Spectrum and Risk Factors of Complications After Gastric Bypass

Objective  To study the spectrum of and risk factors for complications after gastric bypass (GBP).

Design  Prospective cohort study.

Setting  Academic tertiary referral center.

Patients  All morbidly obese patients who underwent open or laparoscopic GBP between January 2003 and December 2006.

Main Outcome Measures  Complications were stratified by grade: grade I, only bedside procedure; grade II, therapeutic intervention but without lasting disability; grade III, irreversible deficits; and grade IV, death. Data were analyzed using logistic regression to identify independent risk factors of complications after GBP. Predictors investigated were age, race, sex, marital and insurance status, body mass index, obesity-associated comorbidities, American Society of Anesthesiologists Physical Status Class, operating room time, open or laparoscopic approach, and surgeon experience.

Results  Of the 404 morbidly obese patients who underwent consecutive open (n = 72) or laparoscopic (n = 332) GBP, 74 (18.3%) experienced 107 complications. Grade I and II complications were more frequent after open GBP (grade I, 19.4% after open vs 3.9% after laparoscopic operations, P < .001; grade II, 20.8% after open vs 8.4% after laparoscopic operations, P < .001), and 55% were wound related. Grades III and IV complications occurred in only 4 patients (1%), and frequency was similar for open and laparoscopic cases. Three factors were independently predictive of complications: diabetes mellitus (odds ratio [OR], 1.9; 95% confidence interval [CI], 1.1-3.3; P = .02), early surgeon experience (OR, 2.5; 95% CI, 1.4-4.2; P = .001), and open approach (OR, 3.9; 95% CI, 2.1-7.3; P < .001).

Conclusions  Complications occurred in 18.3% of patients, but 95% were treated without leading to lasting disability. Presence of diabetes, early surgeon experience, and an open approach are risk factors of complications.

The exponential increase in the number of bariatric operations performed in the past 15 years is because of 5 factors: the obesity epidemic, the recognition of obesity as a health hazard, the poor results with nonsurgical methods, the reproducible good results with surgical techniques, and the introduction of laparoscopic techniques.¹ In the United States, the most common operation offered to treat morbid obesity is gastric bypass (GBP),² which has postoperative complication rates between 10% and 25% and a mortality rate of 0.3%.^3-6 A better understanding of the severity or spectrum of complications and the predictors of complications after GBP can be used to develop strategies to prevent complications, thereby improving overall results.

We sought to delineate the frequency and spectrum of complications after open and laparoscopic GBP in a large cohort of patients treated at a university tertiary referral center and to determine the independent factors associated with complications. A literature review of other large series reporting complications and proposed strategies for prevention, early identification, and treatment are discussed.

The study population consisted of all morbidly obese patients who underwent consecutive primary open or laparoscopic GBP, performed by 1 of the 3 attending surgeons (G.M.C., S.J.R., or A.M.P.) between January 2003 and December 2006 in the bariatric surgery program at the University of California San Francisco Medical Center. All surgeons started their bariatric academic practice at the medical center and had completed fellowship training that included advanced laparoscopic surgery and bariatric surgery. Preoperative, perioperative, and postoperative data were collected prospectively and entered into a database. The medical records of all patients were systematically reviewed to search for possible missed complications. Patients with previous bariatric operations were excluded from this analysis. All patients met the 1991 National Institutes of Health Consensus Development Conference⁷ and University of California San Francisco Bariatric Surgery Program criteria for bariatric operation: body mass index (BMI) (calculated as weight in kilograms divided by height in meters squared) greater than 40 or BMI between 35 and 40 with high-risk obesity-related comorbid conditions for at least 5 years, a documented attempt of medically supervised therapy for weight loss for at least 6 months, and participation in a 1-hour educational and screening session with a dietician and a bariatric surgeon. All patients also underwent preoperative psychological, nutritional, and comprehensive medical evaluations. Informed consent was obtained from all patients. The study was conducted after institutional review board approval.

