- Email: [email protected]
Development of a sleeve gastrectomy risk calculator
Surgery for Obesity and Related Diseases ] (2015) 00–00
Original article
Development of a sleeve gastrectomy risk calculator Ali Aminian, M.D., Stacy A. Brethauer, M.D., Maryam Sharafkhah, M.S., Philip R. Schauer, M.D.* Bariatric and Metabolic Institute, Cleveland Clinic, Cleveland, Ohio Received November 11, 2014; accepted December 12, 2014
Abstract
Background: Laparoscopic sleeve gastrectomy (LSG) is rapidly gaining popularity. Estimating the risk of postoperative adverse events can improve surgical decision-making and informed patient consent. The objective of this study was to develop and validate a risk prediction model for early postoperative morbidity and mortality after LSG. Methods: Cases of primary LSG in the American College of Surgeons-National Surgical Quality Improvement Program (ACS-NSQIP) data set at year 2012 (n ¼ 5871) and 2011 (n ¼ 3130) were identified to develop and examine the validity of model. The composite primary outcome was defined as presence of any of 14 serious adverse events within the 30-days after LSG. Multiple logistic regression analysis was performed and a risk calculator was created to predict the primary outcome. Results: Thirty-day postoperative mortality and composite adverse events rates of 5871 LSG cases were .05% and 2.4%, respectively. Of the 52 examined baseline variables, the final model contained history of congestive heart failure (odds ratio [OR] 6.23; 95% CI 1.25–31.07), chronic steroid use (OR 5.00; 95% CI 2.06–12.15), male sex (OR 1.68; 95% CI 1.03–2.72), diabetes (OR 1.62; 95% CI 1.07–2.48), preoperative serum total bilirubin level (OR 1.57; 95% CI 1.11–2.22), body mass index (OR 1.03; 95% CI 1.01–1.05), and preoperative hematocrit level (OR .95; 95% CI .89–1.00). The risk model was then validated with the 2011 data set and was used to create an online risk calculator with a relatively good accuracy (c-statistic .682). Conclusions: This risk assessment scoring system, which specifically estimates serious adverse events after LSG, can contribute to surgical decision-making, informed patient consent, and prediction of surgical risk for patients and referring physicians. (Surg Obes Relat Dis 2015;]:00– 00.) r 2015 Published by Elsevier Inc. on behalf of American Society for Metabolic and Bariatric Surgery.
Keywords:
Bariatric; Sleeve gastrectomy; NSQIP; Risk; Complication; Morbidity; Mortality; Calculator; Morbid obesity
The growing epidemic of obesity, along with the relative ineffectiveness of conventional weight reduction therapies in severely obese individuals, has led to a remarkable rise in the number of bariatric surgical procedures in the past 2 decades [1]. Among all bariatric This study was presented as one of top 10 studies at Obesity Week, Boston, MA, November 2–7, 2014. * Correspondence: Philip R. Schauer, M.D., Bariatric and Metabolic Institute, Cleveland Clinic, 9500 Euclid Avenue, M61, Cleveland, OH 44195. E-mail: [email protected]
procedures, laparoscopic sleeve gastrectomy (LSG) has had a period of rapid growth in recent years. According to some estimates, LSG has become the most common bariatric procedure performed in the United States [2]. The LSG is a safe and relatively simple procedure and is associated with a low complication rate, even in high-risk patients who cannot tolerate a complex surgical procedure due to co-morbidities or anatomic limitations. In addition to achieving significant and durable weight loss, LSG is associated with long-term favorable metabolic effects [3–7].
http://dx.doi.org/10.1016/j.soard.2014.12.012 1550-7289/r 2015 Published by Elsevier Inc. on behalf of American Society for Metabolic and Bariatric Surgery.
2
A. Aminian et al. / Surgery for Obesity and Related Diseases ] (2015) 00–00
Despite the presence of robust data on safety and efficacy of bariatric surgery, many patients and physicians do not consider surgery to treat the disease of obesity. One reason may be an inaccurate belief of the benefit-to-risk ratio of medical versus surgical treatment of obesity [8,9]. Estimating the risk of postoperative adverse events can improve surgical decision-making and informed patient consent. In addition, there would be a considerable benefit in identifying modifiable preoperative factors that are associated with increased risk of postsurgical adverse events. Limitations of a few available predicted risk models include few baseline variables, combination of open and laparoscopic procedures, or being applicable mainly to gastric bypass [10–13]. The aim of this study was to develop a specific and valid preoperative risk calculator for estimation of early postoperative morbidity and mortality after LSG based on a national data set. Methods Data were extracted from the American College of Surgeons-National Surgical Quality Improvement Program (ACS-NSQIP) database. The ACS-NSQIP prospectively collects data on more than 150 variables, including standardized and audited demographic variables, co-morbidities, laboratory values, and 30-day postoperative mortality and morbidity outcomes for patients undergoing major surgical procedures in the United States (374 participating sites in 2012 and 315 sites in 2011). The ACS has used several mechanisms to ensure that the data collected are of the highest consistency and reliability [14]. Morbidly obese patients aged 418 undergoing LSG with the Current Procedural Terminology (CPT) code 43775 were included. Patients who underwent procedures that are commonly performed along with obesity surgery including concurrent endoscopy, liver biopsy, abdominal wall hernia repair, hiatal hernia repair, cholecystectomy, and procedures to manage intraoperative complications were included. However, revisional bariatric procedures and cases with unrelated concurrent procedures such as appendectomy and hysterectomy were excluded. Cases of LSG at year 2012 and 2011 were identified to make and then to examine the validity of risk model, respectively. The primary outcome was 30-day postoperative composite adverse events, which was defined as presence of any of 14 serious adverse events, including organ/space surgical site infection, stroke, coma, myocardial infarction, cardiac arrest, acute renal failure, deep vein thrombosis, pulmonary embolism, reintubation, failure to wean from mechanical ventilation, sepsis, septic shock, need for transfusion, and death. Independent demographic variables were sex, race, age (as a continuous and a categorical variable), height, weight, and body mass index (BMI; as a continuous and a categorical variable). Examined co-morbidities included
diabetes (and insulin usage), hypertension, history of pulmonary diseases (dyspnea and chronic obstructive pulmonary diseases), history of cardiac diseases (angina, myocardial infarction, congestive heart failure, cardiac interventions and surgeries), history of peripheral vascular diseases, history of kidney diseases (acute renal failure and being on dialysis), history of cerebrovascular diseases (transient ischemic attack and stroke), recent history of chemotherapy or radiotherapy, steroid use for chronic conditions, and bleeding disorders. Other baseline variables included smoking status, alcohol use, functional status (dependant/independent), and the American Society of Anesthesiologists (ASA) score. Preoperative laboratory variables included serum sodium, creatinine, blood urea nitrogen, albumin, aspartate aminotransferase, bilirubin, alkaline phosphatase, hematocrit, white blood cell count, platelet count, partial thromboplastin time, prothrombin time, and international normalized ratio. The effect of concomitant cholecystectomy was also assessed. All variables were clearly defined in the ACS-NSQIP database user guide [14]. Data analysis was performed using STATA (version 12, StataCorp, College Station, TX). To explore the risk factors associated with the primary outcome, the univariate analysis was performed using the Student’s t test for continuous variables and Pearson χ2 test or Fisher’s exact test for categorical variables. Multiple logistic regression with stepwise variable selection was used to construct a model for prediction of the primary outcome. Both the backward stepwise elimination and the forward stepwise selection methods were used to build a model. Independent variables with a significant association (P o .1) with the primary outcome in univariate analyses were entered into a model. With backward elimination procedure, only significant risk factors (P o .05) were kept in the model. Afterward, a forward stepwise selection was also used to find a stable model. The calibration of the model was tested using the Hosmer-Lemeshow goodness-of-fit test. The discriminatory capability of the model was assessed using the c-statistic which is the same as the area under the receiver operating characteristic (ROC) curve [10–12]. Then, the risk model based on the 2012 ACS-NSQIP data set was validated using the 2011 data set. The regression equation used to generate the model was used to construct a free online version of the calculator using the Cleveland Clinic Risk Calculator Constructor (http://www.r-calc.com). Results Baseline characteristics and co-morbidities of 5871 patients who underwent LSG in 2012 have been summarized in Table 1. Patients had a mean age of 43.8 ⫾ 11.2 years and BMI of 45.9 ⫾ 8.1 kg/m2. Eighty percent of the cohort was female, 76% were white, and 22% had diabetes.
