risk of AKI across all blood pressure ranges in patients limitations were the impetus for the post hoc analysis.
with low risk and a strong association in patients of the The design of the post hoc analysis, which compares
highest risk (Mathis et al., 2020). The acute profound trial participants to a propensity-weighted compari-
intraoperative hypotension could be hard to prevent and son group, is able to control for observable differences
result in more kidney injury. Our trial did allow the free- between groups. The weighted models may not fully
dom to anesthesia providers to ignore the alerts from control for differences between the trial participants
devices. Although we achieved a 35% reduction in hypo- and comparison patients on variables that are not
tension time, it may not be enough to convey kidney pro- observed or measured in the registry data. Also, as
tection. Also, operations only account for a small portion noted above, the trial was not designed or powered to
of patients’ hospital stays, and hypotension could occur assess AKI, which we used as a secondary outcome in
while patients are in postoperative care units, intensive our post hoc analysis.
care units, and floors, which would not be prevented by Despite the limitations, this analysis is a valuable addi-
this trial. tion to the literature on the use of predictive algorithms
It is worth noting that the measures of hypotension to guide hemodynamic monitoring and prevention of
duration in the prospective single-arm trial (in Table 1) intraoperative hypotension. It is the first multicenter
and the post hoc analysis with registry data (in Table 3) study of this predictive algorithm with the largest sample
are not identical. Data on the duration of hypotension size and adds to the existing single-center trials. Addi-
for the trial participants were extracted directly from the tionally, while the post hoc design cannot provide as
EV1000 monitor every 20 s, while the post hoc analysis strong a causal relationship as a randomized control trial,
outcome was calculated using blood pressure data from the observational design with propensity-score weighted
the anesthesia record at 1 min intervals. The EV 1000 models is a rigorous method for estimating impacts
monitor was attached to patients after the arterial line in situations when a randomized trial is not possible for
was placed and disconnected at the end of operations, ethical, practical, or financial reasons. Our IPTW model
which may not count all data from the anesthesia record provided a well-balanced match to compare to trial par-
and could miss volatile blood pressure swing at induc- ticipants. This analysis demonstrates that real-world data
tion and emergence phases. Importantly, however, in can be used in conjunction with trial data to advance the
the post hoc analysis that compares trial participants to evidence base.
cotemporaneous control patients, the hypotension data There is now a large body of observational evidence to
source and calculation method are the same for both show that intraoperative hypotension increases patient
groups, which makes the comparison valid. Additionally, risks for adverse outcomes related to perfusion, and there
in the prospective single-arm trial, 17 cases of AKI were is a consensus statement recommending that anesthesia
reported by sites as adverse events by using the AKIN providers maintaining systolic arterial pressure above
definition, whereas in the post hoc analysis, 45 instances 100mmHg and MAP above 60 to 70 mmHg to attempt
of AKI were identified with KDIGO diagnosis criteria. to reduce patient risk (Sessler et al., 2019). Neverthe-
It should be recognized that as a safety measure, kidney less, prevention and management of hypotension should
function was followed until postoperative day 3 in the be targeted to the underlying physiological changes of
prospective trial, while the registry data used for post hoc volume status, cardiac contractility, and vascular tone.
analysis reported kidney function until 7 days after oper- Aggressive or inappropriate volume overload or over-
ation, which may explain the discrepancy. use of vasopressors without accounting for hypovolemia
This study has several limitations. The trial group was could lead to worsening surgical outcomes. A recent mul-
subject to the Hawthorne effect; providers, aware of the ticenter study demonstrated an increased incidence of
trial, may have been more inclined towards aggressive AKI with decreased administration of crystalloid, shorter
correction of hypotension. In the initial analysis, the duration of hypotension, and higher usage of vasopres-
trial outcomes were compared to a historical compari- sor (Chiu et al., 2022; Shin et al., 2018). EV 1000 monitor
son cohort; however, the data source differed between could potentially provide a quick insight into the patient’s
groups. The blood pressure data for trial participants hemodynamic status. More research is needed on the
was pulled from the EV 1000 monitors, whereas for appropriate treatment for intraoperative hypotension
the historical comparison cohort, it came from a clini- in order to maintain hemodynamic stability most effec-
cal registry. Additionally, because we did not receive tively while minimizing overtreatment (Chiu et al., 2022;
patient-level data for the comparison cohort, we were Shin et al., 2018). Trials are also needed to more rigor-
unable to control for differences in the patient and pro- ously determine whether the use of predictive algorithms
cedure between the groups. There is also a difference to prevent hypotension can reduce organ system damage
Bao et al. Perioperative Medicine (2024) 13:13 Page 11 of 12
and other complications, such as AKI, myocardial injury, restrictions apply to the availability of these data, which were used under
license for the current study, and so are not publicly available. Data are how-
postoperative delirium, and mortality. ever available from the authors upon reasonable request and with permission
of the Multicenter Perioperative Outcomes Group.