Open or laparoscopic approaches were used at the discretion of the attending surgeon and the choice was based on the surgeon's clinical assessment and judgment of the likelihood and safety of performing the operation with each approach. The gastrojejunostomy suturing technique was also used at the discretion of the attending surgeon. When conversion from a laparoscopic to an open approach was needed, the case was counted as an open case. Gastric bypass was performed laparoscopically with 6 to 7 ports. Apart from the incision access, the GBP technique was similar for open and laparoscopic approaches. A 3.5-mm linear stapler (United States Surgical Corporation, Norwalk, Connecticut) transected the stomach to create a 15- to 30-mL gastric pouch. A retrocolic gastrojejunostomy was performed in the first 76 cases and an antecolic gastrojejunostomy route was used for the rest. A hand-sewn anastomosis was used in 17 patients (4.2%), a circular anastomosis with a 21-mm stapler (Ethicon Endosurgery, Cincinnati, Ohio) was used in 14 (3.5%), and a 25-mm stapler was used in 166 patients (41.1%). Lastly, a 1.5-cm gastrojejunostomy was created using a 3.5-mm linear stapler in 207 patients (51.2%). Orogastric methylene blue was used routinely to test for anastomotic leaks. A Roux limb of 100 (n = 342 [84.7%]) or 150 cm (n = 62 [15.3%]) was measured and a completely stapled side-to-side jejunojejunostomy was created. The mesenteric defect at the jejunojejunostomy was routinely closed. A 10 French Blake drain was placed close to the gastrojejunostomy.

Routine upper gastrointestinal swallow studies were obtained on the first postoperative day for the first 299 patients and a selective approach was used afterwards.⁸ The patient was given only clear liquids on the first operative day and then was advanced to a full liquid diet on the second postoperative day. Patients were seen postoperatively at 2 weeks, every 3 months for the first year, and annually thereafter.

All complications occurring in the 90-day postoperative period and to the last patient follow-up were included in this analysis. To identify the spectrum of complications, we stratified them into 4 grades based on the 1992 classification system proposed by Clavien et al⁹ and used by Parikh et al.¹⁰ Grade I complications were defined as events carrying “minor risks”; as not life threatening; as not requiring use of drugs other than analgesics, antireflux agents, antipyretics, antiemetics, antidiarrheals, or drugs required for urinary retention or low urinary tract infections; as requiring only interventions that can be performed at the bedside; and as never being associated with hospitals stay greater than twice the median stay for the procedure. Grade II complications were subdivided into grade IIA complications, events requiring use of drug therapy, total parenteral nutrition or blood transfusions, or a hospital stay greater than twice the median stay; and grade IIB complications, events requiring therapeutic intervention, such as imaging procedures, therapeutic endoscopy, or reoperation (not requiring organ resection or anastomotic revision), without lasting disability. Grade III complications were those resulting in organ resection or irreversible deficits or lasting disability. Grade IV was death as a result of a complication. If a patient had more than 1 unrelated complication, the highest-grade complication was used to calculate the percentage of patients who incurred a complication. Unrelated complications were included in the total number of complications. If a complication required repeated interventions, it was counted as a single complication. If a complication of a lower grade evolved into a higher-grade complication, only the highest grade was used for analysis. Complications were distinguished from sequelae of the primary operation, which consisted of diarrhea, bloating, and gallstone formation. Postsurgical biliary colic requiring cholecystectomy was not considered a complication.

The numerical independent variables studied at the univariate level were age, BMI, and operating room time (the total time spent in the operating room suite, including induction of and recovery from anesthesia and surgery time). To simplify clinical interpretation and for optimal use of the logistic regression method, the potentially predictive variables age, BMI, and operating room time were converted to categorical data. For age, patients were studied in 2 groups (< 60 years vs ≥ 60 years). For BMI and operating room time, patients were studied in 3 groups determined by distribution in the cohort in terciles (BMI of 35-44.4, 44.5-51.9, and 52-103; operating room time of 105-250 minutes, 251-295 minutes, and 296-647 minutes).

The categorical variables studied at the univariate and multivariate level were sex, race (white, Hispanic, black, or other), marital status (married vs single), insurance status (private vs state carrier), approach for the procedure (open or laparoscopic GBP), and presence or absence of these obesity-associated comorbidities: diabetes mellitus, hypertension, degenerative joint disease, obstructive sleep apnea, hyperlipidemia, and psychiatric diseases. Additional categorical variables consisted of the noted American Society of Anesthesiologists Physical Status Class (the classification system adopted by the American Society of Anesthesiologists for assessing preoperative physical status), the operating room time, and the first 50 cases from each of the 3 surgeons that were added together to form the 150 cases in the early experience group, which were then compared with the sum of the 3 surgeons' subsequent cases (51st case and beyond).

A univariate analysis assessed the unadjusted association of each variable on the risk of complications. Continuous variables were studied using their numerical values and previously defined categorical variables. The χ² test or Fisher exact test were used to compare proportions between groups, and the Mann-Whitney test was used to compare the distribution of continuous variables between individual groups. Statistical significance was considered to be P < .05.

