The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs.
Cirrhosis is a risk factor for postoperative morbidity and mortality after general surgical procedures. However, the impact of cirrhosis on outcomes of surgical resection for gastrointestinal (GI) malignancies has not been described. The authors' objective was to characterize early postoperative and transitional outcomes in cirrhotic patients undergoing GI cancer surgery.
Query of the National Inpatient Sample Database (2005-2008) identified 106,729 patients who underwent resection for GI malignancy; 1479 (1.4%) had cirrhosis. The association of cirrhosis with postoperative outcomes was examined. The primary outcome measure was in-hospital mortality. Secondary outcomes included length-of-stay (LOS) and discharge to long-term care facility (LTCF).
Cirrhotic patients had higher risk of in-hospital mortality (8.9% vs 2.8%, P < .001), longer LOS (11.5 ± 0.26 vs 10.0 ± 0.03 days, P < .001), and higher rate of discharge to LTCF (19.0% vs 15.7%, P < .001). Mortality was highest in patients with moderate to severe liver dysfunction (21.5% vs 6.5%, P < .001). On multivariate analysis, cirrhosis was an independent predictor of in-hospital mortality (odds ratio [OR], 3.0; 95% confidence interval [CI] 2.5-3.7) and nonhome discharge (OR, 1.7; 95% CI, 1.4-2.0). In cirrhotic patients, moderate to severe liver dysfunction was the only independent predictor of in-hospital mortality (OR, 4.03; 95% CI, 2.7-5.9), but did not predict discharge disposition.
The incidence and prevalence of cirrhosis has been increasing in the United States for the past 4 decades.1, 2 A significant proportion of these cases are related to chronic hepatitis C virus (HCV) infection,2 and it is projected that the prevalence of HCV-related cirrhosis will continue to increase until 2040.1, 3 The mortality from of cirrhosis and chronic liver failure, however, continues to decline.4 As a result, cirrhotic patients are becoming increasingly more prone to age-related, nonhepatic surgical diseases. Among the most commonly encountered general surgical diseases in this population have been abdominal wall hernias and cholelithiasis, both of which are amenable to operative treatment in well-compensated patients with low morbidity. However, given the generally benign nature of these conditions, nonsurgical treatment and/or observation are reasonable alternatives in patients awaiting medical optimization or for those with more advanced liver disease who will never be surgical candidates.5, 6
With advancing age, cirrhotics are also becoming increasingly more prone to gastrointestinal (GI) malignancies. Unlike the more common general surgical diseases, GI cancers almost always require relatively urgent surgery for definitive/curative treatment, and effective nonsurgical therapeutic alternatives are limited. As a result, surgical oncologists are often faced with the challenging decision of either performing major operations on these high-risk patients, or refusing treatment altogether for otherwise curable neoplasms.
Traditionally, cirrhosis has been considered a contraindication to major abdominal surgery. Mortality rates have been reported as high as 17.5% to 38% for cirrhotic patients undergoing any abdominal operation,7-9 with higher rates noted for emergent procedures and for patients with coexisting portal hypertension.10 Most studies, however, have been performed in single centers, with small sample sizes and/or in heterogeneous populations of patients.11-14 The impact of cirrhosis on morbidity and mortality specifically for major abdominal cancer resections has not been well described. In addition, the risks of adverse transitional outcomes in cirrhotic patients who do survive surgery have not been described in the GI cancer population.
Given this lack of data, we conducted a population-based study to describe the surgical risk specifically in cirrhotic patients undergoing major abdominal resections for GI malignancies. We hypothesized that: 1) patients with cirrhosis would have an increased risk of mortality and adverse transitional outcomes, 2) a higher-risk population could be identified in whom surgery should be avoided, and 3) predictors of mortality and transitional outcomes could be described.
MATERIALS AND METHODS
The National Inpatient Sample (NIS) database, developed by the Agency for Healthcare Research and Quality as part of the Healthcare Cost and Utilization Project (HCUP), was purchased for the years 2005-2008 from the HCUP Central Distributor after completing a standard Data Use Agreement. The NIS is the largest inpatient database in the United States, incorporating roughly 8 million hospital stays annually, with participation currently encompassing 42 states and approximately 20% of all hospital discharges nationally. It contains patient-specific information including demographic data, admission diagnoses, comorbidities, procedures performed, in-hospital mortality, length of stay (LOS), and discharge disposition, as well as data pertaining to hospital characteristics.
Sample Population/Data Collection
The NIS database was queried for all patients with a diagnosis of solid GI malignancy who underwent major cancer resection for the corresponding organ system. International Classification of Diseases, 9th Edition, Clinical Modification (ICD-9-CM) coding and definitions are presented in Table 1.
