Perioperative Hypotension in Gynaecologic Oncologic Surgery: HPI-ClearSight Versus Arterial Waveform Analysis

Overview

Background Intraoperative hypotension is associated with increased morbidity and mortality. The Hypotension Prediction Index (HPI) is an advancement of the arterial waveform analysis to predict intraoperative hypotension minutes before episodes occur enabling preventive treatments. This study will test the hypothesis that a hemodynamic treatment protocol based on HPI working with non-invasive ClearSight system reduces intraoperative hypotension when compared to standard goal directed therapy (GDT) in patients undergoing gynaecologic oncologic surgery. Methods A retrospective analysis of 68 adult consecutive patients undergoing gynaecologic oncologic surgery with non-invasive arterial pressure monitoring using either index guidance (HPI) or classic ClearSight system waveform analysis depending on availability (ClearSight, n = 36; HPI, n = 32) will be conducted. A hemodynamic GDT protocol was applied in both groups. The primary endpoint will be the incidence and duration of hypotensive events defined as MAP \<65 mmHg evaluated by time-weighted average of hypotension.

INTRODUCTION Intraoperative hypotension (IOH) is a common adverse event during noncardiac surgery \[1,2\]. It is associated with an increased incidence of acute kidney injury, myocardial injury, neurological deficiencies, as well as an increased 30-day operative mortality \[3-7\]. Organ injuries were shown to be associated with the depth, frequency, and duration of hypotensive episodes \[4,8\]. Data indicate that a mean arterial pressure of 65 mmHg serves as a threshold to predict myocardial and kidney injury \[9-11\]. Recently developed minimally invasive methods of the arterial waveform analysis allow calculation of cardiac output, stroke volume, stroke volume variation, and systemic vascular resistance. This deeper insight into hemodynamics enables goal-directed therapy approaches; however, the given information only allows to react to events instead of preventing them \[12,13\]. Interestingly, recent data indicate a significant reduction in postoperative organ dysfunction by preventing IOH, suggesting a potential benefit of early intervention \[14\]. Surgical interventions for cancer mass reduction are among the most invasive operations in gynaecology. Extensive surgical trauma results in severe intra- and perioperative volume shifts and unstable hemodynamics \[15\]. Major abdomino-pelvic surgery for gynaecological oncologic surgery (GOS) may include hysterectomy, oophorectomy, omentectomy, colectomy, removal of the lymph nodes and peritoneal stripping. Surgery is carried out through laparoscopy or a laparotomic incision and is often associated with significant blood loss \[16\]. Hypotension during GOS is common, and as it is associated with the potential for harm, it requires prompt evaluation and treatment \[16\]. Extensive fluid resuscitation in peritoneal cancer patients is associated with a poor postoperative outcome independent from the underlying malignant disease and avoiding fluid overload is recommended \[15\]. Therefore, the prediction of intraoperative hypotension and consequently its prevention by proactive treatment may show beneficial effects for patients. Machine learning, a discipline within computer science used to analyse large data sets and to develop predictive models, has evident applications in health care \[17,18\]. Several attempts to use algorithms as an aid in anaesthesiology practice recently received renewed attention, with the aim of optimizing patients' perioperative status, primarily focusing on detection of early hemodynamic instability and prediction of hypotension \[19\]. The Hypotension Prediction Index - HPI (Edwards Lifesciences, Irvine, USA) is an algorithm based on the complex analysis of features in high fidelity arterial pressure waveform recordings developed to observe subtle signs that could predict the onset of hypotension in surgical and intensive care unit patients \[20\]. HPI is a unitless number that ranges from 1 to 100, and as the number increases, the likelihood or risk of a hypotensive event (defined as a mean arterial pressure \[MAP\] \<65 mmHg for more than 1 minute) occurring in the near future increases \[20\]. A validation study conducted on surgical patients reported high sensitivity and specificity of HPI for predicting hypotension 5, 10 and 15 min before the event occurred \[21\]. The development of the algorithm was based on invasive arterial line waveform data; however, only a small fraction of patients having noncardiac surgery requires invasive arterial monitoring \[22\]. Estimation of arterial pressure waveforms from a non-invasive finger cuff (ClearSight) is well established, and MAP measured by ClearSight system could be considered as an alternative for mean radial arterial pressure \[23,24\]. The HPI algorithm, is recently reported to work also with a non-invasive waveform estimate \[25-27\]. This study will compare a hemodynamic management based on the HPI algorithm working with non-invasive ClearSight system with a classic ClearSight-based one in terms of incidence, duration, and severity of intraoperative hypotensive events evaluated by time-weighted average of hypotension in patients undergoing GOS. MATERIALS AND METHODS This single-center retrospective observational study was approved by the Internal Ethics Committee (ID 3664, protocol number 10077/21). Written informed consent will be obtained from all the patients involved. The primary endpoint is the incidence, duration, and severity of hypotensive events (defined as MAP \< 65 mmHg for at least 