Skeletal Muscle Depletion and Nutrition Support Affected Postoperative Complications in Patients Underwent Pancreatoduodenectomy

Postoperative complications remain high in patients underwent pancreaticoduodenectomy (PD). Body composition has been shown closely related to the outcome in critical and surgical patients. The aim of the present study was to investigate whether preoperative skeletal muscle condition and postoperative nutrition would affect major complications in patients underwent PD. This retrospective study included 265 patients underwent PD with a CT scan of the abdomen within 1 week before surgery. Body compartment data was extracted from the L3 level of the CT scan, which included skeletal muscle area and density, subcutaneous and visceral fat area. Univariable and multivariable regression analyses were performed to investigate correlations between the above variables and postoperative complications. Furthermore, a subgroup analysis was conducted to explore the relationship between postoperative nutrition strategy and the outcome.


Introduction
Pancreaticoduodenectomy (PD) is an important surgical procedure for malignant and benign disease of pancreatic head and periampullary region. With the development of surgery and critical care, the mortality has dropped signi cantly in the past decades. However, the rate of postoperative complications remains high and hinders the recovery and long-term outcome. Although, several technical strategies have been proposed to reduce perioperative complications, their availability is still controversial. Therefore, assessment and identi cation of the high-risk patients have progressively gained attention.
Recently, the loss of skeletal muscle, or sarcopenia, was recognized as an independent factor affecting the prognosis in cancer patient [1]. Similarly, critical patients with more depletion of muscle mass tend to have longer ventilation and lower survival [2,3]. Patients undergoing major surgery, such as PD, constituted a large portion of intensive care unit (ICU) patients, and received more attention of research for the comprehensive premorbid data. Lately, preoperative visceral obesity and sarcopenia have been shown associated with major complications in patients with pancreatic ductal adenocarcinoma (PDAC) [4,5]. Meanwhile, several studies began to reveal the close relationship between sarcopenia and postoperative pancreatic stula (POPF) in patients receiving PD [6][7][8]. However, the de nition and diagnosis of sarcopenia varied greatly in these studies and body composition differed in various population [9,10]. So it is necessary to validate the predictive value of skeletal muscle depletion on postoperative complications in PD from a large cohort population.
Patients receiving PD are always admitted to ICU electively or in response to complications. In critical patients after major upper gastrointestinal surgery where early enteral nutrition (EN) is limited, the timing and caloric/protein target of parenteral nutrition (PN) remain to be a controversial issue [11,12]. Data from a meta-analysis showed a signi cant decrease of infectious morbidity with preoperative and early postoperative PN compared with standard therapy (STD) [13]. However, the bene cial effect appeared vanished if PN given only postoperatively and there could be even worse outcome with early nutrition [13]. A recent randomized control trial (RCT) demonstrated no difference in 60-day mortality whether critical patients (> 60% surgical admission) received early PN or STD [14]. These discrepancies may result from the heterogeneous critical patients with different baseline nutritional conditions or diseases. Routine nutrition risk assessment, such as NRS-2002 and NUTRIC may have limited power to identify those surgical patients requiring imperative nutrition support, since they mainly focus on weight alteration or general conditions. Given the pivotal role of skeletal muscle in nutritional assessment, we hypothesized skeletal muscle depletion might impact the e ciency of nutrition therapy in patients undergoing PD. Therefore, this study aimed to evaluate the effects of preoperative skeletal muscle condition and postoperative PN on major complications in patients underwent PD procedure.