Conclusion
The use of prediction software for blood pressure man- Declarations
agement was associated with a clinically meaningful Ethics approval and consent to participate
reduction in the duration of intraoperative hypotension. The trial was approved by a central Institutional Review Board (IRB; Western
Further studies must investigate whether predictive IRB approval #1-1131056-1) and 8 local IRBs. Written informed consent was
obtained from each subject in the prospective trial, but not from subjects who
algorithms to prevent hypotension can reduce adverse were retrospectively included in the comparison groups.
patient outcomes.
Consent for publication
Not applicable.
Abbreviations
AKI Acute kidney injury Competing interests
ASA American Society of Anesthesiologist XB, HC, CC, GM, LM, DP, MP, SR, MT, GW, MW declare that they have no com-
BMI Body mass index peting interests.
CI Confidence interval KD received grants or contracts in the past 36 months from Anesthesia Patient
EHR Electronic Health Record Safety Foundation regarding NORA, and lecture honoraria from Northwest
HPI Hypotension Prediction Index Anesthesia Seminars; she serves as Committee Chair for the ASA Committee
IOH Intraoperative hypotension on Practice Parameters.
IPTW Inverse probability of treatment weighted DD received grants or contracts in the past 36 months from NIH – NIA and
KDIGO Kidney disease: improving global outcomes Masimo Inc, and consulting fees from Masimo Inc.
MAP Mean arterial pressure SK received grants or contracts in the past 36 months from Edwards Lifes-
MPOG Multicenter Perioperative Outcomes Group ciences, PCORI, and Hemosonics; honoraria from ROTEM for lecture; and
SD Standard deviation reimbursement from Edwards Lifesciences for travel to a research meeting.
KM received grants or contracts and consulting fees in the past 36 months
from Edwards Lifesciences and Phillips Healthcare.
Supplementary Information NS received grants or contracts in the past 36 months from Edwards Lifes-
The online version contains supplementary material available at https://doi. ciences, Blue Cross Blue Shield Michigan, National Institute of Aging, American
org/10.1186/s13741-024-00369-9. Heart Association.
Author details
Additional file 1: Appendix 1. Statistical Analysis and Reporting Plan 1
Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital,
Additional file 2: Appendix 2. Statistical analysis plan Boston, MA, USA. 2 Department of Anesthesiology, University of Michigan
Medical School, Ann Arbor, MI, USA. 3 Department of Anesthesiology, Henry
Additional file 3: Supplementary Figure S1. HPI Trial Diagram
Ford Health System, Detroit, MI, USA. 4 Department of Anesthesiology
Additional file 4: Supplementary Figure S2. Data Flow Diagram and Critical Care, Hospital of the University of Pennsylvania, Philadelphia, PA,
USA. 5 Department of Anesthesiology and Perioperative Medicine, Oregon
Health & Science University, Portland, OR, USA. 6 Department of Anesthe-
Acknowledgements sia, Stanford University, Stanford, CA, USA. 7 Department of Anesthesiology
Kathryn Cody, Ph.D. and Ariel Muller, MS (Massachusetts General Hospital, and Pain Medicine, University of Washington School of Medicine, Seattle, WA,
Boston, MA) contributed to the statistical analysis and model building. Anna USA. 8 Department of Anesthesiology, Indiana University School of Medicine,
L. Christensen, Ph.D. (Principal Researcher, Mathematica, Washington, DC) Indianapolis, IN, USA. 9 Department of Anesthesiology, Duke University School
assisted with drafting the manuscript. of Medicine, Durham, NC, USA. 10 Department of Anesthesiology, University
of Colorado Anschutz Medical Campus, Aurora, CO, USA. 11 Department
Authors’ contributions of Anesthesiology, College of Physicians & Surgeons of Columbia University,
XB, SK, DP, MW, GM, SR, DD, MP, KD, LM, MT, CC, GW, and KM were each trial New York, NY, USA. 12 Department of General Anesthesiology, Cleveland Clinic,
site-PIs or co-PIs who made substantial contributions to the conception, Cleveland, OH, USA.