Forward stepwise logistic regression was performed to assess the joint effect of all potentially predictive variables and to define those that were independently associated with the presence of complications. To stay in the model, variables were required to be significant at P < .05. Significant variables were then analyzed in a multivariate model to obtain the predictive effect of each, adjusted for the presence of the other significant variables (adjusted odds ratio). SPSS, version 13.0.1 (SPSS Inc, Chicago, Illinois), was used for all statistical analyses.

Four hundred four morbidly obese patients (83.2% women; median age, 44 years; median BMI, 48 [range, 35-103]) underwent consecutive open (17.8%) or laparoscopic (82.2%) GBP operations (Table 1). There were 11 conversions from laparoscopic to open GBP (3.2%). The reasons for conversions were extensive intra-abdominal adhesions in 5 patients, enlarged liver with difficult exposure in 3, stapler failure and need for anastomosis revision in 2, and inability to maintain pneumoperitoneum in 1. Median operating room time was 273 minutes (interquartile range [IQR], 240-310 minutes) and was shorter for the laparoscopic cases than for open cases (266 minutes [IQR, 235-300] for laparoscopic vs 322 minutes [IQR, 277-390] for open GBP; P < .001). Median length of hospital stay was 4 days (range, 1-47 days) and was shorter for the laparoscopic cases (3 days for laparoscopic vs 6 days for open GBP; P < .001). Patients who underwent open GBP had higher BMIs (57.4 [IQR, 49.6-67.0] for open cases vs 46.4 [IQR, 42.4-52.4] for laparoscopic cases; P < .001) and were likelier to be men (31.9% in open cases vs 13.6% in laparoscopic cases; P < .001). No other demographic or clinical characteristics differed significantly. All patients completed a minimum 3 months of follow-up (mean, 11.9 months [range, 3-48]). Weight loss and follow-up data at 12 months were available in 82% of the 249 patients with a minimum of 12 months after the operation. Mean excess weight loss at 12 months was 64% and was similar for patients who had laparoscopic (65%) or open GBP (59%; P = .14).

The overall complication rate for GBP was 18.3% (107 complications in 74 patients). Complications were significantly more common in patients with diabetes mellitus (P = .02), longer operating room time (P = .01), those who had open GBP (P < .001), and those whose operation took place early in a surgeon's experience (P < .001) (Table 1). The median BMI for patients with complications was greater than that for patients without a complication (Table 1).

Grade I and II complications composed 94.5% of all complications, were more frequent after open GBP, and most were wound related (Table 2). Additional analysis of grade II complications showed that grade IIA complications had higher prevalence after open GBP (1.5% for laparoscopic cases vs 12.5% for open cases; P < .001), while grade IIB complications with the need for invasive intervention occurred at similar rates (6.9% for laparoscopic vs 8.3% for open cases; P = .62). Grade III and IV complications occurred in only 4 patients (1%), and frequency was similar for open and laparoscopic cases.

Three factors were independently predictive of complications after GBP: clinical diagnosis of diabetes mellitus (odds ratio, 1.9; 95% confidence interval, 1.1-3.3; P = .02), early surgeon experience (odds ratio, 2.5; 95% confidence interval, 1.4-4.2; P = .001), and open approach (odds ratio, 3.9; 95% confidence interval, 2.1-7.3; P < .001).

Gastric bypass is the most common bariatric operation performed in the United States and successfully provides significant and long-term weight loss,¹¹ improvements in quality of life,^12,13 resolution of obesity-associated comorbidities,¹⁴ and likely extension of life span.¹⁵ Complications after other surgical procedures increase costs¹⁶ and patient disability,¹⁷ have an adverse impact on long-term survival,¹⁸ and may be expected after GBP as well. Consequently, prevention and proper treatment of complications should be continuously pursued, a task that requires a clear understanding of the spectrum of complications that occur after GBP as well as their risk factors.

Our study initially delineated the spectrum and severity of complications after GBP. The most common complication was wound infection or seroma (5.9%), followed by stricture of the gastrojejunostomy (3.5%) and bleeding (2%). Grade I and II complications occurred in 17.3% of patients and occurred more frequently after open procedures. Grade III and IV complications occurred in only 1%, and frequency was similar for open and laparoscopic cases. Finally, we showed that 95% of patients with complications after GBP had no detectable long-term disability after treatment.

