Robot-Assisted versus Open Pancreatoduodenectomy: Identifying Perioperative Anesthetic Factors associated with Postoperative Morbidity. A retrospective cohort study.

Robot-assisted pancreatoduodenectomy (RAPD) poses several challenges concerning perioperative anesthetic guidance compared to open pancreatoduodenectomy (OPD), e.g. combined pneumoperiotoneum with reversed-Trendelenburg positioning. The primary objective of this observational study is to specify these anesthetic differences of RAPD versus OPD and secondly to identify independent anesthetic factors associated with patient morbidity following RAPD. of major postoperative morbidity (CD < III) and RAPD patients above versus below median intraoperative blood loss (high and low intraoperative blood loss, respectively). Normal distribution of data was assessed using a combination of visual inspection of histograms and Q-Q plots and the Shapiro-Wilk test. Numerical data were presented as mean (standard deviation, SD) or median (interquartile range, IQR), as appropriate. Categorical data were presented with frequencies and percentages. An independent sample T-test or Mann-Whitney U-test was performed in comparing numerical data, a χ2 or Fisher’s exact test in categorical data. In order to identify independent predictors of major morbidity following RAPD (CD ≥ III), a logistic regression model was constructed using a backward stepwise approach. Baseline and perioperative variables showing a signicant difference between minor (CD < III) versus major (CD ≥ III) morbidity after RAPD and of suspected clinical relevance in predicting major morbidity after RAPD were included in the regression model. Results herein were presented as odds ratio (OR) with corresponding 95% condence interval (CI). Throughout the study two-tailed P-values of < 0.05 were considered statistically signicant. Statistical analysis was carried out using IBM SPSS Statistics (version 24.0, Armonk, NY, USA; IBM Corp.). Values are presented as number (proportion) or depending on normality distribution of cases as mean ± SD or median (interquartile range). X n where n represents the number of missing cases. ASA, American Society of Anesthesiologists Classication; BMI, Body Mass Index; CD, Clavien Dindo; CVA, Cerebro Vascular Accident; e-GFR, Estimated Glomerular Filtration Rate; OPD, Open Pancreatoduodenectomy; RAPD, Robot-Assisted Pancreatoduodenectomy; TIA, Transient Ischemic Attack.

surgery [1]. Various cohort studies already compared surgical outcomes of robot-assisted pancreatoduodenectomy (RAPD) with conventional open pancreatoduodenectomy (OPD). Despite increased surgical times, RAPD is characterized by lower intraoperative blood loss with equal postoperative mortality and similar oncological outcome [2][3][4]. Although the surgical feasibility of RAPD is extensively reported in literature, little is known about perioperative anesthetic concerns of the robotic approach and how perioperative anesthetic factors affect postoperative outcome.
Determining optimal intraoperative uid regimens during abdominal surgery in relation to postsurgical morbidity is part of ongoing scienti c debate. The 2018 RELIEF trial described an association between a more restrictive intraoperative net uid balance (median 3.7 liters) and an increased rate of kidney injury following major abdominal surgery [5]. However, Grant et al. reported no differences in postoperative major morbidity after randomizing between either a net liberal (12 ml kg/hr) versus net restrictive (6 ml kg/h) intraoperative uid balance for pancreatic surgery (irrespective of the surgical approach) [6]. RAPD presents several speci c challenges concerning the perioperative anesthetic guidance. The patient is exposed to pneumoperitoneum, reversed-Trendelenburg positioning during the majority of surgical time, increasing central venous pressure, expected higher blood carbon dioxide levels and decreasing cardiac output [7][8][9]. We therefore hypothesize differences between RAPD and OPD in perioperative demands for vasopressor and uid administration.
The primary objective of this retrospective cohort study is to evaluate perioperative anesthetic differences of RAPD versus OPD. Secondly, this study aims to identify independent anesthetic factors associated with patient morbidity following RAPD.