Table 1. ICD-9-CM Coding Schemes Used in the Selection of Patients and Definition of Variables
Abbreviations: GI, gastrointestinal; ICD-9-CM, International Classification of Diseases, 9th Edition, Clinical Modification; NIS, National Inpatient Sample.
Major GI resection
Procedure 1 or 2
424, 424.0, 424.1, 424.2, 439.9
435, 436, 437, 438.1, 439, 439.1
457, 457.1, 457.3-457.6, 457.9, 458, 458.1-458.3
485, 485.0-485.2,485.9, 486,486.1-486.4,486.9
Bile duct resection
525.1-525.3, 525.9, 526, 527
Two codes from different categories above
Duodenum and pancreas
571.2, 571.5, 571.6
Cirrhosis + any of following: 572.2, 572.3, 572.4, 572.8, 456.0, 456.1, 456.2, 456.20, 456.21
Cirrhosis without any of the above high-risk diagnosis codes
The main risk factor of interest was the presence of cirrhosis (Table 1). Demographic information collected included age and sex, hospital data included the teaching status of the hospital (teaching vs nonteaching), and procedure type was based on organ system as detailed in Table 1. Comorbidities were described using the Charlson-Deyo Comorbidity Index for administrative data using ICD-9-CM codes, and were categorized into 4 groups (0, 1, 2, ≥3) as previously described.15 Malignancy and liver disease were not included in the comorbidity score, given that these factors were either present in all patients (malignancy) or were already accounted for by another variable in our analysis (cirrhosis).
The outcomes studied were in-hospital mortality, defined as death during hospitalization for the index cancer resection, LOS during the index hospitalization, and discharge disposition in surviving patients. Discharge disposition was categorized as either discharge to home or discharge to other facility (rehabilitation, skilled nursing, or long-term acute care facility).
The primary outcome measure of interest was in-hospital mortality. Secondary outcomes analyzed included LOS and discharge disposition in surviving patients. The primary risk factor of interest was the presence or absence of cirrhosis as coded above. Demographic, procedural, comorbidity, hospital, and outcome data were compared between groups using the Student's t-test for continuous variables and the chi-square test for categorical variables. Multivariate logistic regression analysis was performed to determine the association of cirrhosis with in-hospital mortality controlling for other significant clinical factors. A separate logistic regression analysis was performed using discharge disposition as the outcome of interest.
To identify independent predictors of outcome specifically in cirrhotic patients, multivariate logistic regression analysis was performed in the cirrhosis subpopulation alone. Cirrhosis was categorized by degree of liver dysfunction as outlined in Table 1, with the moderate to severe dysfunction group representing those patients with a diagnosis of portal hypertension or its sequelae.
Summary statistics for continuous variables are presented as means with standard deviations or standard errors of the mean for comparison tests, and proportions for categorical variables. Results of regression analyses are presented as odds ratios (ORs) with 95% confidence intervals (CIs), along with P values for the appropriate statistical tests of significance. In all cases, a P value of <.05 was considered statistically significant. As recommended by the NIS, a complex samples analysis was also performed adjusting for NIS survey sampling characteristics with no significant change in results. SPSS Statistics 17.0.0 (SPSS Inc., Chicago, Ill) was used for all analyses.
In all, 106,729 patients who underwent major abdominal resection specifically for GI malignancy were identified; 1479 (1.4%) had a diagnosis of cirrhosis. Demographic and procedural data for the entire population and a comparison by presence of cirrhosis are presented in Table 2. The mean age of the entire population was 66.7 ± 13.3 years, and sex was nearly equally distributed (51.9% male). The majority of patients (59.5%) had no additional comorbidities, with only a minority (5.6%) having severe comorbidities (Charlson-Deyo Comorbidity Index ≥3) present at the time of hospitalization. The large majority of operations performed were colorectal resections (81.8%). Other major resections represented smaller but significant groups of patients, with the exception of the extremely small number of patients undergoing bile duct resection (0.01%).
Table 2. Characteristics of the Study Population and Comparison by Presence of Cirrhosis, N=106,729
Abbreviations: CBD, common bile duct; SD, standard deviation.
Age, y ± SD
66.7 ± 13.3
67.8 ± 0.04
66.2 ± 0.30
Charlson/Deyo Comorbidity Index
On univariate comparison, cirrhotics were younger (66.2 ± 0.30 vs 67.8 ± 0.04 years), were more likely to be male (61.3% vs 51.8%), had higher comorbidity scores, and were significantly more likely to have operations performed at a teaching hospital (62.7% vs 49.4; all P values <.001). There was also a significant difference in procedure type between cirrhotic and noncirrhotic patients, with the general trend representing a higher rate of liver resections in the cirrhotic group (29.4% vs 1.5%; P < .001).