Our overall complication rate of 18.3% is consistent with previous reports that describe a range between 15% and 20%.^3-6 Other studies have shown complications to be associated with a variety of factors, including access (laparoscopic vs open),^6,19-22 greater BMI,^23,24 older age,⁵ early surgeon experience,^20,22,25 hospital volume,³ operative time,^6,26 and obesity-related comorbidities.^21,27

In our series, the rate of wound-related complications was significantly lower after laparoscopic GBP than open GBP (2.7% vs 22.2%, P < .01). Wound-related complications were responsible for more than 59% of all complications in patients that had open GBP, similar to the other published series.^6,22

Gastrojejunostomy strictures following GBP have been reported in 3% to 27% of patients.^28,29 The frequency of strictures was significantly higher after a 21-mm circular stapler was used (23.1%) and similarly low for the linear stapler (4.1%) and 25-mm circular (2.6%) techniques. All patients were successfully treated with outpatient endoscopic balloon dilatation (60%, 1 session only) without further complications.

In a systematic literature review of 3464 cases, the frequency of postoperative bleeding in the gastrointestinal tract was higher after laparoscopic GBP than after open GBP (1.93% vs 0.6%, respectively, P = .008).²² However, the incidence of gastrointestinal bleeding after both laparoscopic and open GBP in other series appears similarly low.³⁰ In our series, the rate of bleeding complications was similar for laparoscopic and open cases (2.1% vs 1.4%, respectively, P = .69), but this most likely represents a sample size effect, because all but one bleeding complication occurred in laparoscopic cases. The main potential sources of bleeding are intraluminal (the gastric remnant, gastrojejunostomy, and jejunojejunostomy staple lines) and intraperitoneal (mesenteric staple line and dissection planes). Whereas bleeding from the gastrojejunostomy may manifest as hematemesis and permit endoscopic intervention, bleeding from the other sources may pose a diagnostic and treatment challenge. The decision to transfuse, use therapeutic endoscopy, or reoperate is based on a combination of clinical factors and local experience. A recently published randomized trial found that staple line reinforcement was associated with less bleeding from staple lines.³¹

Small-bowel obstruction after laparoscopic GBP has been reported in 1.5% to 5% of patients and can be caused by adhesions, internal herniation through the mesenteric defects, and obstruction at the jejunojejunostomy owing to narrowing or malpositioning (a “kink”) of the biliopancreatic and Roux limbs, among other less common causes.^32,33 The incidence of small-bowel obstruction is reportedly higher when a retrocolic Roux limb is used. The antecolic Roux limb eliminates the most common site for obstructions: the transverse mesocolon.³³ In our series, only 5 patients (1.2%) experienced small-bowel obstruction, but 4 required an operative repair. Because radiologic examination with computed tomography, upper gastrointestinal series, or both may show false-negative results,^33,34 early surgical intervention should be considered in patients with sudden or severe symptoms to prevent the formation of gangrenous bowel.

Gastrogastric fistulae, which have been reported in up to 6% of patients after GBP, are usually technical complications resulting from the incomplete division of the stomach during the creation of the gastric pouch.⁴ In our study, gastrogastric fistulae developed in 3 patients (0.7% of all patients). If the fistula is related to incomplete division of the stomach during the creation of the gastric pouch, we recommend laparoscopic division because a gastrogastric fistula is associated with long-term failure and weight regain and also permits gastric acid secretions to flow back into the gastric pouch and gain contact with the Roux limb, thus leading to the possibility of stomal ulcerations and perforations.

One of the most feared complications after GBP is leakage from the gastrojejunostomy. Many centers use routine postoperative radiological testing of the gastrojejunostomy via an upper gastrointestinal series. We have found this approach unreliable⁸ and prefer selective use of computed tomographic scanning with oral and intravenous contrast in the small subset of patients who have clinical manifestations of gastrojejunostomy leaks, such as tachycardia and/or a high white blood cell count. Only 1 gastrojejunostomy leak was found in our cohort (0.2%) and was treated with laparoscopic repair and drainage.

Gluteal and lumbar rhabdomyolysis occurs in a small subset of patients; is associated with high BMI, central obesity, and prolonged operations^35,36; and can lead to acute renal failure³⁷ and death.^35,38 The one patient in our study who had gluteal rhabdomyolysis developed acute renal failure, which required short-term dialysis. We advocate routine measuring of creatine phosphokinase levels after operations in patients with BMIs higher than 50 and those with central obesity.³⁶ Early detection may permit vigorous intravenous hydration and urine alkalinization to be used to prevent further complications, such as acute renal failure.³⁸

