In-hospital mortality after resection of biliary tract cancer in the United States
Abstract
Objective:
To assess perioperative mortality following resection of biliary tract cancer within the U.S.
Background:
Resection remains the only curative treatment for biliary tract cancer. However, current data on operative mortality after surgical resections for biliary tract cancer are limited to small and single-center studies.
Methods:
Using the Nationwide Inpatient Sample 1998–2006, a cohort of patient-discharges was assembled with a diagnosis of biliary tract cancer, including intrahepatic bile duct, extrahepatic bile duct, and gall bladder cancers. Patients undergoing resection, including hepatic resection, bile duct resection, pancreaticoduodenectomy, and cholecystectomy, were retained. The primary outcome measure was in-hospital mortality. Categorical variables were analyzed by chi-square. Multivariable logistic regression was performed to identify independent predictors of in-hospital mortality following resection.
Results:
31 870 patient-discharges occurred for the diagnosis of biliary tract cancer, including 36.2% intrahepatic ductal, 26.7% extrahepatic ductal, and 31.1% gall bladder. Of the total, 18.6% underwent resection: mean age was 69.3 years (median 70.0); 60.8% were female; 73.7% were white. Overall inpatient surgical mortality was 5.6%. Independently predictive factors of mortality included patient age ≥50 (vs. <50; age 50–59 odds ratio [OR] 5.51, 95% confidence interval [CI] 1.70–17.93; age 60–69 OR 7.25, 95% CI 2.29–22.96; age ≥ 70 OR 9.03, 95% CI 2.86–28.56), the presence of identified comorbidities (congestive heart failure, OR 3.67, 95% CI 2.61–5.16; renal failure, OR 4.72, 95% CI 2.97–7.49), and admission designated as emergent (vs. elective; OR 1.82, 95% CI 1.39–2.37).
Conclusion:
Increased in-hospital mortality for patients undergoing biliary tract cancer resection corresponded to age, comorbidity, hospital volume, and emergent admission. Further study is warranted to utilize these observations in promoting early detection, diagnosis, and elective resection.
Introduction
After hepatocellular cancer, cholangiocarcinoma is the second most common hepatobiliary malignancy.1 The incidence of the different types of cholangiocarcinoma has shifted since 1975, with an increase in intrahepatic cholangiocarcinoma and a slight decrease in extrahepatic cholangiocarcinoma, possibly as a result of an increase in chronic liver disease, hepatitis C and primary sclerosing cholangitis.2 Medical therapy is not effective; resection provides the only cure.3,4
Multiple single-centre studies have examined outcomes of surgical treatment of biliary tract cancer. Most have been small retrospective studies given the low prevalence of the disease, and have demonstrated that a minority of patients with gall bladder and ductal cancer have resectable disease, and that curative resection is associated with the best survival benefit.5–7 The major goal of this study was to examine factors associated with in-hospital mortality for biliary tract cancers using a large, population-based database. The Nationwide Inpatient Sample (NIS) was used to determine the effect of covariates on mortality.
Methods and materials
A retrospective, population-based analysis was performed using discharge records from the Nationwide Inpatient Sample (NIS) for the years 1998–2006. The NIS is the largest national all-payer hospital inpatient care database in the United States. Data exists for nearly 8 million hospital discharges per year from a stratified sample of 20% of non-federal US hospitals from participating states, including academic hospitals.
The NIS is supported by the Healthcare Cost and Utilization Project (HCUP) and contains all-payer discharge information for 100% of patient discharges from 1045 hospitals in 38 states (http://www.hcup-us.ahrq.gov/nisoverview.jsp#Data). An NIS-implemented weighting strategy allows population-based estimates to be drawn at the national level. Sampled hospitals are assigned appropriate weights based on the number of hospitals they represent in the database for a given year.8 Statistical analyses were performed on weighted data and reported as weighted frequencies.
Study population
The primary measured outcome for this study was in-hospital mortality, defined as death prior to discharge regardless of cause or time from operation. To identify all patient discharges with the diagnosis of biliary tract cancer, diagnostic and procedural codes were selected from the International Classification of Disease Ninth Revision Clinical Modification (ICD-9 CM).9 The initial cohort was limited to patients with malignant lesions of the biliary tract; those undergoing resection of biliary tract cancers were then identified based on ICD-9 procedure codes (Table 1).
