Introduction

The World Health Organization projects that by 2015 more than 1.6 billion adults will be overweight and 700 million will be obese [1]. Obesity has been shown to be a risk factor for gastrointestinal cancer, and thus the effect of obesity on surgical outcomes in a cancer population is a timely issue [2–5]. Just like other sites of malignancy, an increased risk of gastric cancer has been noted in the obese population, probably caused by higher abdominal pressure and the resulting gastroesophageal reflux [6, 7]. Obese cancer patients have often been perceived as being at high risk of surgical complications. In fact, several technical disadvantages of gastrectomy for obese patients are expected, including poorer surgical visibility, blood oozing from soft tissues, a dissection plane hindered by adipose tissue, difficulty with anastomosis, and so forth. Several studies have demonstrated increased postoperative morbidity and mortality after gastrectomy with extended (≥D2) LN dissection in obese patients [8–13]. The effect of obesity on oncologic outcomes has been examined to a more limited extent. While some data have indicated obesity as an adverse factor impairing the oncologic adequacy of D2 dissection for gastric cancer [11], other studies have not noted similar results [14]. However, obesity has been defined very broadly in these studies and most authors have included small numbers of “really” obese patients. Significantly, little is known to date about early surgical outcomes of obese patients undergoing “over-D1” lymph node dissection. In the present study we aimed to assess the impact of body mass index (BMI) as a surrogate marker for obesity in a patient cohort that has undergone total gastrectomy with “over-D1” LN dissection (TG-D1+) for gastric adenocarcinoma. The objectives include analysis of perioperative and cancer-related outcomes. Our hypothesis, based on previous literature, was that obesity would have negative effects on perioperative and pathologic outcomes of patients with gastric cancer undergoing a potentially curative total gastrectomy.

Methods

A retrospective case-matched study was performed on 191 consecutive patients who underwent total gastrectomy for gastric cancer between January 2005 and September 2011 at the 2nd Division of General Surgery, Brescia Civic Hospital, Brescia, Italy. Exclusion criteria were as follows: surgery for gastric stump carcinoma and for tumors arising at or crossing the esophagogastric junction; preoperative chemotherapy; palliative resections; LN dissections different from “over-D1”; M1 gastric cancer patients; laparoscopy-assisted gastrectomies; and the presence of another malignancy. We have 161 patients left for the analysis. Patient weight and height were retrieved from anesthesia records, as charted on the day of surgery. BMI was defined as the patient’s weight in kilograms divided by height (in meters) squared. Patients were divided into three categories based on BMI: normal-weight (BMI < 25 kg/m2), overweight (BMI ≥ 25–<30 kg/m2), or obese (BMI ≥ 30 kg/m2), according to the World Health Organization definition [15]. A total of 63 (39.1 %) subjects were classified as normal-weight patients; 73 (45.3 %) were classified as overweight, and 25 (15.6 %) were classified as obese patients. These three groups were matched for the analysis of perioperative and cancer-related (oncologic) outcomes. All patients were preoperatively evaluated by esophagogastroduodenoscopy and abdominal computed tomography (CT). In patients suspected of having metastasis to distant organs, positron emission tomography (PET)/CT scan, chest CT, or magnetic resonance imaging was selectively performed. Anesthesia records were accessed to define the American Society of Anesthesiologists (ASA) score, the intraoperative estimated blood loss (EBL), and operative time. The ASA score was used to compare patient’s preoperative overall physical health [16]. Common co-morbidities included diabetes mellitus, hypertension, cerebrovascular disease, coronary artery disease, arrhythmias, pulmonary disease, and liver and renal dysfunction. The postoperative complications were categorized according to the Clavien-Dindo classification [17] (Appendix). Risk factors analyzed for postoperative complications included gender, age, ASA score, BMI, pTNM stage, long-term anti-platelet therapy, operative time, and splenectomy. Postoperative pathology reports were accessed to determine tumor location; differentiation, number, and status of resected LN, and T status. The final pathologic stage was classified according to the 7th edition of the AJCC Cancer Staging Manual [18]. As the AJCC Cancer Staging Manual recommends that a minimum of 16 lymph nodes be examined, the patients most likely to be under-staged (examined lymph nodes ≤15) were removed from the oncologic outcomes analysis.