Others have reported the unusual complication of acute portal vein thrombosis after GBP³⁹ without operative injury to the portal venous system, which developed in one of our patients. This complication has also been reported after other laparoscopic procedures, such as Nissen fundoplication and cholecystectomy.^40,41 The etiology of acute portal vein thrombosis is not well understood but is likely related to an undiagnosed hypercoagulable state and insufflation of the pneumoperitoneum, leading to a low flow in the portal venous system and thrombosis. Patients usually present with abdominal pain; a computed tomographic scan is usually diagnostic. The treatment for acute portal vein thrombosis should be individualized on the basis of extent of the venous thrombosis and whether bowel ischemia is present. Immediate anticoagulation with or without arterial or venous (transhepatic or transjugular) thrombolysis and thrombectomy has been used with variable results.^39,42-44

We found that 3 factors were independently predictive of complications after GBP: clinical diagnosis of diabetes mellitus, open approach, and surgeon experience. In patients with diabetes undergoing an operation, morbidity is increased as a result of impaired myocardial and vascular function, a higher incidence of wound infection, an increased likelihood of postoperative renal failure, among other factors.^45,46 Routine measurement of intraoperative blood glucose levels and appropriate insulin administration has been an integral part of our postoperative care routine and should be standard practice.^47,48

Laparoscopic approaches are generally thought to be more technically challenging than open procedures, and the incidence and types of complications differ between the two.²² However, laparoscopic GBP has clearly demonstrated advantages, including shorter operative time and hospital stay, less postoperative pain, and a faster convalescence.^33,49 A limitation of our study is baseline differences (BMI and sex) between patients who had open or laparoscopic approaches. However, sex was clearly not associated with complications and is thus unlikely to confound the association we detected with surgical approach; BMI was included as a potential predictor for complications in the logistic regression model and was not selected as an independent predictor. Long-term (12-month) weight loss was similar for both approaches.

Surgeon experience and hospital volume are reportedly important variables that influence the frequency of complications in laparoscopic approaches.^21,50,51 Although the learning curve for laparoscopic bariatric operations may consist of more than 100 cases,^5,51 our complication rate was significantly lower and remained low after 50 cases for all 3 surgeons. The hospital volume and outcome relationship has been previously described as enhancing outcomes after GBP.^52,53 High-volume hospitals (> 13 000 total operative cases, 157 bariatric cases/y) are reportedly associated with a lower overall complication rate than low-volume hospitals (< 3000 cases, 15 cases/y).³ This may reflect the presence of appropriate structural characteristics and formalized processes of care in high-volume hospitals, such as appropriate beds and operating room tables, availability of diagnostic technology, and rehabilitation facilities equipped for the care of bariatric patients, which are indispensable structural components at the hospital level.³

Age and higher BMI were not significant independent predictors of postoperative morbidity in our study. Nevertheless, we approach older patients with caution because of recent findings that suggest that the risk-benefit ratio of bariatric surgery among older patients in terms of increased life span is not that favorable.⁵ Several studies have shown that bariatric surgery can be accomplished safely in super obese patients.^54,55 However, higher BMI should still be an important consideration in patient selection, as it correlates with a higher prevalence of comorbidities, longer operative times, and higher rates of conversion to open procedures.⁵⁶

In conclusion, presence of diabetes, open surgery, and early surgeon experience are associated with increased risk of complications after GBP. Routine proctoring of a surgeon's early experience and increased use of laparoscopic techniques may decrease complication rates. Bariatric surgery in high-volume centers with experienced surgeons and multidisciplinary care available is safe and yields the expected benefits of GBP: long-term and sustained weight loss along with improvement of obesity-associated diseases.

Correspondence: Guilherme M. Campos, MD, PhD, Department of Surgery, University of California, 521 Parnassus Ave, Room C-341, San Francisco, CA 94143-0790 (camposg@surgery.ucsf.edu).

Accepted for Publication: May 25, 2007.

Author Contributions:Study concept and design: Campos and Ciovica. Acquisition of data: Campos, Ciovica, and Takata. Analysis and interpretation of data: Campos, Ciovica, Rogers, Posselt, Vittinghoff, and Cello. Drafting of the manuscript: Campos, Ciovica, and Takata. Critical revision of the manuscript for important intellectual content: Campos, Ciovica, Rogers, Posselt, Vittinghoff, and Cello. Statistical analysis: Campos and Vittinghoff. Obtained funding: Campos. Administrative, technical, and material support: Campos and Ciovica. Study supervision: Campos, Takata, and Cello.

Financial Disclosure: Dr Campos has received educational grants from Ethicon Endosurgery and the United States Surgical Corporation.

Funding/Support: This publication was supported in part by grant KL2 RR024130 from the National Center for Research Resources, a component of the National Institutes of Health and the National Institutes of Health Roadmap for Medical Research.

Additional Contribution: Pamela Derish, MA, reviewed the manuscript.