Sleeve Gastrectomy Risk Calculator / Surgery for Obesity and Related Diseases ] (2015) 00–00
Postoperative adverse events are presented in Table 2. Thirty-day postoperative mortality and composite adverse events rates were .05% and 2.4%, respectively. Early reoperation was observed in 1.4%. Of the 52 examined baseline variables, the following were associated with increased risk of postoperative composite adverse events (including mortality) in multivariate analysis: history of congestive heart failure (odds ratio [OR] 6.23; 95% confidence interval [CI] 1.25 to 31.07), steroid use for chronic conditions (OR 5.00; 95% CI 2.06 to 12.15), male sex (OR 1.68; 95% CI 1.03 to 2.72), diabetes (OR 1.62; 95% CI 1.07 to 2.48), preoperative serum total bilirubin level (OR 1.57; 95% CI 1.11 to 2.22), body mass index (OR 1.03; 95% CI 1.01 to 1.05), and preoperative hematocrit level (OR .95; 95% CI .89 to 1.00). These factors are detailed in
3
Table 3. The multiple logistic regression equation was the following: L¼ –3:497 þ ð:487*DiabetesÞþ ð:0307*BMIÞ þ ð:517*Male GenderÞ þ ð1:83*CHFÞ þ ð1:61*Steroid UseÞ þ ð:451*BilirubinÞ –ð:054*HematocritÞ The estimated probability of an adverse event for a given patient is calculated using the following formula: Estimated probability of composite adverse event ð100%Þ ¼ EXP½L=ð1 þ EXP½LÞ The notation EXP is equivalent to ex, where “e” is the base of natural logarithm (2.718). The model demonstrated a good calibration (HosmerLemeshow goodness-of-fit test, χ2 ¼ 16.02, P ¼ .591) and
Table 1 Some of the independent variables (2012 ACS-NSQIP data set, n ¼ 5871) Variable
Mean ⫾ SD or %
Definition [14]
Age (years) Gender, female BMI (kg/m2) Race, white Diabetes
43.8 ⫾ 11.2 77.9 45.9 ⫾ 8.1 76.0 22.2
Hypertension Dyspnea
48.1 12.4
Previous cardiac surgery Previous PCI MI Angina CHF
1.1 1.4 .2 .4 .2
COPD
1.3
Positive cardiac history Renal failure Dialysis Cerebrovascular disease Chronic steroid use
3.4 .1 .3 1.5 1.3
Bleeding disorders Smoking Functional status, independent ASA Class, III
.9 10.2 99.6
Creatinine (mg/dL) Albumin (g/dL) Bilirubin, total (mg/dL) Hematocrit (%) Concomitant cholecystectomy
1.0 ⫾ .4 4.1 ⫾ .4 .5 ⫾ .5 40.3 ⫾ 3.6 2.1
— — Calculated based on the most recent weight and height — Diabetes requiring medication. Patients whose diabetes was controlled by diet alone are not included. Patients with insulin resistance who routinely take antidiabetic agents are included. Hypertension requiring medication. Chronic dyspnea (upon exertion or at rest) within the 30-days before surgery that reflects a chronic disease state. Any major cardiac surgical procedures. PCI at any time, including either balloon dilation or stent placement. History of MI in the 6 months before surgery. History of typical angina in the 30 days before surgery. Only newly diagnosed CHF in the 30 days before surgery or a diagnosis of chronic CHF with new signs or symptoms in the 30 days before surgery fulfill this definition. COPD resulting in any 1 or more of the following: functional disability, previous hospitalization, on chronic medications, and abnormal pulmonary function test. History of angina, MI, CHF, cardiac interventions, and surgeries Preoperative acute renal failure based on a significant increase in BUN or creatinine Currently on dialysis for acute or chronic renal failure History of transient ischemic attack or stroke. Regular administration of oral or parenteral corticosteroid or immunosuppressant medications, within the 30 days before the surgery for a chronic medical condition fulfills this definition. A one-time pulse, limited short course, a taper of less than 10 days duration, topical, inhalational, and rectal usage would not qualify. Any congenital or acquired bleeding disorder. Chronic aspirin therapy does not fulfill the definition. Current smoker within 1 year. This variable focuses on the patient's abilities to perform activities of daily living in the 30 days before surgery. The ASA Physical Status Classification of the patient’s present physical condition on a scale from 1 to 5. Preoperative serum creatinine. Preoperative serum albumin. Preoperative serum total bilirubin. Preoperative hematocrit. Simultaneous sleeve gastrectomy and cholecystectomy.
64.5
Abbreviations: ACS-NSQIP ¼ American College of Surgeons-National Surgical Quality Improvement Program; BMI ¼ body mass index; PCI ¼ percutaneous coronary intervention; MI ¼ myocardial infarction; CHF = congestive heart failure; COPD ¼ chronic obstructive pulmonary disease; BUN ¼ blood urea nitrogen; ASA ¼ American Society of Anesthesiology.
A. Aminian et al. / Surgery for Obesity and Related Diseases ] (2015) 00–00
4
Table 2 Serious adverse events (2012 ACS-NSQIP data set, n ¼ 5871) Variable
N (%)
Definition [14]
30-day composite adverse event Mortality Organ/space surgical site infection Blood transfusion
140 (2.4%)
Presence of any of the following 14 adverse events:
3 (.05%) 28 (.5%)
Deep vein thrombosis Pulmonary embolism Unplanned intubation
20 (.3%) 12 (.2%) 9 (.2%)
Mechanical ventilation 448 hours Acute renal failure
12 (.2%)
Stroke
1 (0%)
Coma Myocardial infarction Cardiac Arrest
0 4 (.1%) 3 (.1%)
Sepsis Septic shock Reoperation
26 (.4%) 5 (.1%) 85 (1.4%)
Death Infection involving any anatomic structure rather than the surgical incision which appears to be related to the operation. At least 1 unit of packed or whole red blood cells given from the surgical start time up to and including 72 hours postoperatively. Documented with imaging studies and requiring therapy Based on positive V-Q scan, CT scan, pulmonary arteriogram, or any other definitive modality. Placement of an endotracheal tube or other similar breathing tube and ventilator support which was not intended or planned. Total duration of ventilator-assisted respirations during postoperative hospitalization greater than 48 hours. Worsening of renal function postoperatively requiring dialysis, in a patient who did not require dialysis preoperatively. An embolic, thrombotic, or hemorrhagic vascular accident or stroke with motor, sensory, or cognitive dysfunction that persists for 24 or more hours. Being unconscious or unresponsive to all stimuli for greater than 24 hours. Documented with abnormal ECG and/or troponin. The absence of cardiac rhythm or presence of chaotic cardiac rhythm requiring the initiation of CPR, which includes chest compressions. Presence of systemic inflammatory response syndrome in response to an infectious process Sepsis and documented organ and/or circulatory dysfunction. Return to the operating room
71 (1.2%)
2 (0%)
Abbreviations: ACS-NSQIP ¼ American College of Surgeons-National Surgical Quality Improvement Program; V-Q ¼ ventilation-perfusion; CT ¼ computed tomography; ECG ¼ electrocardiogram.