acquisition, and interpretation of data. NS provided comparison group data
and made substantial contributions to the conception, acquisition, and inter- Received: 6 August 2023 Accepted: 25 February 2024
pretation of data. HC made substantial contributions to the conception and
analysis of data. All authors approved the submitted version and agreed to be
accountable for their own contributions.
Funding
The study sponsor is Edwards Lifesciences LLC, One Edwards Way, Irvine, CA
92614. References
American Society of Anesthesiologists. Standards for Basic Anesthetic
Availability of data and materials Monitoring. https://www.asahq.org/standards-and-guidelines/standards-
Data from the trial: There is a sub-set of data that support the findings of this for-basic-anesthetic-monitoring. Updated October 20, 2010. Affirmed
study that are available from Edwards Lifesciences, but restrictions apply December 13, 2020. Accessed August 22, 2022.
to the availability of these data, which were used under license for the Bijker JB, van Klei WA, Kappen TH, van Wolfswinkel L, Moons KG, Kalkman CJ.
current study, and so are not publicly available. Data are however available Incidence of intraoperative hypotension as a function of the chosen
from the authors upon reasonable request and with permission of Edwards definition: literature definitions applied to a retrospective cohort using
Lifesciences. automated data collection. Anesthesiol. 2007;107:213–20.
Data from the comparison group: The data that support the findings of this Chiu C, Fong N, Lazzareschi D, Mavrothalassitis O, Kothari R, Chen L, et al.
study are available from the Multicenter Perioperative Outcomes Group but Fluids, vasopressors, and acute kidney injury after major abdominal
Bao et al. Perioperative Medicine (2024) 13:13 Page 12 of 12
surgery between 2015 and 2019: a multicentre retrospective analysis. Br J from baseline or absolute thresholds, and acute kidney and myocardial
Anaesth. 2022;129:317–26. injury after noncardiac surgery: a retrospective cohort analysis. Anesthe-
Davies SJ, Vistisen ST, Jian Z, Hatib F, Scheeren TWL. Ability of an arterial wave- siol. 2017;126:47–65.
form analysis-derived Hypotension Prediction Index to predict future Schneck E, Schulte D, Habig L, Ruhrmann S, Edinger F, Markmann M, et al.
hypotensive events in surgical patients. Anesth Analg. 2020;130:352–9. Hypotension Prediction Index based protocolized haemodynamic
Disease K. Improving Global Outcomes (KDIGO) Acute Kidney Injury Work management reduces the incidence and duration of intraoperative
Group. KDIGO Clinical Practice Guideline for Acute Kidney Injury. Kidney hypotension in primary total hip arthroplasty: a single centre feasibility
inter Suppl. 2012;2:1–138. randomised blinded prospective interventional trial. J Clin Monit Com-
Frassanito L, Giuri PP, Vassalli F, Piersanti A, Longo A, Zanfini BA, et al. Hypoten- put. 2020;34:1149–58.
sion Prediction Index with non-invasive continuous arterial pressure Sessler DI, Bloomstone JA, Aronson S, Berry C, Gan TJ, Kellum JA, et al. Periop-
waveforms (ClearSight): clinical performance in gynaecologic oncologic erative Quality Initiative-3 workgroup. Perioperative Quality Initiative con-
surgery. J Clin Monit Comput. 2022a;36:1325–32. sensus statement on intraoperative blood pressure, risk and outcomes for
Frassanito L, Sonnino C, Piersanti A, Zanfini BA, Catarci S, Giuri PP, et al. Per- elective surgery. Br J Anaesth. 2019;122:563–74.