Patients and data
This is a retrospective analysis of surgical patients underwent PD from January 2014 to December 2019 in a tertiary hospital (The First A liated Hospital of Soochow University). Patients were included if they were aged 18 years or older, underwent PD procedure, and had an preoperative abdominal CT scan made no more than 1 week before surgery. The excluded criteria were with preoperative organ failure, poor quality CT unable to analyze, and incomplete medical records.
Patient data included age, sex, weight, height, admission diagnosis, body mass index (BMI), NRS2002 scores and routine preoperative blood test. The primary outcome parameter of this research was major postoperative complications of Clavien-Dindo Classi cation (CDC) ≥ 3 [15]. Among these complications (CDC ≥ 3), POPF was graded according to the International Study Group for Pancreatic Fistula classi cation [16], and nosocomial infection included con rmed incision infection, abdominal infection and lung infection, which need speci c treatment, such as upgrading antibiotics, secondary closure and additional drainage. Second outcome included length of hospitalization, the postoperative time to discharge, and 28-day mortality. The average calories delivery (ACD) and average protein delivery (APD) were calculated from the total calories and protein patients received both EN and PN in the rst week after surgery. The initial parenteral nutrition time (IPNT) was de ned as the time interval from the end of the surgery to the beginning of PN in hours.
Surgical procedure and perioperative management All patients received Whipple or pylorus-preserving PD procedure conducted by experienced pancreatic surgeons. A two-layer duct to mucosa pancreato-jejunostomy with either Child or Roux-en-Y technique was used for reconstruction with at least two surgical drains in the abdomen, one next to the bilio-jejunal anastomosis and one closed to the pancreato-jejunal anastomosis. Somatostatin or its analogue was routinely administered for three to seven days. Enteral nutrition was initiated as early as possible and parenteral nutrition was used as a supplementary part according to the patients' condition. Other management included proton pump inhibitor, antibiotics, anti-coagulation and encouraged early mobilization.

CT scan analysis
The pre-operative abdominal CT test was performed no more than 1 week before surgery. Abdominal CT scans were analyzed in anonymized format by two blinded independent researcher trained in radiologic anatomy and body composition analysis. Brie y, a single slice of unenhanced CT scan with a 5-mm thickness at the third lumbar vertebra (L3) level was extracted from picture archiving and communication system (PACS), since both skeletal muscle area (r = 0.83-0.99; p < 0.01) and intermuscular adipose tissue (IMAT) (r = 0.39-0.61; p < 0.05) at this level are closely related to whole body skeletal muscle and IMAT volumes [17]. Skeletal muscles included into the analysis were the internal and external obliques, transversus abdominus, rectus abdominus, psoas, quadratus lumborum, and erector spinae muscles.
The tissue area and density was measured by ImageJ version 1.8.0 (National Institutes of Health). The skeletal muscle index (SMI) was calculated from L3 muscle area (cm) divided by height square (m 2 ).

Statistics
Categorical variables were compared by Fisher Exact-and χ2 tests. Continuous variables with a normal distribution were compared by t test, and Mann-Whitney U test for those non-normally distributed.
Receiver operating characteristic (ROC) curve analysis was used to de ne the cutoff values of body composition best t to predict complications (CDC ≥ 3) in female and male patients separately.
Logistic regression analysis was used to evaluate the association between body composition data, nutrition therapy and major complication. After univariable analysis, a multivariable analysis was performed adjusted for sex, age, and BMI. Any variable that generated a P-value of < 0.1 in univariable analysis was added to the multivariable analysis. Then the predicted probability of the nal multivariable model for complication was assessed by ROC curves. Furthermore, we conducted a subgroup logistic regression analysis to evaluate the effect of nutrition therapy on postoperative complications in patients with low and normal SMI respectively. Restricted cubic splines were t by a logistic regression model with the mean value as the reference standard. In both univariable and multivariable analysis, body composition data was used as categorical variables, with the thresholds from former ROC analysis, and the other numerical variables were conducted as continuous parameters. Computing, Vienna, Austria) were used for statistical analysis. Values with a normal distribution were reported as mean ± standard deviation (SD) and skewed data were expressed as median and 25-75% interquartile range (IQR). All statistical tests were two-sided. A p < 0.05 was considered statistically signi cant.