a moderate discrimination (c-statistic .682). The generated risk model based on 2012 ACS-NSQIP was subsequently validated on the validation data set (2011, n ¼ 3130), which showed a relatively similar performance (c-statistic .63; 95% CI .55 to .71). On the basis of the regression equation and the parameter estimates listed in Table 3, a preoperative risk calculator for LSG was developed. A user-friendly version of the risk calculator is accessible at http://www.r-calc. com under the bariatric surgery formula tab. When the required patients’ values are entered into the calculator, the percent estimate of serious adverse events post-LSG is calculated. A few examples of the estimated probability of an adverse event, including mortality, are as follows:
Estimated risk in a healthy woman with BMI of 38 kg/
m2 and hematocrit of 42% would be 1%. Estimated risk in a diabetic man with BMI of 45 kg/m2 and hematocrit of 42% would be 3.3%. Estimated risk in a woman with BMI of 55 kg/m2 with nonalcoholic steatohepatitis and total bilirubin of 3 mg/ dL and hematocrit of 33% would be 9.6%. Estimated risk in a woman with BMI of 60 kg/m2, with diabetes and history of chronic steroid use (e.g., for asthma or rheumatoid arthritis) and hematocrit of 44% would be 12.6%. Estimated risk in a diabetic man with BMI of 62 kg/ m2, with orthopnea and dyspnea secondary to congestive heart failure and hematocrit of 40% would be 28.5%.
Table 3 Predictive factors of primary outcome based on multivariate analysis Risk factor
Univariate odds ratio
Adjusted odds ratio
95% CI
Estimate
Standard error for the estimate
Congestive heart failure Steroid use for chronic conditions Male sex Diabetes Preoperative serum total bilirubin level Body mass index Preoperative hematocrit level Constant
9.53 5.02 1.71 1.87 1.59 1.03 .96
6.23 5.00 1.68 1.62 1.57 1.03 .95
1.25–31.07 2.06–12.15 1.03–2.72 1.07–2.48 1.11–2.22 1.01–1.05 .89–1.00
1.83 1.61 .52 .49 .45 .03 –.05 –3.50
.82 .45 .25 .22 .18 .01 .03 1.32
Sleeve Gastrectomy Risk Calculator / Surgery for Obesity and Related Diseases ] (2015) 00–00
Notably, statistically significant variables in univariate analyses (P o .05) not selected by the logistic regression model included history of percutaneous coronary intervention (OR 4.48; 95% CI 1.50 to 13.40), preoperative dyspnea (OR 1.80; 95% CI 1.18 to 2.75), hypertension (OR 1.79; 95% CI 1.27 to 2.53), and serum creatinine (OR 1.29; 95% CI 1.05 to 1.59). Discussion Efficacy and safety are 2 essential elements of every treatment modality. Bariatric surgery is the most effective and durable weight loss method and is associated with favorable metabolic outcomes, including improvement of diabetes, hypertension, dyslipidemia, liver steatosis, sleep apnea, and survival rate among morbidly obese individuals [2–7,15]. Improvement in surgical techniques, laparoscopic training, and perioperative care has led to continuous progression of the safety profile of bariatric procedures [9,16]. Mortality of bariatric surgery has decreased substantially from 1.5%–2% 2 decades ago [17] to .05% in the current series of LSG. The modest rate of serious complication, comprising 14 early postoperative adverse events, in this series (2.4%) is consistent with reported complication rate of bariatric surgery based on several large databases (2– 4%) [8–12,17]. Incidence of all individual complications, except postoperative bleeding, was r .5% in this series (Table 2). Of the 52 examined baseline variables, the final model had 7 risk factors, including history of congestive heart failure, chronic steroid use, male gender, diabetes, high BMI, elevated preoperative serum bilirubin level, and low hematocrit. Among them, congestive heart failure and then chronic steroid use displayed the strongest independent associations with the probability of post-LSG adverse events. The significance of them had been proved in previous predictive risk models of various surgeries including bariatric procedures [11,16]. Preoperative optimization of patients with symptomatic heart failure may reduce the risk of elective surgery. Chronic steroid use can impair the healing process and increase infectious complications [18]. Conservative perioperative and intraoperative measures in such patients with immunosuppression may diminish the surgical risk. Male gender [10,13,16,19,20] and high BMI [11–13,16,21,22] have been well studied and are the established risk factors for morbidity and mortality after bariatric surgery. In addition, diabetes [19,20], anemia [19], and liver disease [16] have been the risk factors for development of complications in some reported risk models. Notably, age, which is a commonly reported risk factor in previous studies of bariatric surgery [10,11,13,16,19–22], was not associated with an increased risk of morbidity after LSG in the present multivariate analyses. There are a few risk stratification systems to estimate postoperative morbidity and mortality of bariatric surgery.
5
The well-known and most widely used Obesity SurgeryMortality Risk Score (OS-MRS) was developed based on more than 2000 open and laparoscopic gastric bypass procedures in a single center over a 10-year period from 1995 to 2004. This model is limited by inclusion of old data, consideration of only 12 baseline variables, combination of open and laparoscopic procedures, and being applicable to only gastric bypass [13]. For example, a 50year-old male patient with hypertension and BMI of 55 kg/ m2 has 4 risk factors of OS-MRS and 7.5% risk of mortality after bariatric surgery, which is a substantial overestimation in the current era. Three available risk models, which have been built based on the ACS-NSQIP data set, included patients before 2009, when the LSG was not being captured in the database. Turner et al. made a nomogram to predict early postoperative complications of bariatric surgery with a slight to moderate discriminative ability (c-statistic .629) [21]. A team of researchers led by Forse developed 2 risk calculators to separately predict morbidity [12] and mortality [11] of bariatric surgery. A study originated from Michigan Bariatric Surgery Collaborative analyzed data of more than 25,000 patients undergoing bariatric surgery (LSG, n ¼ 2279) between 2006 and 2010. Using the bariatric specific database, several factors were identified that contributed to the risk of serious postoperative complications including type of bariatric procedure (duodenal switch 4 gastric bypass 4 LSG 4 gastric band), previous history of venous thromboembolism, mobility limitations, coronary artery disease, age over 50 years, pulmonary disease, male gender, and smoking history with a moderate discriminative ability (c-statistic .68) [10]. Using the ACS-NSQIP, which is a robust clinical, validated, and audited database, enables inclusion of multiple preoperative variables from a large sample size of patients from both the academic and community centers. In contrast to some of previous studies, the current risk calculator was directly developed from a regression equation which provides an exact model-based estimate of adverse events after LSG. The user-friendly online version of this calculator can assist patients and physicians in risk assessment, decision making, need for preoperative medical optimization, and informed consent process before LSG. Since this calculator is specific to LSG, it can be used alongside with the previously published calculators which are mainly applicable to gastric bypass to compare safety profile of each procedure. The discriminative ability of the proposed model (cstatistic .68) is just below the absolute cutoff value of .70, which some researchers have suggested is needed for good discrimination. In fact, any predictive model with c-statistic 4 .50 may have clinical value [23]. Interestingly, all reported models for prediction of composite morbidity after bariatric surgery using large databases had a c-statistic o .70 [10,12,21,22]. Several factors can
6
A. Aminian et al. / Surgery for Obesity and Related Diseases ] (2015) 00–00