Methods
The Medical Ethics Committee approved a waiver for informed consent on February 14th 2019 (MEC-2019-0090, Medical Ethics Committee, Erasmus MC University Hospital, P. Box 2040, 3000 CA, Rotterdam, the Netherlands), given the retrospective nature of this observational study. All consecutive patients who underwent either RAPD or OPD between January 1st 2017 and December 31st 2018 were included for analysis. RAPD has been performed in our center (tertiary referring hospital for pancreatic diseases) since January 2017. All procedures were undertaken by a dedicated team of two pancreatic surgeons or a pancreatic surgeon together with a pancreatic surgical fellow. RAPD was executed using the Da Vinci Model S robotic surgical device, which was later switched to the Model Xi (Intuitive Surgical Inc., Sunnyvale, CA, USA). OPD was performed by or under direct supervision of three experienced consultant pancreatic surgeons. Several consultant anesthesiologists were involved in both surgical modalities.
Patients were assigned to RAPD based on patient preference and availability of both the robot and the robotic surgical team, the sole exclusion criterion for RAPD was locally advanced pancreatic cancer.
RAPD was characterized as full-robotic surgery, meaning both the resection and the reconstruction phase were conducted in a robotic fashion. A central venous catheter was incidentally inserted based on the anesthesiologist's discretion. All patients were postoperatively admitted to a High Dependency Unit (HDU) or incidentally to an Intensive Care Unit (ICU). Protocols for postoperative management were identical for HDU and ICU. Standard postoperative analgesic regimen comprised paracetamol (1000mg 4 times daily) and naproxen (750mg 3 times daily). OPD was preferably performed under additional epidural analgesia (routinely using ropivacaine 0.2% combined with sufentanil 1 µg/ml). For RAPD a patient-controlled analgesia (PCA) device (morphine based) was used for postoperative analgesia. Once oral intake was possible again, epidural or PCA analgesia was if necessary converted to oral oxycodone.
Digital patient records were reviewed for patient demographics, intra-and postoperative management and postoperative morbidity. Extracted baseline data included age, sex, body mass index (BMI) and medical history (comprising diabetes mellitus, any pulmonary, cardiac or vascular disease, any history of cerebrovascular accident (CVA) or transient ischemic attack (TIA), hypertension or prior kidney or liver failure). Besides, records were reviewed regarding any history of previous malignancy, abdominal as well as non-abdominal surgery or neoadjuvant chemotherapy. Baseline comorbidity was graded according to the American Society of Anesthesiologist's (ASA) score and Charlson Comorbidity Index (CCI) [10].
Baseline hemoglobin (Hb), platelet count, estimated glomerular ltration rate (e-GFR) and levels of albumin, total bilirubin, creatinine and tumor-marker CA 19-9 were also extracted from patient records.
According to our center's protocol, norepinephrine (NE) was routinely used as rst choice perioperative vasopressor. Perioperative anesthesia records were screened for the NE dose on start and end of surgery as well as the amount of times NE dose perioperative exceeded 0.2 µg kg/min (considered the cut-off for the administration of NE over a central instead of peripheral intravenous cannula). The amount of times NE dosage exceeded 0.2 µg kg/min was expressed as time span in minutes and as percentage of operating room time (time between entering versus leaving the surgical theatre). The operative time was de ned as the time interval between skin incision and wound closure, time to detubation as the time interval between wound closure and removal of the endotracheal tube.
Fluid balances, including the necessity of erythrocyte transfusion, were studied up to 24 hours following surgery. Types of intravenous uid used were NaCl 0.9% (Baxter, Deer eld, United States of America) and Sterofundin ® (Braun, Melsungen, Germany) as crystalloids, Voluven ® and Volulyte ® (Hydroxyethylstarch 130/0.4, Fresenius Kabi, Bad Homburg vor der Höhe, Germany) as colloids. Besides, arterial blood gas (ABG) analyses (including blood pH, partial CO2 (pCO2) pressure, lactate and Hb count) were studied at 3 points in time: rst available sample results during surgery, rst available results upon HDU/ICU admission and rst available results after a minimum of 12 hour HDU/ICU admission. The perioperative pCO 2 levels were subsequently compared to the corresponding end tidal (et) CO 2 level.