Operative procedure

Total gastrectomy always included a formal “over-D1” lymphadenectomy (LN stations 1–8a of the Japanese classification), also called D1.5 or D1+ lymphadenectomy. Briefly, lymphadenectomy comprised en bloc removal of all lymphatic tissue in the left and right paracardial regions, in suprapyloric and infrapyloric sites, along the lesser and greater curvatures, along the left gastric artery, and along the common hepatic artery. Dissection of the splenic artery and the celiac trunk, as well as the hepatoduodenal ligament, was performed only in patients who had enlarged nodes at this area on preoperative or intraoperative staging. In cases of invasion of neighboring structures by the tumor, combined resection was employed, with removal on demand of adjacent organs or part of organs. The spleen was removed only if the tumor was close to or directly invading either the tail of the pancreas or the splenic hilum, and/or suspected LN involvement was evident. Advanced sealing devices were never used. Intestinal continuity was restored by means of mechanical standard Roux-en-Y esophagojejunostomy performed with a 25 mm circular stapling device in all patients. A naso-jejunal tube was inserted intraoperatively and was removed by the third postoperative day (POD 3) in most patients. One or more abdominal drains were routinely placed before abdominal fascia closure. All patients received prophylactic antibiotics (ceftriaxone 2–3 g/day) starting half an hour before the laparotomy and continuing for a mean of 72 h. Subcutaneous thromboembolic prophylaxis with low molecular weight heparin (LMWH) 3,000–6,000 IU/day was administrated to all patients. For patients on long-term aspirin therapy, LMWH replaced antiplatelet therapy for 1 week before operation. Before recommencing oral nutrition, the integrity of the esophagoenteral anastomosis was controlled with water-soluble contrast swallow X-ray 5–6 days after operation.

Evaluation of surgical outcomes

We considered as primary outcome measures the perioperative clinical course, the rates and severity of postoperative complications, and the length of hospital stay. Mortality was defined as lethal outcome during the operation or within the first 30 postoperative days. Postoperative complications, calculated for the same period, were classified as surgical or medical complications. Major bleeding was arbitrarily defined as any bleeding causing hemoglobin reduction ≥3 g/dl or requiring transfusion of ≥2 red blood cell units. The cancer-related parameters were considered secondary outcome measures. These outcomes were studied to assess whether the BMI of the patient was a determinant of safety and oncologic adequacy for total gastrectomy. In addition, preoperative variables in terms of the patients’ baseline characteristics, co-morbidity, and perioperative course were analyzed to determine whether they had predictive values as risk factors for postoperative complications.

Statistical analysis

One-way analysis of variance (ANOVA) was used to compare continuous variables with normal distribution (presented as mean and 95 % confidence interval [95 % CI] of the mean). The Mann–Whitney U test and Kruskal–Wallis one-way analysis of variance (ANOVA) were used to compare continuous variables not normally distributed (presented as median and interquartile range [IQR]). Normality of the distribution of variables was determined with the D’Agostino-Pearson test. The Chi square test or Fisher’s exact test, when appropriate, was used to compare categorical variables. Patients with grade II or greater complications based on the Clavien-Dindo classification [17] (Appendix) were defined as having postoperative morbidity. Correlation analysis between BMI and morbidity grade, operative time, intraoperative EBL, postoperative hospital stay, numbers of retrieved lymph nodes (LN), numbers of metastatic LN, and the ratio of number of metastatic LN to total LN resected (lymph-node ratio) was performed with Spearman’s rank test. Univariate logistic regression analysis was carried out to identify the risk factors for postoperative morbidity. The identified variables were subsequently entered into a multivariate logistic regression analysis in stepwise manner. Odds ratios (OR) and 95 % CI were calculated when indicated. For each analysis, the following potential predictive variables for postoperative complication were taken into account: age, gender, BMI, ASA score, splenectomy, operative time, long-term antiplatelet therapy, and TNM stage. A p value < 0.05 was considered statistically significant. All tests were two-sided. Statistical analysis was performed with statistical software for biomedical research (McCalc Software for Windows, version 10.2.0.0, Mariakerke, Belgium).