Patient characteristics
From all 458 patients underwent PD, 288 patients with available preoperative CT were included in the present study. Twenty-three patients were excluded due to preoperative organ failure or incomplete medical data. Finally, 265 patients (136 males and 129 females) were analyzed (Fig. 1). The mean age and BMI were 59.5 ± 13.9 years and 23.1 ± 3.6 kg/m 2 , respectively (Table 1). Systemic comorbidities, including hypertension, diabetes, were recorded in 107 (40.4%) patients. The top three etiologies for PD in our cohort patients were PDAC (125, 47.1%), ampullar tumor (78, 29.4%) and pancreatic cystic tumor (34, 12.8%). Postoperative management was based on the actual patient conditions. According to current nutritional consensus [19], enteral nutrition was considered rstly, PN was initiated if EN was not able to reach 70% of required calories. Therefore, the average initial parenteral nutrition time (IPNT) was 70.4 ± 13.7 hours after surgery. The average protein delivery (APD) was 1.12 ± 0.32 g/kg/d calculated during seven days postoperatively.

Body composition analysis
A sex-speci c diagnostic analysis for each CT-derived body composition on major complication (CDC ≥ 3) through ROC curve was performed to stratify patient with different risks (  The association between skeletal muscle and complications The detailed complications and secondary outcome were listed in Table 3, which was grouped by SMI and NRS2002. Eighty-one patients (30.6%) developed major complication (CDC ≥ 3). POPF and nosocomial infection were observed in 70 (26.4%) and 66 (24.9%) subjects, respectively. The average time span from surgery to discharge was 16.3 ± 7.2 days. The overall 28-day mortality was 1.13% (n = 3).   (Table 5). Based on this multivariable model, we tested its predictive performance on major complication. ROC curve shown the AUC was 0.776, with a sensitivity of 63.3% and a speci city of 83.3% in its optimal cut-point (Fig. 2). group. However, this result did not happen in patients with normal SMI. Furthermore, multivariable adjusted restricted cubic splines (RCS) for the association between IPNT, APD and major complications were shown in Fig. 3, with the mean values as the reference standard.