contribute to this finding. One possible reason is that data of potentially important factors such as sleep apnea, previous history of venous thromboembolism, surgeon’s experience and case volume were not available in the ACS-NSQIP. The importance of these factors in development of postbariatric surgery adverse events has been shown in several studies including the Longitudinal Assessment of Bariatric Surgery (LABS) study [24,25]. Consideration of these variables could help the risk adjustment model. Using of composite outcome, consisting of various heterogeneous dependent variables, instead of one outcome (e.g., mortality) can also be a contributing factor. Risk factors of postoperative bleeding, DVT, and renal failure are different and risk factors of one of them may not predict the others. Notably, 2 consecutive ACS-NSQIP studies with relatively similar statistical methodology on morbidity (composite outcome) [12] and mortality (single outcome) [11] of bariatric surgery had c-statistic of .69 and .80, respectively. Similarly, the discriminative ability of colorectal surgery risk calculator based on ACS-NSQIP data was better for mortality than prediction of composite morbidity outcome (c-statistic .90 versus .68, respectively) [26]. The other proposed explanation is the relative uniformity of the patient population who undergo bariatric surgery [10]. In addition to moderate discriminative ability, this model does have other limitations. The ACS-NSQIP collects only a sample of procedures from each participating hospital and does not collect data unique to bariatric operations including preoperative co-morbidities such as sleep apnea, history of venous thromboembolism, and pulmonary hypertension as well as postoperative complications such as gastric leak. As was mentioned above, surgeon’s experience was not considered in this risk model. Therefore, the risk assessment tool may be less accurate for surgeons and hospitals with significantly poorer performance than national outcomes. In addition, the data set only includes short-term postoperative outcomes and does not capture complications beyond 30 days after surgery which can lead to underestimation of real risk. However, most serious complications occur in the initial 30 days after LSG. There is a need for development of risk models for various procedures based on standardized bariatric surgery specific multiinstitutional databases. The implementation of the national Metabolic and Bariatric Surgery Accreditation and Quality Improvement Program (MBSAQIP) data set can provide more detailed bariatric surgery specific data elements and short-term and long-term adverse events. Conclusion These national data point to the overall safety of LSG as a treatment for severe obesity. Symptomatic congestive heart failure, chronic steroid use, male gender, diabetes, increasing BMI, preoperative serum bilirubin level, and low hematocrit were identified as major risk factors for post-
LSG serious adverse events. This risk assessment scoring system, which specifically estimates mortality and morbidity after LSG, may contribute to surgical decision-making, informed patient consent, prediction of surgical risk for patients and referring physicians and, therefore, improve patient care. Further studies are warranted to externally validate this risk model in a different population of LSG patients. Disclosures The authors have no commercial associations that might be a conflict of interest in relation to this article. Acknowledgments The authors acknowledge the assistance of Kevin Chagin. ACS NSQIP Disclaimer: The American College of Surgeons National Surgical Quality Improvement Program and the hospitals participating in the ACS NSQIP are the source of the data used herein; they have not verified and are not responsible for the statistical validity of the data analysis or the conclusions derived by the authors. References [1] Buchwald H, Oien DM. Metabolic/bariatric surgery worldwide 2011. Obes Surg 2013;23:427–36. [2] ASMBS. Estimate of bariatric surgery numbers. Available from: http://asmbs.org/2014/03/estimate-of-bariatric-surgery-numbers/. Accessed: Sept 9, 2014. [3] Eid GM, Brethauer S, Mattar SG, Titchner RL, Gourash W, Schauer PR. Laparoscopic sleeve gastrectomy for super-obese patients: fortyeight percent excess weight loss after 6 to 8 years with 93% followup. Ann Surg 2012;256:262–5. [4] Brethauer SA, Hammel JP, Schauer PR. Systematic review of sleeve gastrectomy as staging and primary bariatric procedure. Surg Obes Relat Dis 2009;5:469–75. [5] Carlin AM, Zeni TM, English WJ, et al. The comparative effectiveness of sleeve gastrectomy, gastric bypass, and adjustable gastric banding procedures for the treatment of morbid obesity. Ann Surg 2013;257:791–7. [6] Schauer PR, Bhatt DL, Kirwan JP, et al. Bariatric surgery versus intensive medical therapy for diabetes—3-year outcomes. N Engl J Med 2014;370:2002–13. [7] Brethauer SA, Aminian A, Romero-Talamás H, et al. Can diabetes be surgically cured? Long-term metabolic effects of bariatric surgery in obese patients with type 2 diabetes mellitus. Ann Surg 2013;258: 628–37. [8] Pomp A. Safety of bariatric surgery. Lancet Diabetes Endocrinol 2014;2:98–100. [9] Aminian A, Brethauer SA, Kirwan JP, Kashyap SR, Burguera B, Schauer PR. How safe is metabolic/diabetes surgery? Diabetes Obes Metab Epub 2015;17:198–201. [10] Finks JF, Kole KL, Yenumula PR, et al. Predicting risk for serious complications with bariatric surgery: results from the Michigan Bariatric Surgery Collaborative. Ann Surg 2011;254:633–40. [11] Ramanan B, Gupta PK, Gupta H, Fang X, Forse RA. Development and validation of a bariatric surgery mortality risk calculator. J Am Coll Surg 2012;214:892–900. [12] Gupta PK, Franck C, Miller WJ, Gupta H, Forse RA. Development and validation of a bariatric surgery morbidity risk calculator using
Sleeve Gastrectomy Risk Calculator / Surgery for Obesity and Related Diseases ] (2015) 00–00
[13]
[14]
[15]
[16]
[17] [18] [19] [20]
the prospective, multicenter NSQIP data set. J Am Coll Surg 2011;212:301–9. DeMaria EJ, Portenier D, Wolfe L. Obesity surgery mortality risk score: proposal for a clinically useful score to predict mortality risk in patients undergoing gastric bypass. Surg Obes Relat Dis 2007;3:134–40. American College of Surgeons. National Surgical Quality Improvement Program Participant User Data Profile. Available from: http:// site.acsnsqip.org/participant-use-data-file/. Accessed: May 15, 2014. Aminian A, Daigle CR, Romero-Talamás H, et al. Risk prediction of complications of metabolic syndrome before and 6 years after gastric bypass. Surg Obes Relat Dis 2014;10:576–82. Benotti P, Wood GC, Winegar DA, et al. Risk factors associated with mortality after Roux-en-Y gastric bypass surgery. Ann Surg 2014;259: 123–30. Kim JH, Wolfe B. Bariatric/metabolic surgery: short- and long-term safety. Curr Atheroscler Rep 2012;14:597–605. Ismael H, Horst M, Farooq M, et al. Adverse effects of preoperative steroid use on surgical outcomes. Am J Surg 2011;201:305–8. Livingston EH. Development of bariatric surgery-specific risk assessment tool. Surg Obes Relat Dis 2007;3:14–20. Nguyen NT, Nguyen B, Smith B, et al. Proposal for a bariatric mortality risk classification system for patients undergoing bariatric surgery. Surg Obes Relat Dis 2013;9:239–46.
7
[21] Turner PL, Saager L, Dalton J, et al. A nomogram for predicting surgical complications in bariatric surgery patients. Obes Surg 2011;21:655–62. [22] Maciejewski ML, Winegar DA, Farley JF, et al. Risk stratification of serious adverse events after gastric bypass in the Bariatric Outcomes Longitudinal Database. Surg Obes Relat Dis 2012;8: 671–7. [23] Wells BJ, Kattan MW, Cooper GS, et al. ColoRectal Cancer Predicted Risk Online (CRC-PRO) calculator using data from the MultiEthnic Cohort Study. J Am Board Fam Med 2014;27:42–55. [24] Longitudinal Assessment of Bariatric Surgery (LABS) Consortium, Flum DR, Belle SH, King WC, et al. Perioperative safety in the longitudinal assessment of bariatric surgery. N Engl J Med 2009;361: 445–54. [25] Smith MD, Patterson E, Wahed AS, et al. Relationship between surgeon volume and adverse outcomes after RYGB in Longitudinal Assessment of Bariatric Surgery (LABS) study. Surg Obes Relat Dis 2010;6:118–25. [26] Cohen ME, Bilimoria KY, Ko CY, et al. Development of an American College of Surgeons National Surgery Quality Improvement Program: morbidity and mortality risk calculator for colorectal surgery. J Am Coll Surg 2009;208:1009–16.
