TMCnet News
Model for End-stage Liver Disease (MELD) in Predicting Postoperative Mortality of Patients Undergoing Colorectal Surgery [American Surgeon, The](American Surgeon, The Via Acquire Media NewsEdge) The Model for End-stage Liver Disease (MELD) score was previously shown to predict perioperative mortality in patients with cirrhosis undergoing a variety of nontransplant surgical procedures. We sought to determine its usefulness in predicting postoperative mortality in patients undergoing colorectal procedures. National Surgical Quality Improvement Program data were gathered for adult patients undergoing elective and emergent colorectal procedures (Current Procedural Terminology codes 44005 through 45563 excluding appendectomy) during 2005 and 2006 at participating centers. The preoperative MELD score was calculated for all patients and assessed using logistic regression modeling. A total of 10,033 patients met study inclusion criteria. Overall 30-day mortality was 6.6 per cent. In all patients undergoing colorectal surgery, MELD was an independent predictor of mortality (2.95 [2.27 to 3.84]). Other independent predictors included age, functional status, American Society of Anesthesiologists classification, ascites, esophageal varices, disseminated cancer, chronic steroid use, cardiac disease, renal failure, malnutrition, sepsis, emergency, and ventilator dependence. The MELD score is an independent predictor of mortality in patients undergoing colorectal procedures. These data can be used to assign risk and assist in clinical decision-making. THE MODEL FOR END-STAGE Liver disease (MELD) was originally described as a prognostic indicator for patients undergoing a transjugular intrahepatic portosystemic shunt procedure.1 It has subsequently been validated as a measure of short-term mortality risk in patients with end-stage liver disease.2 Previously, MELD has been shown to predict 30- and 90-day mortality in patients with cirrhosis undergoing a variety of nontransplant surgical procedures.3, 4 Prior studies demonstrating the ability of MELD to predict postoperative mortality included a variety of procedures with variable risk. The purpose of the current study was to assess the ability of MELD to predict 30- day mortality in patients undergoing intra-abdominal colorectal procedures using a large national database. Methods Study Design The Institutional Review Board at the University of Virginia Health System approved this study. An observational multi-institutional cohort study was conducted through query of the participant use data file available to all participants of the National Surgical Quality Improvement Program (NSQIP). Data were analyzed on adult patients (older than age 18 years) undergoing colorectal procedures identified by the Current Procedural Terminology codes 44005 through 45563 (excluding appendectomy) during 2005 and 2006. This included both elective and emergent cases. NSQIP methodology has been previously described5, 6 and includes national data from a random sampling of general surgery and vascular procedures performed at participating academic and community centers (193 participating centers at the time of query). Over 100 variables are collected and entered into the database by clinically trained staff. These variables include demographic data, patient comorbidities, select 30-day preoperative laboratory values, operative type, and 30-day morbidity, and mortality. For a complete listing of the NSQIP database variables, please refer to the user guide found on the NSQIP web site (www.acsnsqip.org). The presence of cirrhosis is not one of the data points collected in the participant use data file of the NSQIP database. We used the presence of ascites or esophageal varices as surrogate markers to identify patients with potential liver disease. As such, the patients with ascites or esophageal varices were selected and analyzed separately. MELD Calculation Preoperative MELD was calculated using preoperative values of creatinine (mg/dL), total bilirubin (mg/dL), and international normalized ratio (INR). Patients were excluded from analysis if all three of the included laboratory values were not available. The modified MELD score used by the United Network for Organ Sharing was used in this study:MELD43.78 × loge (bilirubin in mg/dL) + 11.2 × loge (INR) + 9.57 × loge (creatinine in mg/dL) + 6.43. A bilirubin, creatinine, or INR value of less than 1.0 was rounded to 1.0 and the maximum creatinine allowed was 4.0 mg/dL. All patients on dialysis received a creatinine of 4.0 mg/dL. The MELD score was then rounded to the nearest integer. A score less than 10 is considered in the normal range. MELD can be calculated online at www. mayoclinic.org/meld/mayomodel6.html. Statistical Analysis Univariate analysis was initially conducted to identify demographic and clinically relevant variables associated with 30-day mortality. Continuous variables were compared using Student's t test and categorical variables were compared