Table 1
International Classification of Disease Ninth Revision Clinical Modification (ICD-9 CM) diagnostic and procedural codes
| Diagnostic Code | Diagnosis | Procedural Code | Procedure |
|---|---|---|---|
| 156.0 | Malignancy of the gallbladder | 52.7 | Pancreaticoduodenectomy, radical |
| 155.1 | Malignancy of intrahepatic bile ducts | 51.63, 51.69 | Bile duct resection (common bile duct or NOS) |
| 156.1 | Malignancy of extrahepatic bile ducts | 51.21, 51.22 | Open cholecystectomy |
| Laparoscopic cholecystectomy | 51.23, 51.24 | ||
| 156.8 | Malignancy of contiguous sites | 50.22 | Partial hepatectomy |
| 230.8 | Carcinoma in situ of the biliary system | 50.3 | Total hepatic lobectomy |
Statistical analysis
All data were analysed using SAS version 9.1 (SAS Institute, Cary, NC) advanced survey procedures. Univariate analyses of categorical variables were performed using χ2-tests, with P-values of <0.05 considered statistically significant. Trend analyses were performed using the Mantel–Haenszel χ2-test. Independent variables included age group (<50, 50–59, 60–69, ≥70), gender, race, admission status (elective vs. emergent), diagnosis type (gall bladder cancer, intrahepatic ductal, extrahepatic ductal and contiguous sites), procedure type (cholecystectomy, hepatic resection, duct resection and pancreaticoduodenectomy), year of resection and hospital type (teaching vs. non-teaching). Analyses also included medical comorbidities from validated Elixhauser comorbidity software for use with national datasets (congestive heart failure, diabetes mellitus, liver disease, obesity and peripheral vascular disease).10
The NIS classifies admission type into multiple categories, including ‘emergency’, ‘urgent’ and ‘elective.’ For these analyses, ‘emergency’ and ‘urgent’ were combined into the variable ‘emergent’ to provide a dichotomous comparison. Definitions of admission type are derived from the UB-04, a universal billing form developed by the American Hospital Association and insurance industry to codify health care data collection; information regarding this form is available through the Centers for Medicare & Medicaid Services website (http://www.cms.hhs.gov/Manuals). ‘Emergency’ admission type is defined as one in which ‘the patient required medical intervention as a result of severe, life-threatening or potentially disabling conditions’. ‘Urgent’ admission type requires ‘immediate attention for care and treatment of a physical or mental disorder’. ‘Elective’ admission type permits ‘adequate time to schedule the availability of a suitable accommodation’. 11
A multivariable logistic regression was assembled with in-hospital mortality as the dependent variable. Unadjusted and adjusted odds ratios were calculated to determine the effects of the identified covariates on the outcome measure of in-hospital mortality. Patient race was excluded from multivariate analysis secondary to prohibitive quantities of missing data. Logistic regression was repeated after excluding cholecystectomy (open and laparoscopic) as a covariate to address bias associated with the likelihood that gall bladder cancers may frequently present and be treated as acute cholecystitis. The adjusted odds ratios in this subgroup were calculated to determine the independent effect of admission type on in-hospital mortality.
Results
Patient characteristics and univariate analyses are presented in Tables 2 and and3,3, respectively. Predictors of in-hospital mortality were examined using a multivariable logistic regression model and are reported in Table 4. Results from the reconstructed multivariable logistic regression model excluding patients having undergone laparoscopic or open cholecystectomy are presented in Table 5.
Table 2
Patient and hospital characteristics for all biliary tract cancer resections, 1998–2006
| Factor | Total | Weighted (%) |
|---|---|---|
| Total No. patients with Biliary tract cancer | 31 870 | 157 266 |
| Resected patients | 5 942 | 29 315 (18.6) |
| Age | ||
| Mean (median) | 69.3 (70.0) | |
| <50 | 479 | 2 366 (8.1) |
| 50–59 | 832 | 4 117 (14.1) |
| 60–69 | 1 363 | 6 749 (23.0) |
| ≥70 | 3 264 | 16 063 (54.8) |
| Gender | ||
| Men | 2 319 | 11 471 (39.1) |
| Women | 3 623 | 17 824 (60.8) |
| Race | ||
| White | 3 333 | 16 479 (73.7) |
| Black | 374 | 1 819 (8.1) |
| Hispanic | 507 | 2 451 (11.0) |
| Asian | 173 | 857 (3.8) |
| Native American | 24 | 120 (0.5) |
| Unknown | 127 | 623 (2.8) |
| Comorbid Conditions | ||
| Congestive heart failure | 2 281 | 11 270 (7.2) |
| Diabetes | 5 499 | 27 093 (17.2) |
| Renal failure | 855 | 4 170 (2.7) |
| Chronic lung disease | 3 398 | 16 723 (10.6) |
| Peripheral vascular disease | 751 | 3 709 (2.4) |
| Obesity | 574 | 2 790 (1.8) |
| Type of Hospital | ||
| Teaching | 3 247 | 16 312 (55.6) |
| Non-teaching | 2 694 | 12 998 (44.2) |
| Admission Type | ||
| Emergency | 2 628 | 12 929 (50.6) |
| Elective | 2 551 | 12 640 (49.4) |
Table 3
Univariate analysis of mortality (weighted): all resections
| Factor | n (%) | P-value |
|---|---|---|
| Overall mortality | 1638 (5.6) | N/A |
| Gender | 0.005 | |
| Men | 763 (2.6) | |
| Women | 874 (3.0) | |
| Age (years) | <0.001 | |
| <50 | 19 (0.1) | |
| 50–59 | 131 (0.4) | |
| 60–69 | 323 (1.1) | |
| ≥70 | 1160 (4.0) | |
| Year of procedure | 0.527 | |
| 1998 | 240 (0.8) | |
| 1999 | 167 (0.6) | |
| 2000 | 218 (0.7) | |
| 2001 | 176 (0.6) | |
| 2002 | 139 (0.5) | |
| 2003 | 193 (0.7) | |
| 2004 | 163 (0.6) | |
| 2005 | 156 (0.5) | |
| 2006 | 185 (0.6) | |
| Diagnosis type | 0.493 | |
| Gall bladder CA | 876 (3.0) | |
| Intrahepatic ductal CA | 221 (0.8) | |
| Extrahepatic ductal CA | 375 (1.3) | |
| Contiguous sites | 166 (0.6) | |
| Procedure type | 0.104 | |
| Cholecystectomy | 1042 (3.6) | |
| Hepatic resection | 281 (1.0) | |
| Duct resection | 92 (0.3) | |
| Pancreaticoduodenectomy | 223 (0.8) | |
| Hospital type | 0.072 | |
| Teaching | 830 (2.8) | |
| Non-teaching | 807 (2.8) | |
| Admission type* | <0.001 | |
| Emergent | 964 (3.6) | |
| Elective | 497 (1.9) | |
| Comorbid conditions | ||
| Renal failure | 181 (0.6) | <0.001 |
| Congestive heart failure | 360 (1.2) | <0.001 |
| Peripheral vascular disease | 58 (0.2) | 0.036 |
| Chronic lung disease | 259 (0.9) | 0.016 |
| Diabetes | 233 (0.8) | 0.247 |
| Obesity | 23 (0.1) | 0.050 |
CA, cancer.