Original article
Development of a sleeve gastrectomy risk calculator Ali Aminian, M.D., Stacy A. Brethauer, M.D., Maryam Sharafkhah, M.S., Philip R. Schauer, M.D.* Bariatric and Metabolic Institute, Cleveland Clinic, Cleveland, Ohio Received November 11, 2014; accepted December 12, 2014
Abstract
Background: Laparoscopic sleeve gastrectomy (LSG) is rapidly gaining popularity. Estimating the risk of postoperative adverse events can improve surgical decision-making and informed patient consent. The objective of this study was to develop and validate a risk prediction model for early postoperative morbidity and mortality after LSG. Methods: Cases of primary LSG in the American College of Surgeons-National Surgical Quality Improvement Program (ACS-NSQIP) data set at year 2012 (n ¼ 5871) and 2011 (n ¼ 3130) were identified to develop and examine the validity of model. The composite primary outcome was defined as presence of any of 14 serious adverse events within the 30-days after LSG. Multiple logistic regression analysis was performed and a risk calculator was created to predict the primary outcome. Results: Thirty-day postoperative mortality and composite adverse events rates of 5871 LSG cases were .05% and 2.4%, respectively. Of the 52 examined baseline variables, the final model contained history of congestive heart failure (odds ratio [OR] 6.23; 95% CI 1.25–31.07), chronic steroid use (OR 5.00; 95% CI 2.06–12.15), male sex (OR 1.68; 95% CI 1.03–2.72), diabetes (OR 1.62; 95% CI 1.07–2.48), preoperative serum total bilirubin level (OR 1.57; 95% CI 1.11–2.22), body mass index (OR 1.03; 95% CI 1.01–1.05), and preoperative hematocrit level (OR .95; 95% CI .89–1.00). The risk model was then validated with the 2011 data set and was used to create an online risk calculator with a relatively good accuracy (c-statistic .682). Conclusions: This risk assessment scoring system, which specifically estimates serious adverse events after LSG, can contribute to surgical decision-making, informed patient consent, and prediction of surgical risk for patients and referring physicians. (Surg Obes Relat Dis 2015;]:00– 00.) r 2015 Published by Elsevier Inc. on behalf of American Society for Metabolic and Bariatric Surgery.
Keywords:
Bariatric; Sleeve gastrectomy; NSQIP; Risk; Complication; Morbidity; Mortality; Calculator; Morbid obesity
The growing epidemic of obesity, along with the relative ineffectiveness of conventional weight reduction therapies in severely obese individuals, has led to a remarkable rise in the number of bariatric surgical procedures in the past 2 decades [1]. Among all bariatric This study was presented as one of top 10 studies at Obesity Week, Boston, MA, November 2–7, 2014. * Correspondence: Philip R. Schauer, M.D., Bariatric and Metabolic Institute, Cleveland Clinic, 9500 Euclid Avenue, M61, Cleveland, OH 44195. E-mail: [email protected]
procedures, laparoscopic sleeve gastrectomy (LSG) has had a period of rapid growth in recent years. According to some estimates, LSG has become the most common bariatric procedure performed in the United States [2]. The LSG is a safe and relatively simple procedure and is associated with a low complication rate, even in high-risk patients who cannot tolerate a complex surgical procedure due to co-morbidities or anatomic limitations. In addition to achieving significant and durable weight loss, LSG is associated with long-term favorable metabolic effects [3–7].
http://dx.doi.org/10.1016/j.soard.2014.12.012 1550-7289/r 2015 Published by Elsevier Inc. on behalf of American Society for Metabolic and Bariatric Surgery.
2
A. Aminian et al. / Surgery for Obesity and Related Diseases ] (2015) 00–00
Despite the presence of robust data on safety and efficacy of bariatric surgery, many patients and physicians do not consider surgery to treat the disease of obesity. One reason may be an inaccurate belief of the benefit-to-risk ratio of medical versus surgical treatment of obesity [8,9]. Estimating the risk of postoperative adverse events can improve surgical decision-making and informed patient consent. In addition, there would be a considerable benefit in identifying modifiable preoperative factors that are associated with increased risk of postsurgical adverse events. Limitations of a few available predicted risk models include few baseline variables, combination of open and laparoscopic procedures, or being applicable mainly to gastric bypass [10–13]. The aim of this study was to develop a specific and valid preoperative risk calculator for estimation of early postoperative morbidity and mortality after LSG based on a national data set. Methods Data were extracted from the American College of Surgeons-National Surgical Quality Improvement Program (ACS-NSQIP) database. The ACS-NSQIP prospectively collects data on more than 150 variables, including standardized and audited demographic variables, co-morbidities, laboratory values, and 30-day postoperative mortality and morbidity outcomes for patients undergoing major surgical procedures in the United States (374 participating sites in 2012 and 315 sites in 2011). The ACS has used several mechanisms to ensure that the data collected are of the highest consistency and reliability [14]. Morbidly obese patients aged 418 undergoing LSG with the Current Procedural Terminology (CPT) code 43775 were included. Patients who underwent procedures that are commonly performed along with obesity surgery including concurrent endoscopy, liver biopsy, abdominal wall hernia repair, hiatal hernia repair, cholecystectomy, and procedures to manage intraoperative complications were included. However, revisional bariatric procedures and cases with unrelated concurrent procedures such as appendectomy and hysterectomy were excluded. Cases of LSG at year 2012 and 2011 were identified to make and then to examine the validity of risk model, respectively. The primary outcome was 30-day postoperative composite adverse events, which was defined as presence of any of 14 serious adverse events, including organ/space surgical site infection, stroke, coma, myocardial infarction, cardiac arrest, acute renal failure, deep vein thrombosis, pulmonary embolism, reintubation, failure to wean from mechanical ventilation, sepsis, septic shock, need for transfusion, and death. Independent demographic variables were sex, race, age (as a continuous and a categorical variable), height, weight, and body mass index (BMI; as a continuous and a categorical variable). Examined co-morbidities included
diabetes (and insulin usage), hypertension, history of pulmonary diseases (dyspnea and chronic obstructive pulmonary diseases), history of cardiac diseases (angina, myocardial infarction, congestive heart failure, cardiac interventions and surgeries), history of peripheral vascular diseases, history of kidney diseases (acute renal failure and being on dialysis), history of cerebrovascular diseases (transient ischemic attack and stroke), recent history of chemotherapy or radiotherapy, steroid use for chronic conditions, and bleeding disorders. Other baseline variables included smoking status, alcohol use, functional status (dependant/independent), and the American Society of Anesthesiologists (ASA) score. Preoperative laboratory variables included serum sodium, creatinine, blood urea nitrogen, albumin, aspartate aminotransferase, bilirubin, alkaline phosphatase, hematocrit, white blood cell count, platelet count, partial thromboplastin time, prothrombin time, and international normalized ratio. The effect of concomitant cholecystectomy was also assessed. All variables were clearly defined in the ACS-NSQIP database user guide [14]. Data analysis was performed using STATA (version 12, StataCorp, College Station, TX). To explore the risk factors associated with the primary outcome, the univariate analysis was performed using the Student’s t test for continuous variables and Pearson χ2 test or Fisher’s exact test for categorical variables. Multiple logistic regression with stepwise variable selection was used to construct a model for prediction of the primary outcome. Both the backward stepwise elimination and the forward stepwise selection methods were used to build a model. Independent variables with a significant association (P o .1) with the primary outcome in univariate analyses were entered into a model. With backward elimination procedure, only significant risk factors (P o .05) were kept in the model. Afterward, a forward stepwise selection was also used to find a stable model. The calibration of the model was tested using the Hosmer-Lemeshow goodness-of-fit test. The discriminatory capability of the model was assessed using the c-statistic which is the same as the area under the receiver operating characteristic (ROC) curve [10–12]. Then, the risk model based on the 2012 ACS-NSQIP data set was validated using the 2011 data set. The regression equation used to generate the model was used to construct a free online version of the calculator using the Cleveland Clinic Risk Calculator Constructor (http://www.r-calc.com). Results Baseline characteristics and co-morbidities of 5871 patients who underwent LSG in 2012 have been summarized in Table 1. Patients had a mean age of 43.8 ⫾ 11.2 years and BMI of 45.9 ⫾ 8.1 kg/m2. Eighty percent of the cohort was female, 76% were white, and 22% had diabetes.