using the Fisher's exact test. Based on these results, backward stepwise multivariable regression modeling was used to identify significant independent risk factors for mortality. A significance level of 0.05 was used for entry into the model. A random sample of 80 per cent of patients was used for model creation; the remaining cases were used for validation. Minimal attenuation (less than 2%) of statistical performance was observed during internal validation. The capacity of the models to discriminate between high-risk and low-risk subjects was measured using the C statistic. A C statistic value of 0.5 indicates that the model provides no predictive discrimination, whereas a value of 1.0 indicates perfect discrimination between patients who lived and died. The fitness of the model was assessed using the Nagelkerke R2 statistic, which ranges from 0 for models that provide no predictive information to 1 for models that predict perfectly. The predicted probability of death generated by the logistic regression model was stratified by MELD score to evaluate the ability of MELD to predict mortality. All P values are two-tailed. P values # 0.05 are considered statistically significant. Except as otherwise specified, values expressed are mean ± standard error or percentage of the group of origin. Statistical analyses were performed with SPSS software, Version 16.0. Results Patient Selection Thirty four thousand four hundred ninety-four patients met criteria for inclusion. Of these, MELD could be calculated in 10,033 (30%). Demographic variables are demonstrated in Table 1. The average MELD score for the patient population was 9.5 ± 0.04 and the distribution of patients by MELD score is seen in Figure 1. The vast majority of patients had normal MELD scores; 69.8 per cent of patients had MELD scores less than 10. Overall 30-day mortality for patients with a MELD score less than 10 was 2.9 per cent. Mortality stratified by MELD score is demonstrated in Figure 2. There was a steep incremental increase in mortality with increasing MELD score. The procedure types are demonstrated in Table 2. Fourteen per cent were laparoscopic procedures and the remaining 86 per cent were laparotomies. Predictors of Mortality The variables listed in Table 3 were associated with mortality through univariate analysis. Multivariate analysis with backward stepwise logistic regression using the variables listed in this table as independent variables was conducted with all-cause 30-day mortality as the dependent variable. The results of this analysis are demonstrated in Table 4. Age, MELD score, functional status before the illness, American Society of Anesthesiologists classification, the presence of ascites, esophageal varices, disseminated cancer, chronic steroid use, coronary artery disease, acute renal failure, greater than 10 per cent weight loss within six months of surgery, preoperative sepsis, emergency procedures, and ventilator dependence were independent predictors of mortality. The Nagelkerke R2 was 0.366. Predicted Probability of Death The predicted mortality for each patient was calculated using the previous multivariate regression models and stratified by the MELD score for all patients and patients with esophageal varices and ascites. This analysis was conducted for all patients and stratified by elective and emergent procedures given the inherent differences in these two patient populations. As demonstrated in Figure 3A (elective cases) and 3B (emergent cases), there is a strong correlation between predicted mortality and MELD score. Comparing emergent (Fig. 3B) and elective (Fig. 3A) procedures, the correlation is more pronounced with emergent procedures.With regard to elective procedures (Fig. 3A), there is a roughly 0.5 per cent risk of death with each point in the MELD score up to 20 where the risk of death is approximately 1 per cent per point thereafter. This estimation is slightly higher per MELD point in the presence of ascites or esophageal varices. The association between mortality and MELD score is increased in patients undergoing emergent procedures with a one-to-one ratio up to a MELD of 15 and a roughly two-to-one ratio of mortality per MELD score thereafter (Fig. 3B). Discussion In this large multi-institutional observational study, MELD was found to be an independent predictor of mortality in all patients undergoing colorectal surgery. This is similar to other studies demonstrating the predictive ability of MELD in predicting postoperative mortality.3, 4 However, unlike prior studies, this study included patients without known liver disease. Although not designed for use in patients without known liver disease, these data suggest that MELD is a marker for the severity of illness in a variety of patients. There are other more sophisticated prognostic tools for predicting mortality in patients undergoing colorectal surgery.7, 8 However, these are highly complex and not easily calculated at the bedside. Although not as predictive as