Table 3 presents a comparison of primary and secondary outcomes by presence of cirrhosis. Cirrhotics had a significantly higher rate of in-hospital mortality (8.9% vs 2.8%), a longer mean LOS (11.5 ± 0.26 vs 10.0 ± 0.03 days), and a higher proportion of surviving patients discharged to nonhome facilities (19.0 vs 15.7%; all P values <.001).
Table 3. Comparison of Short-Term and Transitional Outcomes by Presence of Cirrhosis
No, n = 105,250
Yes, n = 1479
Abbreviation: SEM, standard error of the mean.
Length of stay, mean ± SEM
10.0 ± 0.03
11.5 ± 0.26
Disposition in surviving patients
The impact of cirrhosis on mortality stratified by procedure is presented in Figure 1. There was a significant increase in risk of death in cirrhotics undergoing any operation (all within-procedure P values <.05), with the highest absolute risk of death noted in cirrhotics undergoing gastrectomy (12.7%) and esophagectomy (15.6%).
On univariate logistic regression analysis (Table 4), cirrhosis was significantly associated with increased in-hospital mortality (OR, 3.4; 95% CI, 2.9-4.1; P < .001). After controlling for age, sex, comorbidity score, procedure, and hospital teaching status, cirrhotics still had a 3-fold increased risk of mortality compared with noncirrhotic patients (OR, 3.0; 95% CI, 2.5-3.7; P < .001). Table 4 also presents similar univariate and multivariate logistic regression analyses with discharge disposition as the dependent outcome variable. On both univariate and multivariate analyses, cirrhosis was associated with an increased risk of nonhome discharge, up to 70% greater than baseline when adjusting for other factors (OR, 1.7; 95% CI, 1.4-2.0; P < .001).
Table 4. Logistic Regression Analysis Evaluating the Effect of Cirrhosis on In-Hospital Mortality and Discharge Disposition Using Univariate and Multivariate Analyses
After stratifying the cirrhotic group by degree of liver dysfunction, cirrhotic patients with moderate to severe liver dysfunction were noted to have a 3-fold increase in mortality compared with those with mild dysfunction (21.5% vs 6.5%, respectively; P < .001). However, there was no difference in discharge disposition by degree of liver dysfunction (19.0% vs 19.0%; P = 1.0). On multivariate analysis (Table 5), moderate to severe liver dysfunction was associated with a 4-fold increase in mortality (OR, 4.0; 95% CI, 2.7-5.9; P < .001). Age, hospital teaching status, and Charlson-Deyo Comorbidity Index were not independently predictive of in-hospital mortality.
Table 5. Multivariate Logistic Regression of In-Hospital Mortality and Discharge Disposition (Nonhome Discharge) in Cirrhotic Patients Only (n=1479)
Degree of liver dysfunction was not associated with nonhome discharge (OR, 1.09; 95% CI, 0.72-1.65; P = .67 for moderate to severe disease), whereas increasing age (OR, 1.07; 95% CI, 1.05-1.08; P < .001) and Charlson-Deyo Comorbidity Index (OR, 2.5; 95% CI, 1.5-4.2; P < .001 for Charlson-Deyo Comorbidity Index ≥3) were independently predictive of a higher likelihood of nonhome discharge. Teaching hospitals were less likely to discharge patients to a long-term facility (OR, 0.61; 95% CI, 0.45-0.81; P = .001).
With the increasing prevalence and survival associated with HCV-related cirrhosis, the incidence of solid GI malignancies in cirrhotics is likely to increase. As in all patients, a significant proportion of these malignancies are potentially curable by surgical resection.
However, individuals with cirrhosis represent a high-risk population in whom surgery is often avoided because of concerns about perioperative complications and mortality. As a result, a significant number of patients may be denied appropriate therapy for potentially curable neoplasms.
The extent of risk is not uniform in the cirrhotic population, and numerous scoring systems have been developed to more accurately quantify risk in individual patients.16, 17 Although there are some inconsistencies, the general rule is that increasing severity of liver dysfunction is associated with greater risk. The difficulty with applying predictive indices broadly, however, is that they have been for the most part studied in small populations, single centers, and/or in heterogeneous surgical cohorts. Whether, and to what degree, these data are applicable to the cancer population is, therefore, unclear. An added concern is that, unlike the more common general surgical diseases encountered in cirrhotics (ie, abdominal wall hernias and cholelithiasis), effective nonsurgical alternatives are generally not available for solid GI malignancies. As a result, oncologic surgeons are often faced with the decision of either providing suboptimal care for otherwise curable neoplasms, or performing major operations on high-risk patients with limited applicable data.18-20
The purpose of our study was to describe operative risk in a more uniform population of patients undergoing major resections specifically for GI malignancies. We have demonstrated that major abdominal operations in patients with cirrhosis are associated with a significantly greater risk of adverse events, including a 3-fold increase in the risk of postoperative mortality compared with baseline after accounting for comorbidity and other relevant factors. The impact of cirrhosis was significant for all major abdominal cancer operations. However, the highest absolute risk of operative mortality was noted in cirrhotic patients undergoing upper GI operations (gastrectomy and esophagectomy), possibly as a result of the increased complexity of these procedures (which often require exposure of >1 body cavity) and/or because of the postoperative redistribution of portal venous blood flow. In cirrhotic patients, moderate to severe liver dysfunction was an independent predictor of 4-fold greater mortality compared with the presence of mild dysfunction, whereas age and comorbid conditions were not associated with increased mortality in this group.