Sleeve Gastrectomy Risk Calculator / Surgery for Obesity and Related Diseases ] (2015) 00–00
Postoperative adverse events are presented in Table 2. Thirty-day postoperative mortality and composite adverse events rates were .05% and 2.4%, respectively. Early reoperation was observed in 1.4%. Of the 52 examined baseline variables, the following were associated with increased risk of postoperative composite adverse events (including mortality) in multivariate analysis: history of congestive heart failure (odds ratio [OR] 6.23; 95% confidence interval [CI] 1.25 to 31.07), steroid use for chronic conditions (OR 5.00; 95% CI 2.06 to 12.15), male sex (OR 1.68; 95% CI 1.03 to 2.72), diabetes (OR 1.62; 95% CI 1.07 to 2.48), preoperative serum total bilirubin level (OR 1.57; 95% CI 1.11 to 2.22), body mass index (OR 1.03; 95% CI 1.01 to 1.05), and preoperative hematocrit level (OR .95; 95% CI .89 to 1.00). These factors are detailed in
3
Table 3. The multiple logistic regression equation was the following: L¼ –3:497 þ ð:487*DiabetesÞþ ð:0307*BMIÞ þ ð:517*Male GenderÞ þ ð1:83*CHFÞ þ ð1:61*Steroid UseÞ þ ð:451*BilirubinÞ –ð:054*HematocritÞ The estimated probability of an adverse event for a given patient is calculated using the following formula: Estimated probability of composite adverse event ð100%Þ ¼ EXP½L=ð1 þ EXP½LÞ The notation EXP is equivalent to ex, where “e” is the base of natural logarithm (2.718). The model demonstrated a good calibration (HosmerLemeshow goodness-of-fit test, χ2 ¼ 16.02, P ¼ .591) and
Table 1 Some of the independent variables (2012 ACS-NSQIP data set, n ¼ 5871) Variable
Mean ⫾ SD or %
Definition [14]
Age (years) Gender, female BMI (kg/m2) Race, white Diabetes
43.8 ⫾ 11.2 77.9 45.9 ⫾ 8.1 76.0 22.2
Hypertension Dyspnea
48.1 12.4
Previous cardiac surgery Previous PCI MI Angina CHF
1.1 1.4 .2 .4 .2
COPD
1.3
Positive cardiac history Renal failure Dialysis Cerebrovascular disease Chronic steroid use
3.4 .1 .3 1.5 1.3
Bleeding disorders Smoking Functional status, independent ASA Class, III
.9 10.2 99.6
Creatinine (mg/dL) Albumin (g/dL) Bilirubin, total (mg/dL) Hematocrit (%) Concomitant cholecystectomy
1.0 ⫾ .4 4.1 ⫾ .4 .5 ⫾ .5 40.3 ⫾ 3.6 2.1
— — Calculated based on the most recent weight and height — Diabetes requiring medication. Patients whose diabetes was controlled by diet alone are not included. Patients with insulin resistance who routinely take antidiabetic agents are included. Hypertension requiring medication. Chronic dyspnea (upon exertion or at rest) within the 30-days before surgery that reflects a chronic disease state. Any major cardiac surgical procedures. PCI at any time, including either balloon dilation or stent placement. History of MI in the 6 months before surgery. History of typical angina in the 30 days before surgery. Only newly diagnosed CHF in the 30 days before surgery or a diagnosis of chronic CHF with new signs or symptoms in the 30 days before surgery fulfill this definition. COPD resulting in any 1 or more of the following: functional disability, previous hospitalization, on chronic medications, and abnormal pulmonary function test. History of angina, MI, CHF, cardiac interventions, and surgeries Preoperative acute renal failure based on a significant increase in BUN or creatinine Currently on dialysis for acute or chronic renal failure History of transient ischemic attack or stroke. Regular administration of oral or parenteral corticosteroid or immunosuppressant medications, within the 30 days before the surgery for a chronic medical condition fulfills this definition. A one-time pulse, limited short course, a taper of less than 10 days duration, topical, inhalational, and rectal usage would not qualify. Any congenital or acquired bleeding disorder. Chronic aspirin therapy does not fulfill the definition. Current smoker within 1 year. This variable focuses on the patient's abilities to perform activities of daily living in the 30 days before surgery. The ASA Physical Status Classification of the patient’s present physical condition on a scale from 1 to 5. Preoperative serum creatinine. Preoperative serum albumin. Preoperative serum total bilirubin. Preoperative hematocrit. Simultaneous sleeve gastrectomy and cholecystectomy.
64.5
Abbreviations: ACS-NSQIP ¼ American College of Surgeons-National Surgical Quality Improvement Program; BMI ¼ body mass index; PCI ¼ percutaneous coronary intervention; MI ¼ myocardial infarction; CHF = congestive heart failure; COPD ¼ chronic obstructive pulmonary disease; BUN ¼ blood urea nitrogen; ASA ¼ American Society of Anesthesiology.
A. Aminian et al. / Surgery for Obesity and Related Diseases ] (2015) 00–00
4
Table 2 Serious adverse events (2012 ACS-NSQIP data set, n ¼ 5871) Variable
N (%)
Definition [14]
30-day composite adverse event Mortality Organ/space surgical site infection Blood transfusion
140 (2.4%)
Presence of any of the following 14 adverse events:
3 (.05%) 28 (.5%)
Deep vein thrombosis Pulmonary embolism Unplanned intubation
20 (.3%) 12 (.2%) 9 (.2%)
Mechanical ventilation 448 hours Acute renal failure
12 (.2%)
Stroke
1 (0%)
Coma Myocardial infarction Cardiac Arrest
0 4 (.1%) 3 (.1%)
Sepsis Septic shock Reoperation
26 (.4%) 5 (.1%) 85 (1.4%)
Death Infection involving any anatomic structure rather than the surgical incision which appears to be related to the operation. At least 1 unit of packed or whole red blood cells given from the surgical start time up to and including 72 hours postoperatively. Documented with imaging studies and requiring therapy Based on positive V-Q scan, CT scan, pulmonary arteriogram, or any other definitive modality. Placement of an endotracheal tube or other similar breathing tube and ventilator support which was not intended or planned. Total duration of ventilator-assisted respirations during postoperative hospitalization greater than 48 hours. Worsening of renal function postoperatively requiring dialysis, in a patient who did not require dialysis preoperatively. An embolic, thrombotic, or hemorrhagic vascular accident or stroke with motor, sensory, or cognitive dysfunction that persists for 24 or more hours. Being unconscious or unresponsive to all stimuli for greater than 24 hours. Documented with abnormal ECG and/or troponin. The absence of cardiac rhythm or presence of chaotic cardiac rhythm requiring the initiation of CPR, which includes chest compressions. Presence of systemic inflammatory response syndrome in response to an infectious process Sepsis and documented organ and/or circulatory dysfunction. Return to the operating room
71 (1.2%)
2 (0%)
Abbreviations: ACS-NSQIP ¼ American College of Surgeons-National Surgical Quality Improvement Program; V-Q ¼ ventilation-perfusion; CT ¼ computed tomography; ECG ¼ electrocardiogram.