Discussion
Pancreaticoduodenectomy is one of the major procedures for periampullary malignancy and benign diseases. However, the high rate of postoperative complications hinders the rehabilitation and increases medical cost. Other than improving surgical techniques, identifying susceptible patients and re ning support strategy seems crucial. In the present study, we found preoperative skeletal muscle index and density as well as age were prognostic factors affecting postoperative complications. More importantly, through sub-group analysis, the IPNT and APD were independent determinants for complications in patients with lower SMI. These results may provide a practical strategy for distinguishing PD patients of high risk and tailored nutritional therapy.
Previously, Martin et al demonstrated skeletal muscle depletion was a powerful prognostic factor for survival in cancer patients [1]. In critical patients, muscle wasting was prevalent and profound. Up to 60-70% critical patients had muscle wasting at ICU admission [20], and 20% of skeletal muscle reduction was found in ICU patients within rst 10 days [21]. Both the quantity and quality of skeletal muscle affected the course and outcome of the disease [2,3]. In our present study, preoperative muscle depletion was found in 40.75% subjects who would undergo PD. The reported incidence rate of sarcopenia in PD patients varied a lot from 24.2% [22] to 75.5% [7,23], which may due to different population and cut-off value. Pancreatic and periampullary cancer patients were always malnourished at admission. It was necessary to have a thorough and accurate evaluation on the nutritional status. To date, CT was a widely used method for measuring skeletal muscle conditions due to its accuracy [23], and other approaches included bioelectrical impedance analysis and ultrasound. In our univariable analysis, VATA/SMA ratio appeared to be a signi cant factor affecting postoperative morbidity (RR 1.23, 95%CI 1.07-1.56, P = 0.012). This was in accordance with previous studies. David et al have shown high visceral adipose tissue was associated with increased infection rate (OR: 2.4) in pancreatic cancer patients. Moreover, longer survival was observed in patients with high muscle attenuation combined with low visceral adipose tissue (P = 0.011) [4]. In cancer patients underwent PD, Sandini et al had shown a high visceral adipose tissue-to-skeletal muscle ratio signi cantly increased major complication with OR reaching 3.20 [22]. From a prospective study with 284 patients undergoing PD, researchers demonstrated sarcopenic obesity was the only independent predictor for POPF (OR 2.65, 95% CI 1.43-4.93) [7]. In addition, females tend to have higher rate of sarcopenic obesity (BMI > 25) than males (38% vs. 12%) in patients with resectable PDAC and those patients showed higher incidence of major complications (P < 0.001) [5].
Adipose tissue does have some merits, such as energy storage, harm protection, warm keeping. Being a double-edge sword, fat may also play a bad character under some conditions. Excessive visceral fat could handicap the surgical operation, increasing the intraoperative risk. As an endocrine organ, unbalanced adipokines and cytokines may promote postoperative in ammation and develop insulin resistance [24].
In the multivariable regression, we found preoperative L3 level muscle index and density were predictive of postoperative complications. Here, higher muscle density was signi cantly related reduced complications (OR = 0.85, 95%CI 0.64-1.03, p = 0.029). The result was in line with other studies. Amini et al reported higher risk of morbidity and complication was observed in patients with sarcopenia after curative resection for pancreatic adenocarcinoma [25]. Recently, both Nicolas and Minji et al have demonstrated mean muscle attenuation could be a promising parameter to predict complications and POPF after PD [6,7]. In addition to short-term results, long-term outcome also seems to relate to skeletal muscle conditions. Elisabeth et al have shown sarcopenia negatively impacted overall survival in resectable PDCA patients (14 vs. 20 months) [5]. However, several studies demonstrated the amount of muscle mass was not predictive of major complication or morbidity after pancreatic surgery [22,26]. We considered the disagreement may relate to different methods, cut-off values and etiology.
Lower skeletal muscle area and density probably mean lower protein reserve and myosteatosis. In aging and tumor patients, the muscle depletion may be caused by insu cient intake and ensuing proteolysis, which occurred long before admission. Study has shown the serum albumin concentration correlated well with skeletal mass in critical patients [27]. Decreased albumin always indicated poor nutrition status and even leaded to tissue edema, which may hamper surgical procedure and healing. Reduced muscle density was a re ection of alteration in muscle composition within muscle bers, as uid surcharge and fat accumulation [28]. In several researches of oncology, including melanoma, adrenocortical carcinoma, renal cell carcinoma, muscle density was found to be an independent prognostic factor of survival [29][30][31]. An inverse association between muscle attenuation and triglyceride content has been established in healthy adults [32], and it was shown IMAT correlated with muscle density [4]. Myosteatosis may involve insulin resistance, in ammation, mitochondrial damage and oxidative stress [27]. These metabolic changes would further cause a fuel utilization shift from lipids to glucose in muscle, leading to muscle protein depletion and reduced capacity [33].