Results

Characteristics of patients in terms of demographics, preoperative co-morbidities, and preoperative laboratory values are shown in Table 1. Overall, the median age of the study population was 71 years (IQR 62–77 years), the male/female ratio was 93/68, and the mean BMI was 25.7 kg/m2 (95 % CI 25.2–26.3 kg/m2). In particular, the mean BMI of normal-weight patients was 22.6 kg/m2 (95 % CI 22.2–23.1 kg/m2); the mean BMI of overweight patients was 26.3 kg/m2 (95 % CI 26–26.6 kg/m2), and the mean BMI of obese patients was 31.9 kg/m2 (95 % CI 31.4–32.4 kg/m2). The BMI difference between normal-weight patients, overweight patients, and obese patients was statistically significant (p < 0.001). There were no significant differences in age, gender, preoperative hemoglobin, and antiplatelet therapy assumption among the three study groups. Overall the co-morbidity rate was similar in the three study groups, although patients in the overweight and obesity groups showed more severe dysfunctions. Furthermore, diabetes mellitus and arterial hypertension were significantly more frequent among overweight and obese subjects than in those of normal weight. This reflects the significant differences in ASA class (p = 0.025) between the three groups, with 60.3 and 68 % of overweight and obese patients, respectively, in class III–IV versus 38.1 % in the normal weight group. There were no differences between the three study groups in terms of tumor location, Lauren histological classification, and TNM stage (Table 1). As for surgical outcomes (Table 2), extended resections were performed in 38 cases, including two patients with resection of the colon, one with bilateral annessiectomy, one with distal pancreatectomy, and 36 with splenectomy (no significant inter-group differences).

Table 1 Baseline characteristics of the patients who underwent TG-D1+
Full size table
Table 2 Surgical outcomes, postoperative complications and cancer-related parameters after TG-D1+
Full size table

Perioperative morbidity and mortality

Operative time (OT) was longer for obese patients (median 180 min; IQR 149–190 min) compared with both the overweight patients (median 160 min; IQR 140–180 min) and the normal-weight patients (median 142 min; IQR 120–160 min); p = 0.0005. Obese patients were more likely to have a greater operative EBL (median 300 ml) compared with both the overweight patients (median 100 ml) and the normal-weight patients (median 0 ml); p < 0.0001. Similar results were obtained when BMI was modeled as a continuous variable rather than categorical variables: rho (ρ) for OT = 0.36, 95 % CI 0.22–0.5; p < 0.0001; ρ for EBL = 0.42; 95 % CI 0.29–0.54; p < 0.0001. Postoperative complications are listed in Table 2. Medical complications occurred in the overweight and obesity groups two to three times more frequently, respectively, than in the control group (p = 0.031). Pulmonary dysfunction was the most frequent complaint observed in the former groups, and surgical complications, including postoperative bleeding, had a higher incidence in overweight and obese subjects than among normal-weight patients (p = 0.034) (Table 2). Major bleeding was the most common surgical complication, and its incidence was higher in the overweight and obese groups (30.1 and 52 %, respectively) than in the normal-weight group (23.8 %); p = 0.035. However, bleeding did not require more blood transfusions or surgical revision in the inter-group analysis (p = n.s.) (Table 2). Similarly, wound infection was less frequently recorded among normal-weight patients than in the control groups (p = 0.013). Table 3 shows the postoperative complications recorded in the three study groups. Complications (Clavien classification) correlated closely with both the BMI and ASA grade: ρ for BMI = 0.29; 95 % CI 0.14–0.43; p = 0.0003; ρ for ASA = 0.45; 95 % CI 0.32–0.57; p < 0.0001. Although complications were more frequently observed in overweight and obese subjects, they were managed conservatively in most cases and did not require significant extension of hospital stay (grade I–II complications). Hence no significant difference was found when patients with no complication or grade I complications were compared with patients with grade II or greater complications in inter-group analysis (normal-weight versus overweight versus obese patients) (p = 0.648) (Table 3). In fact obese patients had similar median length of stay (11 days, IQR 9–13 days) to overweight (11 days, IQR 9–14 days) and normal-weight (10 days, IQR 9–12 days) patients; p = 0.112. Hospital stay was closely related to postoperative complications (Clavien classification) (ρ = 0.50; 95 % CI 0.38–0.61; p < 0.0001) but not to BMI (ρ = 0.056; 95 % CI −0.1 to 0.21; p = 0.49). The postoperative 30-day death rate was 0.9 % (one patient in the overweight group). The death occurred on POD 3 as a result of septic shock following anastomotic leakage.