a moderate discrimination (c-statistic .682). The generated risk model based on 2012 ACS-NSQIP was subsequently validated on the validation data set (2011, n ¼ 3130), which showed a relatively similar performance (c-statistic .63; 95% CI .55 to .71). On the basis of the regression equation and the parameter estimates listed in Table 3, a preoperative risk calculator for LSG was developed. A user-friendly version of the risk calculator is accessible at http://www.r-calc. com under the bariatric surgery formula tab. When the required patients’ values are entered into the calculator, the percent estimate of serious adverse events post-LSG is calculated. A few examples of the estimated probability of an adverse event, including mortality, are as follows:
Estimated risk in a healthy woman with BMI of 38 kg/
m2 and hematocrit of 42% would be 1%. Estimated risk in a diabetic man with BMI of 45 kg/m2 and hematocrit of 42% would be 3.3%. Estimated risk in a woman with BMI of 55 kg/m2 with nonalcoholic steatohepatitis and total bilirubin of 3 mg/ dL and hematocrit of 33% would be 9.6%. Estimated risk in a woman with BMI of 60 kg/m2, with diabetes and history of chronic steroid use (e.g., for asthma or rheumatoid arthritis) and hematocrit of 44% would be 12.6%. Estimated risk in a diabetic man with BMI of 62 kg/ m2, with orthopnea and dyspnea secondary to congestive heart failure and hematocrit of 40% would be 28.5%.
Table 3 Predictive factors of primary outcome based on multivariate analysis Risk factor
Univariate odds ratio
Adjusted odds ratio
95% CI
Estimate
Standard error for the estimate
Congestive heart failure Steroid use for chronic conditions Male sex Diabetes Preoperative serum total bilirubin level Body mass index Preoperative hematocrit level Constant
9.53 5.02 1.71 1.87 1.59 1.03 .96
6.23 5.00 1.68 1.62 1.57 1.03 .95
1.25–31.07 2.06–12.15 1.03–2.72 1.07–2.48 1.11–2.22 1.01–1.05 .89–1.00
1.83 1.61 .52 .49 .45 .03 –.05 –3.50
.82 .45 .25 .22 .18 .01 .03 1.32
Sleeve Gastrectomy Risk Calculator / Surgery for Obesity and Related Diseases ] (2015) 00–00
Notably, statistically significant variables in univariate analyses (P o .05) not selected by the logistic regression model included history of percutaneous coronary intervention (OR 4.48; 95% CI 1.50 to 13.40), preoperative dyspnea (OR 1.80; 95% CI 1.18 to 2.75), hypertension (OR 1.79; 95% CI 1.27 to 2.53), and serum creatinine (OR 1.29; 95% CI 1.05 to 1.59). Discussion Efficacy and safety are 2 essential elements of every treatment modality. Bariatric surgery is the most effective and durable weight loss method and is associated with favorable metabolic outcomes, including improvement of diabetes, hypertension, dyslipidemia, liver steatosis, sleep apnea, and survival rate among morbidly obese individuals [2–7,15]. Improvement in surgical techniques, laparoscopic training, and perioperative care has led to continuous progression of the safety profile of bariatric procedures [9,16]. Mortality of bariatric surgery has decreased substantially from 1.5%–2% 2 decades ago [17] to .05% in the current series of LSG. The modest rate of serious complication, comprising 14 early postoperative adverse events, in this series (2.4%) is consistent with reported complication rate of bariatric surgery based on several large databases (2– 4%) [8–12,17]. Incidence of all individual complications, except postoperative bleeding, was r .5% in this series (Table 2). Of the 52 examined baseline variables, the final model had 7 risk factors, including history of congestive heart failure, chronic steroid use, male gender, diabetes, high BMI, elevated preoperative serum bilirubin level, and low hematocrit. Among them, congestive heart failure and then chronic steroid use displayed the strongest independent associations with the probability of post-LSG adverse events. The significance of them had been proved in previous predictive risk models of various surgeries including bariatric procedures [11,16]. Preoperative optimization of patients with symptomatic heart failure may reduce the risk of elective surgery. Chronic steroid use can impair the healing process and increase infectious complications [18]. Conservative perioperative and intraoperative measures in such patients with immunosuppression may diminish the surgical risk. Male gender [10,13,16,19,20] and high BMI [11–13,16,21,22] have been well studied and are the established risk factors for morbidity and mortality after bariatric surgery. In addition, diabetes [19,20], anemia [19], and liver disease [16] have been the risk factors for development of complications in some reported risk models. Notably, age, which is a commonly reported risk factor in previous studies of bariatric surgery [10,11,13,16,19–22], was not associated with an increased risk of morbidity after LSG in the present multivariate analyses. There are a few risk stratification systems to estimate postoperative morbidity and mortality of bariatric surgery.
5
The well-known and most widely used Obesity SurgeryMortality Risk Score (OS-MRS) was developed based on more than 2000 open and laparoscopic gastric bypass procedures in a single center over a 10-year period from 1995 to 2004. This model is limited by inclusion of old data, consideration of only 12 baseline variables, combination of open and laparoscopic procedures, and being applicable to only gastric bypass [13]. For example, a 50year-old male patient with hypertension and BMI of 55 kg/ m2 has 4 risk factors of OS-MRS and 7.5% risk of mortality after bariatric surgery, which is a substantial overestimation in the current era. Three available risk models, which have been built based on the ACS-NSQIP data set, included patients before 2009, when the LSG was not being captured in the database. Turner et al. made a nomogram to predict early postoperative complications of bariatric surgery with a slight to moderate discriminative ability (c-statistic .629) [21]. A team of researchers led by Forse developed 2 risk calculators to separately predict morbidity [12] and mortality [11] of bariatric surgery. A study originated from Michigan Bariatric Surgery Collaborative analyzed data of more than 25,000 patients undergoing bariatric surgery (LSG, n ¼ 2279) between 2006 and 2010. Using the bariatric specific database, several factors were identified that contributed to the risk of serious postoperative complications including type of bariatric procedure (duodenal switch 4 gastric bypass 4 LSG 4 gastric band), previous history of venous thromboembolism, mobility limitations, coronary artery disease, age over 50 years, pulmonary disease, male gender, and smoking history with a moderate discriminative ability (c-statistic .68) [10]. Using the ACS-NSQIP, which is a robust clinical, validated, and audited database, enables inclusion of multiple preoperative variables from a large sample size of patients from both the academic and community centers. In contrast to some of previous studies, the current risk calculator was directly developed from a regression equation which provides an exact model-based estimate of adverse events after LSG. The user-friendly online version of this calculator can assist patients and physicians in risk assessment, decision making, need for preoperative medical optimization, and informed consent process before LSG. Since this calculator is specific to LSG, it can be used alongside with the previously published calculators which are mainly applicable to gastric bypass to compare safety profile of each procedure. The discriminative ability of the proposed model (cstatistic .68) is just below the absolute cutoff value of .70, which some researchers have suggested is needed for good discrimination. In fact, any predictive model with c-statistic 4 .50 may have clinical value [23]. Interestingly, all reported models for prediction of composite morbidity after bariatric surgery using large databases had a c-statistic o .70 [10,12,21,22]. Several factors can
6
A. Aminian et al. / Surgery for Obesity and Related Diseases ] (2015) 00–00