1 minute) evaluated by time-weighted average (TWA) MAP of hypotension in the two groups of patients. The TWA-MAP is a combination of severity and duration of the hypotensive events, in relation to the total surgery time. It is calculated by using the sum of the area under the threshold divided by the duration of surgery \[28\]. The threshold for hypotensive events is defined as a MAP below 65 mmHg for at least 1-min duration. Hypotension duration time ends after re-increasing MAP values upon ≥65 mmHg for at least 1 min. Secondary endpoints consists of number of patients with hypotensive events, number of events per (respective) patient, cumulative and average duration of hypotension, combined with numbers of hypotensive events \<65 and \<50 mmHg. The data were collected during a limited period of time in which new hemodynamic monitoring sensors (ClearSight) were evaluated during a time marketing release. The first 68 patients undergoing GOS at IRCCS Policlinico A.Gemelli Foundation between December 2019 and February 2020 were monitored either with classic ClearSight sensor (n= 36) and ClearSight sensor with HPI-enabled (n = 32). Inclusion criteria consisted of elective GOS, age \>18 years. Exclusion criteria were patients not in sinus rhythm, ejection fraction \<30%, severe aortic valve stenosis, emergency surgery, acute myocardial ischemia, pregnancy. Standard monitoring (Life Scope TR, Nihon Kohden Co, Tokyo, Japan) included a 5-lead electrocardiogram, pulse oximetry, non-invasive blood pressure (NIBP) and eventual invasive blood pressure (IBP). In addition to standard monitoring, all patients had a non-invasive hemodynamic monitoring with ClearSight (Edwards Lifesciences, Irvine, CA). After arriving at the operating theatre, NIBP measurement using an automated digital sphygmomanometer on the right arm was started and the ClearSight system was attached to a finger of the left arm of the patients. We connected the ClearSight monitor with an interface cable to the patient monitor. In patients requiring IBP monitoring an arterial cannula was placed contralateral to the ClearSight cuff. The ClearSight reference system was zeroed at the level of the right atrium. The blood pressure value from the finger cuff (CS-BP) was reported on the main monitor. NIBP measurement using the automated digital sphygmomanometer was performed at 5-min intervals. For therapy purposes we defined hypotension as an absolute value of CS-BP MAP \< 65 mmHg. Incidence and duration of hypotensive episodes and interventions were registered. Bradycardia was defined as a heart rate (HR) \< 60 bpm. A large bore i.v. catheter was inserted in a forearm vein, and Cefazoline 2 gr, Dexamethasone 4 mg and Omeprazole 40 mg were administered. General anesthesia was induced with sufentanil 0.2 mcg/kg (ideal body weight), propofol 2 mg/kg (actual body weight), and rocuronium 0.6 mg/kg (ideal body weight). During surgery, patients were positioned in Trendelenburg position with both arms spread out on arm-positioning devices. Anesthesia was maintained with Sevoflurane to maintain a Bispectral Index value between 40-50. Additional boluses of sufentanil and rocuronium were administered when needed. All patients received a GDT hemodynamic protocol to optimize the cardiac output and oxygen supply. As part of the GDT protocol, patients received crystalloid and/or fluids to maintain SVV\<13% or to avoid SV decrease of more than 10% of the baseline. Vasopressor and inotropic medication were administered upon request of the clinician in order to maintain a MAP ≥65. In the HPI group, the attending anesthesiologist was free to use liquid, vasopressors and inotropic drugs, but he was allowed to read on the Hemosphere monitor HPI and secondary parameters (for evaluation of preload, cardiac contractility, and afterload as possible causes for hypotension), and to act in a preventive way in order to avoid hypotension. All data were downloaded from the HemoSphere monitor, including HPI, CS-MAP, CS systolic arterial pressure, CS diastolic arterial pressure. All downloaded data consisted of 20-second interval averaged data points. Data were transferred to a computer for analysis via an USB drive. Every file was appointed with an automated generated code by the machine and was identifiable by an ID number contained within it. The HPI algorithm estimates the probability of occurrence in the near future of a hypotensive event taking the arterial pressure waveform as the input to compute an index value that ranges between 1 and 100. In this study, instead of invasive arterial waveform data, we used the non-invasive arterial pressure waveform of ClearSight. In the HemoSphere monitor, poor quality arterial waveforms were automatically detected by the arterial waveform processing algorithms and excluded from the computation of the 20-sec averages. Statistical analysis Data collected perioperatively and downloaded from the HemoSphere platform will be analyzed using the Acumen Analytics Software (Edwards Lifesciences Corp.). Categorical data will be presented as frequencies with percentages. Differences will be analyzed with the Fisher's exact test. For quantitative variables test for normal distribution with the Shapiro-Wilk test will be used. Data will be presented as means ±standard deviation (SD) in case of normally distribution, otherwise, as median and interquartile range (IQR). Linear variables will be analyzed using the Student's unpaired t-test in normal distributed variables and the Mann-Whitney U-test in non-normally distributed variables. P-values less than 0.05 will be considered statistically significant. Statistical analysis will be performed with SPSS Statistics.