the more sophisticated models, the MELD score is an easily computable value that can be calculated quickly and discussed with patients in the office or hospital setting. Based on Figure 3, clinicians could calculate the MELD score and quote patients an estimated mortality when undergoing elective colon and rectal surgery. These data were obtained from a large national database with associated limitations in data collection. Patients without calculable MELD scores were excluded from analysis. Exclusion of these patients may have skewed these data. Additionally, because liver disease was not collected as a data point, we were unable to identify patients with cirrhosis. Although ascites and esophageal varices aremarkers of liver disease, they are certainly not specific. It may be that in patients with cirrhosis undergoing colorectal surgery, the mortality distribution is different from the data presented in this article. However, it is interesting to note that the mortality curves in patients with ascites or esophageal varices (Fig. 3) are virtually identical to the data presented by Northup et al.3 in patients with known cirrhosis undergoing intra-abdominal surgery. This would suggest that the two patient populations are similar. There are several other implications that may be inferred from these data. Particularly for patients presenting for elective procedures, these data may be highly useful in operative planning. For instance, based on these data, the patient with a MELD of 20 presenting with an endoscopically unresectable polyp has a 15 to 20 per cent risk of death from surgery. One must ponder whether this outweighs the risk of malignant transformation over the expected lifespan of this high-risk patient. We chose MELD as the focus of this article because it is easily calculated and can give clinicians a quantifiable risk to quote patients. However, several other variables were found to independently predict mortality as demonstrated in Table 4. These variables are similar to those reported in the literature.9 Most factors are nonmodifiable. However, several may provide an opportunity for intervention before elective surgery. For instance, preoperative withdrawal of steroids, if appropriate, provision of nutrition support, and optimization of renal disease are potential modifiable risks. Whether these interventions would translate into improved outcomes is beyond the scope of the current study. Conclusion Based on a large data set of over 10,000 patients undergoing intra-abdominal colorectal procedures, the MELD score was found to be an easily calculated independent predictor of 30-day mortality. These data may assist in clinical decision-making and provide additional avenues of further study for risk stratification in patients undergoing colorectal surgery. REFERENCES 1. Malichoc M, Kamath PS, Gordon FD, et al. A model to predict poor survival in patients undergoing transjugular intrahepatic portosystemic shunts. Hepatology 2000;31:864-71. 2. Kamath PS, Wiesner RH, Malinchoc M, et al. A model to predict survival in patients with end-stage liver disease. Hepatology 2001;33:464-70. 3. Northup PD, Wanamaker RC, Lee VD, et al. Model for End-Stage Liver Disease (MELD) predicts nontransplant surgical mortality in patients with cirrhosis. Ann Surg 2005;242: 244-51. 4. Teh SH, Nagorney DM, Stevens SR, et al. Risk factors for mortality after surgery in patients with cirrhosis. Gastroenterology 2007;132:1261-9. 5. Englesbe MJ, Peeletier SJ, Magee JC, et al. Seasonal variation in surgical outcomes as measured by the American College of Surgeons-National Surgical Quality Improvement Program (ACSNSQIP). Ann Surg 2007;246:456-62. 6. Dimick JB, Chen SL, Taheri PA, et al. Hospital costs associated with surgical complications: a report from the private sector National Surgical Quality Improvement Program. J Am Coll Surg 2004;199:531-7. 7. Bowles TA, Sander KM, Colson M, Watters DA. Simplified risk stratification in elective colorectal surgery. ANZ J Surg 2008; 78:24-7. 8. Oomen JL, Cuesta MA, Engel AF. Comparison of outcome of POSSUM, p-Possum, and cr-POSSUM scoring after elective resection of the sigmoid colon for carcinoma or complicated diverticular disease. Scand J Gastroenterol 2007;42: 841-7. 9. Slim K, Panis T, Alves A, et al. Predicting postoperative mortality in patients undergoing colorectal surgery. World J Surg 2006;30:100-6. TRACI L. HEDRICK, M.D.,* BRIAN R. SWENSON, M.D., M.S.,[dagger] CHARLES M. FRIEL, M.D.* From the *Department of Surgery, University of Virginia, Charlottesville, Virginia; and [dagger]St. John's Clinic, Springfield, Missouri Presented at the Annual Meeting of the American Society of Colon and Rectal Surgeons, Boston, MA, June 7-11, 2008. Address correspondence and reprint requests to Traci L. Hedrick, M.D., Assistant Professor of Surgery, Department of Surgery, University of Virginia Health System, P.O. Box 800709, Charlottesville, VA 22908. E-mail: [email protected]. (c) 2013 Southeastern Surgical Congress |