Table 3 Postoperative complications according to the Clavien-Dindo classification [17] after TG-D1+
Full size table

Impact of BMI on oncologic outcomes

The median number of lymph nodes retrieved was 25 (IQR 19–36) per patient. There was a significant difference in the median number of LN harvested when stratified by BMI: obese, 21 (IQR 18–26) versus overweight, 24 (IQR 19–28) versus normal weight, 28 (IQR 24–38); p = 0.031 (Table 2). Overall, 12/161 (7.5 %) patients had ≤15 LN harvested. According to BMI class, the proportion was: normal weight, 4/63 (6.4 %); versus overweight, 6/73 (8.2 %); versus obese, 2/25 (8 %); p = 0.91. Some 70.2 % of patients enrolled in the study had LN metastases (N+). Specifically, the percentage of patients with N + was similar in each BMI category (obese, 64 % versus overweight, 72.6 % versus normal weight, 70 %; p = 0.713) (Table 1). BMI was not associated with a higher incidence of LN with metastatic involvement, as the median number of metastatic LN was 2–3 per patient for each BMI category (p = 0.567) (Table 2). The ratio of the number of metastatic LN to total LN resected per patient (lymph node ratio) was also comparable (p = 0.646) (Table 2). Similar results were obtained when BMI was modeled as a continuous variable rather than as a categorical variable: ρ for number of LN retrieved = 0.2; 95 % CI 0.1–0.3; p = 0.003; ρ for number of metastatic LN = −0.007; 95 % CI −0.16–0.15; p = 0.93; ρ for LN ratio = −0.003, 95 % CI −0.16–0.15; p = 0.97.

Predictors of postoperative complications

Patients with grade II or greater complications were considered positive for postoperative morbidity. At univariate analysis, the patient’s age, BMI, ASA score, antiplatelet therapy assumption, and splenectomy were found to be predictive of postoperative complications. Conversely gender, operative time, and TNM stage were not significantly correlated to postoperative 30-day morbidity or mortality (Table 4). At multivariate analysis, only ASA score and splenectomy were confirmed as independent risk factors for postoperative complications (Table 4). In particular, ASA score was closely associated to both medical (ρ = 0.34; 95 % CI 0.2–0.47; p < 0.0001) and surgical postoperative complications (ρ = 0.42; 95 % CI 0.28–0.54; p < 0.0001). No significant difference was found between the respective areas under the curve (AUC) with receiver operating characteristic (ROC) curve analysis (difference between AUC = −0.022; p = 0.765).

Table 4 Univariate and multivariate logistic regression analyses of potential risk factors for complications (classified according to the Clavien-Dindo classification [17]) after TG-D1+
Full size table