Similar data were recently presented by Csikesz et al,10 who demonstrated increased in-hospital mortality in a heterogeneous population of cirrhotics undergoing 4 index elective operations, including general surgical, vascular, and cardiac procedures. The highest mortality, as in our study, was noted in patients with severe liver dysfunction/portal hypertension. Teh et al18 and Northup et al21 have also described the importance of the degree of liver dysfunction on postoperative mortality for general, cardiothoracic, urologic, and orthopedic surgeries. According to these studies, increasing Model for End Stage Liver Disease scores as well as American Society of Anesthesiologists Class and age were associated with increased risk of adverse events.
Data are relatively limited and somewhat conflicting, however, in cancer-only populations. Gervaz et al11 have demonstrated increased postoperative mortality (13%) in cirrhotic patients undergoing colorectal cancer surgery at a single institution, with increasing mortality observed in the presence of worsening liver dysfunction. Conversely, a Japanese group12 have reported extremely low mortality in cirrhotic patients undergoing gastric cancer surgery, although significantly increased blood loss was noted in patients with higher Child-Pugh class.
Considering these data collectively with our own, major abdominal oncologic surgery in patients with sequelae of portal hypertension should be avoided, particularly in patients with upper GI cancers. Conversely, the subset of patients with mild liver dysfunction may represent a relatively low-risk group, independent of age and comorbid conditions, although the risk of adverse perioperative events is still higher than baseline. In the absence of other surgical contraindications, surgery can be offered reasonably to these patients, who have potentially curable malignancies, provided that the patients are appropriately counseled preoperatively with respect to their risk profile.
In addition, we have demonstrated that, with respect to transitional outcomes and disposition, cirrhotic patients who survived their cancer operation were significantly more likely to be discharged to a long-term care facility compared with noncirrhotic patients. This finding held true for both patients with mild and moderate to severe liver dysfunction and persisted even when controlling for other significant clinical factors such as age and comorbidity. Interestingly, in the cirrhotic subpopulation the degree of liver dysfunction did not increase the risk for nonhome discharge, which was more a function of age and comorbidity. The study of transitional outcomes is a more recent topic of research, and therefore data are limited in this particular population. Other groups have demonstrated adverse transitional outcomes in the elderly, in patients with significant comorbidities, and in patients with poor functional status.22-24 These studies, along with our own, underscore the need to plan for long-term care in this and other high-risk populations, and to include the likelihood of nonhome discharge disposition in the preoperative conversation, particularly in elderly patients with extensive comorbidity.
Our data should be interpreted in light of several limitations, which are common to most retrospective analyses using administrative data. First, given limitations inherent to the NIS administrative database, we could not differentiate between emergent and elective procedures, which may be significant given that the urgency of an operation is known to affect outcomes in cirrhotic patients. However, our coding method, which allowed inclusion only of patients whose primary diagnosis was a malignancy (as opposed to an acute intra-abdominal event) and whose primary operation was a specified major cancer resection, likely significantly limited the inclusion of patients undergoing emergent procedures. In addition, there is an element of selection bias in our study, with the study group likely consisting of a cautiously selected group of patients. Therefore, our results may underestimate risk. Hence, a comparable degree of caution should be used in applying our data and in selecting even well-compensated cirrhotics for surgery. Another limitation specific to the database is the lack of laboratory information, which precluded the ability to stratify based on the more commonly used risk indices in cirrhotics.
In conclusion, major abdominal resections for GI malignancy in patients with cirrhosis are associated with worse postoperative short-term and transitional outcomes. Those with moderate to severe liver dysfunction and portal hypertension are at particularly increased risk of postoperative death, whereas those with mild liver dysfunction may represent a low-risk group in whom curative operations are feasible and relatively safe. Although there is still increased risk of morbidity and mortality, the proper selection of cirrhotic patients with mild liver dysfunction for GI oncologic resections can result in good outcomes. These data should be incorporated into the decision-making process when considering curative oncologic resection in this high-risk population.
The authors acknowledge financial support and/or resources in the preparation of this article from the Houston Health Services Research and Development Center of Excellence (HF P90-020) at the Michael E. DeBakey Veterans Administration Medical Center.