contribute to this finding. One possible reason is that data of potentially important factors such as sleep apnea, previous history of venous thromboembolism, surgeon’s experience and case volume were not available in the ACS-NSQIP. The importance of these factors in development of postbariatric surgery adverse events has been shown in several studies including the Longitudinal Assessment of Bariatric Surgery (LABS) study [24,25]. Consideration of these variables could help the risk adjustment model. Using of composite outcome, consisting of various heterogeneous dependent variables, instead of one outcome (e.g., mortality) can also be a contributing factor. Risk factors of postoperative bleeding, DVT, and renal failure are different and risk factors of one of them may not predict the others. Notably, 2 consecutive ACS-NSQIP studies with relatively similar statistical methodology on morbidity (composite outcome) [12] and mortality (single outcome) [11] of bariatric surgery had c-statistic of .69 and .80, respectively. Similarly, the discriminative ability of colorectal surgery risk calculator based on ACS-NSQIP data was better for mortality than prediction of composite morbidity outcome (c-statistic .90 versus .68, respectively) [26]. The other proposed explanation is the relative uniformity of the patient population who undergo bariatric surgery [10]. In addition to moderate discriminative ability, this model does have other limitations. The ACS-NSQIP collects only a sample of procedures from each participating hospital and does not collect data unique to bariatric operations including preoperative co-morbidities such as sleep apnea, history of venous thromboembolism, and pulmonary hypertension as well as postoperative complications such as gastric leak. As was mentioned above, surgeon’s experience was not considered in this risk model. Therefore, the risk assessment tool may be less accurate for surgeons and hospitals with significantly poorer performance than national outcomes. In addition, the data set only includes short-term postoperative outcomes and does not capture complications beyond 30 days after surgery which can lead to underestimation of real risk. However, most serious complications occur in the initial 30 days after LSG. There is a need for development of risk models for various procedures based on standardized bariatric surgery specific multiinstitutional databases. The implementation of the national Metabolic and Bariatric Surgery Accreditation and Quality Improvement Program (MBSAQIP) data set can provide more detailed bariatric surgery specific data elements and short-term and long-term adverse events. Conclusion These national data point to the overall safety of LSG as a treatment for severe obesity. Symptomatic congestive heart failure, chronic steroid use, male gender, diabetes, increasing BMI, preoperative serum bilirubin level, and low hematocrit were identified as major risk factors for post-
LSG serious adverse events. This risk assessment scoring system, which specifically estimates mortality and morbidity after LSG, may contribute to surgical decision-making, informed patient consent, prediction of surgical risk for patients and referring physicians and, therefore, improve patient care. Further studies are warranted to externally validate this risk model in a different population of LSG patients. Disclosures The authors have no commercial associations that might be a conflict of interest in relation to this article. Acknowledgments The authors acknowledge the assistance of Kevin Chagin. ACS NSQIP Disclaimer: The American College of Surgeons National Surgical Quality Improvement Program and the hospitals participating in the ACS NSQIP are the source of the data used herein; they have not verified and are not responsible for the statistical validity of the data analysis or the conclusions derived by the authors. References [1] Buchwald H, Oien DM. Metabolic/bariatric surgery worldwide 2011. Obes Surg 2013;23:427–36. [2] ASMBS. Estimate of bariatric surgery numbers. Available from: http://asmbs.org/2014/03/estimate-of-bariatric-surgery-numbers/. Accessed: Sept 9, 2014. [3] Eid GM, Brethauer S, Mattar SG, Titchner RL, Gourash W, Schauer PR. Laparoscopic sleeve gastrectomy for super-obese patients: fortyeight percent excess weight loss after 6 to 8 years with 93% followup. Ann Surg 2012;256:262–5. [4] Brethauer SA, Hammel JP, Schauer PR. Systematic review of sleeve gastrectomy as staging and primary bariatric procedure. Surg Obes Relat Dis 2009;5:469–75. [5] Carlin AM, Zeni TM, English WJ, et al. The comparative effectiveness of sleeve gastrectomy, gastric bypass, and adjustable gastric banding procedures for the treatment of morbid obesity. Ann Surg 2013;257:791–7. [6] Schauer PR, Bhatt DL, Kirwan JP, et al. Bariatric surgery versus intensive medical therapy for diabetes—3-year outcomes. N Engl J Med 2014;370:2002–13. [7] Brethauer SA, Aminian A, Romero-Talamás H, et al. Can diabetes be surgically cured? Long-term metabolic effects of bariatric surgery in obese patients with type 2 diabetes mellitus. Ann Surg 2013;258: 628–37. [8] Pomp A. Safety of bariatric surgery. Lancet Diabetes Endocrinol 2014;2:98–100. [9] Aminian A, Brethauer SA, Kirwan JP, Kashyap SR, Burguera B, Schauer PR. How safe is metabolic/diabetes surgery? Diabetes Obes Metab Epub 2015;17:198–201. [10] Finks JF, Kole KL, Yenumula PR, et al. Predicting risk for serious complications with bariatric surgery: results from the Michigan Bariatric Surgery Collaborative. Ann Surg 2011;254:633–40. [11] Ramanan B, Gupta PK, Gupta H, Fang X, Forse RA. Development and validation of a bariatric surgery mortality risk calculator. J Am Coll Surg 2012;214:892–900. [12] Gupta PK, Franck C, Miller WJ, Gupta H, Forse RA. Development and validation of a bariatric surgery morbidity risk calculator using
Sleeve Gastrectomy Risk Calculator / Surgery for Obesity and Related Diseases ] (2015) 00–00
[13]
[14]
[15]
[16]
[17] [18] [19] [20]
the prospective, multicenter NSQIP data set. J Am Coll Surg 2011;212:301–9. DeMaria EJ, Portenier D, Wolfe L. Obesity surgery mortality risk score: proposal for a clinically useful score to predict mortality risk in patients undergoing gastric bypass. Surg Obes Relat Dis 2007;3:134–40. American College of Surgeons. National Surgical Quality Improvement Program Participant User Data Profile. Available from: http:// site.acsnsqip.org/participant-use-data-file/. Accessed: May 15, 2014. Aminian A, Daigle CR, Romero-Talamás H, et al. Risk prediction of complications of metabolic syndrome before and 6 years after gastric bypass. Surg Obes Relat Dis 2014;10:576–82. Benotti P, Wood GC, Winegar DA, et al. Risk factors associated with mortality after Roux-en-Y gastric bypass surgery. Ann Surg 2014;259: 123–30. Kim JH, Wolfe B. Bariatric/metabolic surgery: short- and long-term safety. Curr Atheroscler Rep 2012;14:597–605. Ismael H, Horst M, Farooq M, et al. Adverse effects of preoperative steroid use on surgical outcomes. Am J Surg 2011;201:305–8. Livingston EH. Development of bariatric surgery-specific risk assessment tool. Surg Obes Relat Dis 2007;3:14–20. Nguyen NT, Nguyen B, Smith B, et al. Proposal for a bariatric mortality risk classification system for patients undergoing bariatric surgery. Surg Obes Relat Dis 2013;9:239–46.
7
[21] Turner PL, Saager L, Dalton J, et al. A nomogram for predicting surgical complications in bariatric surgery patients. Obes Surg 2011;21:655–62. [22] Maciejewski ML, Winegar DA, Farley JF, et al. Risk stratification of serious adverse events after gastric bypass in the Bariatric Outcomes Longitudinal Database. Surg Obes Relat Dis 2012;8: 671–7. [23] Wells BJ, Kattan MW, Cooper GS, et al. ColoRectal Cancer Predicted Risk Online (CRC-PRO) calculator using data from the MultiEthnic Cohort Study. J Am Board Fam Med 2014;27:42–55. [24] Longitudinal Assessment of Bariatric Surgery (LABS) Consortium, Flum DR, Belle SH, King WC, et al. Perioperative safety in the longitudinal assessment of bariatric surgery. N Engl J Med 2009;361: 445–54. [25] Smith MD, Patterson E, Wahed AS, et al. Relationship between surgeon volume and adverse outcomes after RYGB in Longitudinal Assessment of Bariatric Surgery (LABS) study. Surg Obes Relat Dis 2010;6:118–25. [26] Cohen ME, Bilimoria KY, Ko CY, et al. Development of an American College of Surgeons National Surgery Quality Improvement Program: morbidity and mortality risk calculator for colorectal surgery. J Am Coll Surg 2009;208:1009–16.