Of particular interesting, our subgroup analysis revealed that delayed initiation of supplementary parental nutrition (SPN) was associated with higher complications in low SMI patients undergoing PD (OR 1.89, 95% CI 1.43-2.49, p = 0.032). It was widely recognized that early enteral feeding was preferred in major surgery and critical patients [11,12,19]. However, enteral feeding always accompanied with abdominal distention, nausea, vomiting and underfeeding, especially among patients receiving major upper gastrointestinal surgery. According to ASPEN guideline, PN should be delayed for 5-7 days in the postoperative ICU patients, who may partly represent PD subjects [11]. However, the rationale resources it based on were largely from clinical studies conducted more than twenty years ago [13,34,35], which may differ from current PN pattern including preparation technology and hypocaloric feeding. Actually, in a recent multicenter RCT comparing EN and PN, no difference in mortality, infectious complication rate, and hospital length of stay was observed in critical patients [36]. In malnourished patients, PN resulted in a signi cantly lower mortality with a tendency towards lower rates of infection [37]. The timing of starting PN has long been debated in critical and surgical patients [11,12,19,38]. Our present study, to our knowledge, for the rst time demonstrated only in low SMI patients, early PN was bene cial for reduced complication. This result was in line with Heidegger's research, which found in malnourished patients and those with special risks SPN should be considered on day 4 after ICU admission [39]. Indeed, many studies have yielded negative results on early PN support in critical and surgical patients [38,40].
However, a recent Cochrane meta-analysis could not draw a clear conclusion as whether early or late SPN was better in critical patients due to low-quality evidence [41]. The heterogeneity and different baseline nutrition status may have impacted the results. ESPEN guideline suggested PN should be considered in high nutrition risk or malnourished patients [12,19]. But, no clear prescription was provided as a lack of integrating studies. Since skeletal muscle was a crucial indicator for nutrition status and affected patients outcome both in short and long term [2,3,21], we tested whether it could differentiate the e ciency of parental nutrition support in PD patients. The present results may help identify those patients who may bene t early PN support and provide some evidence for early SPN in sarcopenic patients.
Some mechanism researches have shown early PN may suppress muscle cell autophagy and associated with muscle weakness [42,43]. In addition, increased amount of adipose tissue within the muscle compartments was observed in critical patients receiving early PN [40]. However, these outcomes did not mean worse endpoints. Perhaps, it may just represent a metabolic process. Anyhow, early nutrition support and su cient protein supply can ensure necessary substrates in malnourished patients, even though it may not affect muscle wasting.
Here we also have shown higher protein supply was a protective factor for postoperative complication in low SMI patients (OR 0.76, 95%CI 0.53-0.89, p = 0.021). This result was supported by previous observational researches [44,45], which demonstrated higher protein delivery was associated with improved survival. Bene ts of high protein delivery were also found in RCT trials. Ferrie et al shown higher level of amino acids supply (1.1 g/kg vs o.9 g/kg) was associated with improved patients focused outcomes, such as handgrip strength and muscle mass, despite without differences in mortality [46]. Doig et al found higher intravenous amino acid therapy was related to improvement in renal function [47].
Ishibashi et al shown higher protein intake (> 1.5 g/kg/d) was associated with less total body protein loss [48]. Mechanistic studies have shown a 3-h high level (1 g/kg/d) amino acid infusion was able to improve protein balance from negative to positive in critical patients [49]. Sundstrom et al also reported a supplemental intravenous amino acid infusion sustained a positive protein balance for 24 h [50]. Current guidelines recommend adequate protein supply (≥ 1.2 or 1.3 g/kg/d) in critical patients, but few data exists as to major surgery patients [19]. Therefore, su cient protein supply appears bene cial in surgical patients, but high quality evidence is needed.
Our study did have some limitations that should be addressed. First, the patient population underwent PD procedure was due to various reasons from trauma, benign disease to cancer. This heterogeneity may reduce the quality of evidence to guide clinical case. Second, we only analyzed the relationship between preoperative skeletal muscle and short-term postoperative complications. The extent of muscle wasting after surgery was not documented, which may also in uence the short and long-term outcomes. Thus, dynamic courses of skeletal muscle conditions and its impact on prognosis as well as nutrition support should be examined in the future. Third, the patients who received late PN may have more severe conditions, such as hemodynamic instability, which may add bias to the analysis.

Conclusion
Low skeletal muscle index and density were independent risk factors for postoperative complications in PD patients. Therefore, preoperative body composition analysis by CT scan seems potential for the strati cation of complication risk. For patients with low SMI, early PN initiation and su cient protein supply appears to be a protective factor for postoperative complications in PD patients, but large RCT trials are required to validate this result.