Patient records were screened for the total hours of postoperative HDU/ICU stay and the rate of prolonged HDU/ICU admission, the latter de ned as a HDU/ICU stay exceeding 24 hours. The length of postoperative hospital stay was counted in days starting from the rst day following surgery. Postoperative morbidity was assessed 90 days after surgery using the Clavien Dindo (CD) morbidity score as well as the Comprehensive Complication Index [11,12]. Major morbidity was de ned as a CD score ≥ III. Kidney failure was graded according to European Society of Anesthesiologists' (ESA) Results 126 consecutive patients were included in this study during the two-year inclusion period (n=64 RAPD, n=62 OPD).
European Perioperative Clinical Outcome (EPCO) standards [13]. Pain scores, expressed as Numeric Rating Scale (NRS), were analyzed on postoperative days 1 and 3. Mortality rates were studied on postoperative days 30 and 90.
Baseline and perioperative variables were analyzed for RAPD versus OPD, RAPD patients suffering major postoperative morbidity (CD ≥ III) vs. absence of major postoperative morbidity (CD < III) and RAPD patients above versus below median intraoperative blood loss (high and low intraoperative blood loss, respectively). Normal distribution of data was assessed using a combination of visual inspection of histograms and Q-Q plots and the Shapiro-Wilk test. Numerical data were presented as mean (standard deviation, SD) or median (interquartile range, IQR), as appropriate. Categorical data were presented with frequencies and percentages. An independent sample T-test or Mann-Whitney U-test was performed in comparing numerical data, a χ2 or Fisher's exact test in categorical data. In order to identify independent predictors of major morbidity following RAPD (CD ≥ III), a logistic regression model was constructed using a backward stepwise approach. Baseline and perioperative variables showing a signi cant difference between minor (CD < III) versus major (CD ≥ III) morbidity after RAPD and of suspected clinical relevance in predicting major morbidity after RAPD were included in the regression model. Results herein were presented as odds ratio (OR) with corresponding 95% con dence interval (CI). Throughout the study two-tailed P-values of < 0.05 were considered statistically signi cant. Statistical analysis was carried out using IBM SPSS Statistics (version 24.0, Armonk, NY, USA; IBM Corp.).      Postoperative morbidity after RAPD: major (CD ≥ III) vs. non-major (CD < III) A higher rate of baseline hypertension was observed in the RAPD subgroup with major postoperative morbidity (17/28, 60.7% vs. 11/36, 30.6%, p=0.023, Table 1). Median intraoperative colloid administration and blood loss were higher in the RAPD group with major postoperative morbidity (500.0 vs. 0.0 ml, p=0.002 and 350.0 vs. 200.0 ml, p=0.047, respectively, Table 4, additional le 3). Average NE dose was higher at the end of surgery for the RAPD subgroup with major postoperative morbidity (0.09 vs. 0.04 µg kg/min, p=0.726). Upon HDU/ICU admission, lower arterial blood pH as well as higher lactate levels were observed in the RAPD subgroup with major postoperative morbidity (7.32 vs. 7.34, p=0.017 and 1.7 vs 1.3 mmol/l, in the RAPD subgroup without major postoperative morbidity, respectively). A similar trend was observed after a minimum of 12 hour HDU/ICU stay (7.37 vs. 7.39, p=0.016 and 1.4 vs. 0.9 mmol/l, respectively). Hospital stay was more than doubled in the patients with major postoperative morbidity compared to patients (18.0 vs. 7.0 days, p<0.001, Table 5). Within 90 days following surgery, n=2 RAPD patients deceased due to early recurrence of malignant disease (one in both RAPD subgroups).      Values are presented as number (proportion) or depending on normality distribution of cases as mean ± SD or median (interquartile range). X n where n represents the number of missing cases. BGA, Blood Gas Analysis; HDU, High Dependency Unit; ICU, Intensive Care Unit; NE, Norepinephrine; OPD, Open Pancreatoduodenectomy; RAPD, Robot Assisted Pancreatoduodenectomy.   mmol/l in the RAPD group of low intraoperative blood loss, p=0.008). Length of hospital stay was doubled for the RAPD group of high intraoperative blood loss (16.0 vs. 8.0 days in the RAPD group of low intraoperative blood loss, p=0.002, Table 5). A higher rate of postoperative morbidity was observed in the RAPD group with high intraoperative blood loss (Comprehensive Complication Index of 51.5 vs. 24.2 in the RAPD group with low intraoperative blood loss, p<0.001). Besides, a higher rate of major postoperative morbidity (CD ≥ III) was found in the RAPD group characterized by high intraoperative blood loss (20/33, 60.6% vs. 8/31, 25.8% in the RAPD group of low intraoperative blood loss, p=0.006).