Discussion

Several large randomized controlled trials in Europe have reported increased postoperative morbidity after gastrectomy with ≥D2 LN dissection when compared with less extended lymphadenectomy [19–24]. The greater number of obese patients in European countries than in Asian countries may have an impact on this unfavorable outcome. BMI is a widely accepted indicator of obesity. Higher BMI is related to prolonged operating time, increased intraoperative bleeding, and a decreased number of dissected LN. It is also associated with postoperative morbidity, postoperative hospital death, and poor long-term survival after D2 dissection for gastric cancer [8, 11–13, 25]. Therefore the effects of splenectomy or pancreatectomy during D2 dissection in obese patients must be carefully considered [21–24]. It is worth noting that, to date, there are poor data on the effect of BMI on early surgical outcomes and LN retrieval after open TG-D1+. To avoid subjectivity and imprecision in complication reporting, we adopted the Clavien-Dindo classification [17]. It has been increasingly used in the surgical literature and in studies addressing the role of obesity in gastric cancer surgery [26]. First of all, our study revealed a significant difference in the way of presenting symptoms and ASA grade across the BMI spectrum. In fact, obesity turned out to be linked to a greater risk of other co-morbidities, mainly diabetes mellitus and arterial hypertension, both of which significantly affect the ASA grading. We observed greater operative blood loss, higher incidence of pulmonary dysfunction and wound infection in high BMI classes, most notably in the morbidly obese group. However, the majority of such perioperative complications were minor and required either no therapy at all or a simple routine intervention (grade I–II complications). Thus we did not observe a significant increase in complication rate and severity, duration of in-hospital stay, or operative mortality across the BMI spectrum. This is in contrast with reports after D2 dissection, which however dealt with different classification scales [9, 12, 26–28]. Results from our logistic regression analysis are relevant. On univariate analysis, BMI along with the patient’s age, ASA score, antiplatelet therapy assumption, and splenectomy turned out to be predictive of postoperative complications, whereas on multivariate analysis BMI lost statistical significance and only ASA score and splenectomy were confirmed as independent risk factors. Data from other institutions also showed that BMI was not an independent predictor of postoperative complications, but rather a confounding factor when associated with higher ASA score [25, 28]. As expected, the operative time increased with increasing weight and was especially evident in patients with morbid obesity. Similar conclusions have been drawn after D2 LN dissection [9, 12, 25–29]. It is clear that morbid obesity may increase the technical challenge of total gastrectomy even when coupled with “over-D1” lymphadenectomy. We did not find any association between BMI (or morbid obesity) and any of the cancer-related parameters despite a detailed analysis of factors, including tumor site, TNM stage, Lauren classification, lymph node status, and lymph node ratio. This observation seems to be in contrast to increasing epidemiologic evidence of a link between gastric cancer and obesity. Nevertheless, findings similar to those from our study have been reported [29]. Few researchers have assessed whether LN harvest is compromised by obesity. In the present study the increased individual BMI significantly correlated with the decreased number of retrieved LN, indicating that excessive intra-abdominal visceral fat precludes the complete dissection of LN. Furthermore, during the handling of a specimen the isolation of LN from the retrieved soft tissue with abundant fat might be more difficult than retrieval in non-obese patients. Data in the literature are contradictory. Some authors have reported a lower LN yield in obese patients after D2 dissection [11, 25, 29, 30], whereas other studies have failed to corroborate these findings [14, 31]. Our median number of LN retrieved (25) was in line with reports in literature of open TG-D1+, which range from 15 to 30 LN harvested [32, 33]. Also, in our series, 12/161 patients (7.5 %) had ≤15 LN harvested, and this figure is consistent with large published series [19, 23, 34, 35]. Patients with BMI ≥ 25 kg/m2 were more likely to be under-staged, although not statistically. It is of note that the difference in the prevalence and degree of “overweight” patients between Far East and Western countries makes data from Asian institutions hard to compare with studies like this one or others from European or US hospitals. In addition, most Asian patients classified as obese should be classified as overweight for comparison with Western studies [25, 27, 29, 30]. In fact, in Europe and the USA obesity is defined as BMI > 30 kg/m2, which accounts for 20–25 % of these populations [36, 37]. This percentage is significantly higher than the 15 and 1.0 % prevalence of patients having BMI ≥ 25 and BMI > 30 kg/m2, respectively, as reported in a large randomized Japanese study [9]. Significantly, about 45 % of patients enrolled in the present study were overweight, and 15 % were obese. In addition, the mean BMI of our study population was almost 27 kg/m2, which is far higher than values reported in Korean or Japanese studies (range 22–23.5 kg/m2) [25, 26, 30, 31, 38, 39]. This result agrees with the difference in BMI found in a comparative study between Japanese and US patients undergoing surgery for gastric carcinoma [40]. It is noteworthy that the overall incidence of esophagojejunal anastomosis leakage was 1.9 % in the present study, comparable with the lower anastomotic leak rate cited by centers with adequate case loads, which ranges between 1.3 and 15.9 % [41–44]. Meanwhile, the relatively high incidence of duodenal stump dehiscence (6.2 %) needs to be stressed. This finding should be considered as a consequence of a not routinely hand-sewn duodenal stump until 2010. After an internal audit of our achievements to date, we decided to carry out hand sewing of the duodenal stump on a routine basis. Since that decision was taken, no similar complications have been reported.

In summary, our analysis indicates that obesity per se is not associated with an increased perioperative morbidity or mortality after TG-D1+, although obese patients had longer operative time, higher risk of perioperative blood loss, and a higher incidence of wound infection. In contrast, ASA grade was closely associated with postoperative complications. In particular, the presence of a medical co-morbidity, as reflected by ASA class, increased the risk of both surgical and medical complications. This finding must be carefully considered when dealing with obese patients, who usually have a high ASA score: >60 % of our overweight and obese patients were in class III–IV compared with <40 % in the normal-weight group. In the present study, obesity affected operative oncologic outcomes. In particular, the number of lymph nodes retrieved was lower in obese subjects, although lymph node status and lymph node ratio were comparable in obese and non-obese patients.

We are aware that the surgeon factor might be a bias for this study. However, most of the procedures were performed by the chief surgeon (R.F.), and the rest were performed by only the most experienced members of the surgical team (all of them consultant surgeons). Also, the composition of the surgical team remained the same throughout the entire study period. In addition, the short study period was chosen to avoid all possible biases linked to the rotation of the team members and it covers the phase of professional “maturity” of the whole surgical team, corresponding to the “plateau” of the learning curve. We therefore believe that the impact of this variable on the outcomes, if present, is minimal.

In conclusion, this study showed that in experienced units and in the absence of severe ASA score, obese patients should not be denied TG-D1+ based on BMI only. That said, a word of caution is needed about the greater risk of perioperative blood loss, which could lead to potentially serious complications.