formance of the Hypotension Prediction Index with noninvasive arterial Sessler DI, Meyhoff CS, Zimmerman NM, Mao G, Leslie K, Vásquez SM, et al.
pressure waveforms in awake cesarean delivery patients under spinal Period-dependent associations between hypotension during and for four
anesthesia. Anesth Analg. 2022b;134:633–43. days after noncardiac surgery and a composite of myocardial infarction
Gregory A, Stapelfeldt WH, Khanna AK, Smischney NJ, Boero IJ, Chen Q, et al. and death: a substudy of the POISE-2 Trial. Anesthesiol. 2018;128:317–27.
Intraoperative hypotension Is associated with adverse clinical outcomes Shah NJ, Mentz G, Kheterpal S. The incidence of intraoperative hypotension
after noncardiac surgery. Anesth Analg. 2021;132:1654–65. in moderate to high risk patients undergoing non-cardiac surgery:
Grundmann CD, Wischermann JM, Fassbender P, Bischoff P, Frey UH. Hemo- a retrospective multicenter observational analysis. J Clin Anesth.
dynamic monitoring with Hypotension Prediction Index versus arterial 2020;66:109961.
waveform analysis alone and incidence of perioperative hypotension. Shin B, Maler SA, Reddy K, Fleming NW. Use of the Hypotension Predic-
Acta Anaesthesiol Scand. 2021;65:1404–12. tion Index during cardiac surgery. J Cardiothorac Vasc Anesth.
Hallqvist L, Granath F, Fored M, Bell M. Intraoperative hypotension and 2021;35:1769–75.
myocardial infarction development among high-risk patients undergo- Shin CH, Long DR, McLean D, Grabitz SD, Ladha K, Timm FP, et al. Effects of
ing noncardiac surgery: a nested case-control study. Anesth Analg. intraoperative fluid management on postoperative outcomes: a hospital
2021;133:6–15. registry study. Ann Surg. 2018;267:1084–92.
Hatib F, Jian Z, Buddi S, Lee C, Settels J, Sibert K, et al. Machine-learning Solares GJ, Garcia D, Monge Garcia MI, Crespo C, Rabago JL, Iglesias F, et al.
algorithm to predict hypotension based on high-fidelity arterial pressure Real-world outcomes of the hypotension prediction index in the man-
waveform analysis. Anesthesiol. 2018;129:663–74. agement of intraoperative hypotension during non-cardiac surgery: a
Keuffel EL, Rizzo J, Stevens M, Gunnarsson C, Maheshwari K. Hospital costs retrospective clinical study. J Clin Monit Comput. 2023;37(1):211–20.
associated with intraoperative hypotension among non-cardiac surgical Šribar A, Jurinjak IS, Almahariq H, Bandić I, Matošević J, Pejić J, et al. Hypoten-
patients in the US: a simulation model. J Med Econ. 2019;22:645–51. sion prediction index guided versus conventional goal directed therapy
Lorente JV, Ripollés-Melchor J, Jiménez I, Becerra AI, Mojarro I, Fernández- to reduce intraoperative hypotension during thoracic surgery: a rand-
Valdes-Bango P, et al. Intraoperative hemodynamic optimization using omized trial. BMC Anesthesiol. 2023;23(1):101.
the hypotension prediction index vs. goal-directed hemodynamic ther- Stapelfeldt WH, Yuan H, Dryden JK, et al. The SLUScore: A Novel Method for
apy during elective major abdominal surgery: the Predict-H multicenter Detecting Hazardous Hypotension in Adult Patients Undergoing Noncar-
randomized controlled trial. Front Anesthesiol. 2023;2:1–14. https://doi. diac Surgical Procedures. Anesth Analg. 2017;124:1135–52.
org/10.3389/fanes.2023.1193886. Tsoumpa M, Kyttari A, Matiatou S, Tzoufi M, Griva P, Pikoulis E, et al. The use
Maheshwari K, Buddi S, Jian Z, Settels J, Shimada T, Cohen B, et al. Performance of the Hypotension Prediction Index integrated in an algorithm of goal
of the Hypotension Prediction Index with non-invasive arterial pres- directed hemodynamic treatment during moderate and high-risk sur-
sure waveforms in non-cardiac surgical patients. J Clin Monit Comput. gery. J Clin Med. 2021;10:5884.