Predictors of major morbidity following RAPD
After performing univariate logistic regression, variables independently associated with major morbidity (CD ≥ III) following RAPD were baseline medical history of hypertension (OR 3.51, 95% CI 1.

Discussion
RAPD is characterized by higher perioperative vasopressor administration as well as a higher intraoperative net uid balance, compared to OPD. However, colloid administration and erythrocyte transfusion were more often used in OPD compared to RAPD. Rates of major postoperative morbidity following surgery (CD ≥ III) were similar for RAPD and OPD. Patients who developed major morbidity (CD ≥ III) after RAPD required vasopressor administration in higher doses and more often received intraoperative colloids than those without major postoperative morbidity. Administration of colloids as well as increased postoperative lactate levels, were independently associated with major morbidity (CD ≥ III) following RAPD.
When interpreting outcomes following RAPD versus OPD, patient-related factors are more likely to affect postoperative morbidity compared to the surgical modality. We report a relatively small portion of patients marked ASA class III and higher (11/64, 17 outcome' after pancreatic surgery). In our study ASA class itself was not an individual predictor for postoperative major morbidity (CD ≥ III) as baseline hypertension was (OR 3.51, 95% CI 1.24-9.92). This nding, along with higher vasopressor demands in RAPD, implies as association between elemental hemodynamic and cardiovascular condition and postoperative outcome following RAPD. This observation illustrates the necessity of adequate perioperative cardiovascular risk management, e.g. suggesting lower postoperative morbidity following RAPD in case of optimal baseline antihypertensive therapy, invasive monitoring of circulatory status and prompt intervention in case of any deterioration. However, having a medical history of hypertension might include several other conditional factors posing an increased risk for developing (major) postoperative morbidity (e.g. increased BMI, vascular remodeling or pre-existing renal insu ciency) by itself.
The higher use of intraoperative vasopressors in RAPD versus OPD can be explained by differences in patient positioning (reversed-Trendelenburg in RAPD vs. supine in OPD) as well as exposure to pneumoperitoneum, increasing cardiac afterload and decreasing cardiac output [7][8][9]. A higher demand for vasopressor administration in RAPD was not necessarily re ected by worse baseline physical condition. Except for the fact that OPD patients more often received neoadjuvant chemotherapy and baseline hemoglobin levels were lower, no differences in baseline medical conditions could be demonstrated between RAPD and OPD patients. Although the intraoperative use of vasopressors was evident, we feel supported by recently published data that the routine insertion of a central venous catheter is not mandatory in neither RAPD nor OPD patients [19,20].
The 2018 RELIEF Study focused on postoperative outcomes after distinct intraoperative uid strategies during major abdominal surgery, differentiating between an either restrictive (median crystalloid + colloid 2177 ml) or liberal (median crystalloid + colloid 3500 ml) net intraoperative uid balance [5]. Whereas no differences were observed in general postoperative outcomes between both uid approaches, a more liberal intraoperative uid strategy was associated with lower rates of postoperative acute kidney failure (17/1439, 5.0% for liberal vs. 124/1443, 8.6% for restrictive, p<0.001). In our study, median intraoperative net uid balance was 2800 ml and 9/64 (14.5%) of RAPD patients suffered from postoperative acute kidney injury. Bannone et al. suggested an association between a more restrictive perioperative uid balance and an increased risk for developing postoperative pancreatic stula [21]. On the contrary, the 2019 meta-analysis by Garland et al. reported an OR of 0.54 (95% CI 0.31-0.94) for major morbidity following pancreatoduodenectomy surgery after adopting a more restrictive intraoperative uid strategy [22]. The optimal intraoperative uid regime in pancreatoduodenectomy remains point of debate: current insights are contradictory and prospective research in this area should distinguish between laparoscopic and open pancreatic surgery.
We observed an association between the intraoperative administration of colloids and development of Our study has several limitations. First the retrospective single-center study design covering a relatively high, but still limited number of procedures. Due to the limited number of surgeons, the surgical approach was very standardized. This is in contrast to the perioperative anesthetic care, which was provided by several consultant anesthesiologists who followed the available protocols with different levels of adherence. We consider this variability in the anesthesiological approach a mirror of clinical practice and an aspect which deserves special attention in possible prospective trials. Also, although locally advanced pancreatic disease was considered the only exclusion criteria for RAPD, study results could be biased by patient selection for RAPD either OPD surgery.

Conclusions
There are speci c differences and challenges regarding the perioperative anesthetic strategy between RAPD and OPD. RAPD is associated with higher levels of vasopressor administration during surgery and net perioperative uid balance is higher. Contrarily, the levels of colloid and erythrocyte transfusion are higher for OPD compared to RAPD. Baseline hypertension, perioperative colloid administration and Page 21/25 increased lactate levels after surgery are associated with higher rates of major morbidity (CD ≥ III) following RAPD. A more restrictive intraoperative uid regime has previously been shown to increase postoperative (nephrogenic) morbidity, present evidence is nonetheless contradictory. Current data is insu cient to make speci c recommendations to optimal perioperative anesthetic guidance in RAPD. However, perioperative hemodynamic management including preoperative optimization and intraoperative uid-and vasopressor-strategy are suggested to in uence postoperative morbidity and should be the focus of future prospective studies. -Consent for publication: not applicable.

Abbreviations
-Availability of data and materials: the datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
-Competing interests: the authors declare that they have no competing interests.
-Authors' contributions: All authors contributed to the study conception and design. Data collection and subsequent analysis were performed by AvdE. The rst draft of the manuscript was written by AvdE and