2021;35:71–8. van der Ven WH, Terwindt LE, Risvanoglu N, Ie ELK, Wijnberge M, Veelo DP, et al.
Maheshwari K, Khanna S, Bajracharya GR, et al. A randomized trial of continu- Performance of a machine-learning algorithm to predict hypotension in
ous noninvasive blood pressure monitoring during noncardiac surgery. mechanically ventilated patients with COVID-19 admitted to the inten-
Anesth Analg. 2018;127:424–31. sive care unit: A cohort study. J Clin Monit Comput. 2022;36:1397–405.
Maheshwari K, Shimada T, Yang D, Khanna S, Cywinski JB, Irefin SA, et al. Hypo- van Waes JA, van Klei WA, Wijeysundera DN, van Wolfswinkel L, Lindsay TF,
tension Prediction Index for prevention of hypotension during moderate- Beattie WS. Association between intraoperative hypotension and myo-
to high-risk noncardiac surgery. Anesthesiol. 2020;133:1214–22. cardial injury after vascular surgery. Anesthesiol. 2016;124:35–44.
Mathis MR, Naik BI, Freundlich RE, Shanks AM, Heung M, Kim M, et al. Multi- Wesselink EM, Kappen TH, Torn HM, Slooter AJC, van Klei WA. Intraoperative
center Perioperative Outcomes Group Investigators. Preoperative risk and hypotension and the risk of postoperative adverse outcomes: a system-
the association between hypotension and postoperative acute kidney atic review. Br J Anaesth. 2018;121:706–21.
injury. Anesthesiol. 2020;132:461–75. Wijnberge M, Geerts BF, Hol L, Lemmers N, Mulder MP, Berge P, et al. Effect
Murabito P, Astuto M, Sanfilippo F, La Via L, Vasile F, Basile F, et al. Proactive of a machine learning-derived early warning system for intraoperative
management of intraoperative hypotension reduces biomarkers of organ hypotension vs standard care on depth and duration of intraoperative
injury and oxidative stress during elective non-cardiac surgery: a pilot hypotension during elective noncardiac surgery: the HYPE randomized
randomized controlled trial. J Clin Med. 2022;11(2):392. clinical trial. JAMA. 2020;323:1052–60.
Nanji KC, Shaikh SD, Jaffari A, Franz C, Bates DW. A Monte Carlo simulation to Wijnberge M, Schenk J, Bulle E, Vlaar AP, Maheshwari K, Hollmann MW, et al.
estimate the additional cost associated with adverse medication events Association of intraoperative hypotension with postoperative mor-
leading to intraoperative hypotension and/or hypertension in the United bidity and mortality: systematic review and meta-analysis. BJS Open.
States. J Patient Saf. 2021;17:e758–64. 2021a;5:zraa018.
Ranucci M, Barile L, Ambrogi F, Pistuddi V, Surgical and Clinical Outcome Wijnberge M, van der Ster BJP, Geerts BF, de Beer F, Beurskens C, Emal D, et al.
Research (SCORE) Group. Discrimination and calibration properties of the Clinical performance of a machine-learning algorithm to predict intra-
hypotension probability indicator during cardiac and vascular surgery. operative hypotension with noninvasive arterial pressure waveforms: a
Minerva Anestesiol. 2019;85:724–30. cohort study. Eur J Anaesthesiol. 2021b;38:609–15.
Reddy VS, Stout DM, Fletcher R, Barksdale A, Parikshak M, Johns C, et al.
Advanced artificial intelligence-guided hemodynamic management
within cardiac enhanced recovery after surgery pathways: A multi-institu- Publisher’s Note
tion review. JTCVS Open. 2023 Aug;7(16):480–9. Springer Nature remains neutral with regard to jurisdictional claims in pub-
Salmasi V, Maheshwari K, Yang D, Mascha EJ, Singh A, Sessler DI, et al. Relation- lished maps and institutional affiliations.
ship between intraoperative hypotension, defined by either reduction