Table 4
Logistic regression of in-hospital mortality
| Factor | Odds ratio (Unadjusted) | 95% confidence interval | Odds ratio (Adjusted) | 95% confidence interval |
|---|---|---|---|---|
| Gender | ||||
| Women (vs. men) | 0.72 | 0.58–0.90 | 0.73 | 0.56–0.95 |
| Age groups (years) | ||||
| 50–59 (vs. <50) | 4.08 | 1.45–11.50 | 5.51 | 1.70–17.93 |
| 60–69 (vs. <50) | 6.27 | 2.29–17.13 | 7.25 | 2.29–22.96 |
| ≥70 (vs. <50) | 9.69 | 3.58–26.27 | 9.03 | 2.86–28.56 |
| Diagnosis type | ||||
| Intrahepatic ductal CA (vs. GBC) | 1.24 | 0.88–1.75 | 1.42 | 0.97–2.09 |
| Extrahepatic ductal CA (vs. GBC) | 1.09 | 0.83–1.44 | 1.44 | 0.95–2.19 |
| Contiguous sites (vs. GBC) | 0.89 | 0.60–1.31 | 1.04 | 0.68–1.59 |
| Procedure type | ||||
| Cholecystectomy (vs. hepatectomy) | 1.07 | 0.79–1.46 | 0.64 | 0.45–0.91 |
| Duct resection (vs. hepatectomy) | 0.79 | 0.47–1.32 | 0.60 | 0.33–1.10 |
| PD (vs. hepatectomy) | 1.18 | 0.79–1.76 | 0.84 | 0.50–1.42 |
| Year of resection | ||||
| 1998 (vs. 2006) | 1.06 | 0.68–1.65 | 1.31 | 0.81–2.14 |
| 1999 (vs.2006) | 0.76 | 0.47–1.22 | 0.81 | 0.47–1.40 |
| 2000 (vs. 2006) | 1.15 | 0.72–1.82 | 1.29 | 0.77–2.16 |
| 2001 (vs. 2006) | 0.87 | 0.54–1.38 | 0.96 | 0.56–1.63 |
| 2002 (vs. 2006) | 0.76 | 0.44–1.30 | 0.92 | 0.53–1.62 |
| 2003 (vs. 2006) | 0.99 | 0.60–1.62 | 1.05 | 0.60–1.83 |
| 2004 (vs. 2006) | 0.78 | 0.47–1.30 | 1.09 | 0.63–1.88 |
| 2005 (vs. 2006) | 0.77 | 0.48–1.24 | 0.79 | 0.47–1.32 |
| Comorbid conditions | ||||
| Congestive heart failure | 4.45 | 3.34–5.92 | 3.67 | 2.61–5.16 |
| Renal failure | 6.71 | 4.52–9.97 | 4.72 | 2.97–7.49 |
| Peripheral vascular disease | 1.91 | 1.04–3.52 | 1.18 | 0.56–2.49 |
| Diabetes | 0.83 | 0.60–1.14 | 0.72 | 0.50–1.03 |
| Chronic lung disease | 1.47 | 1.08–2.05 | 1.17 | 0.81–1.70 |
| Obesity | 0.42 | 0.17–1.03 | 0.51 | 0.19–1.39 |
| Hospital type | ||||
| Non-teaching (vs. teaching) | 1.23 | 0.97–1.54 | 1.0 | 0.78–1.33 |
| Admission type | ||||
| Emergency (vs. elective) | 1.87 | 1.47–2.38 | 1.82 | 1.39–2.37 |
GBC, gall bladder cancer; PD, pancreaticoduodenectomy.
Table 5
Logistic regression of in-hospital mortality, cholecystectomy patients excluded
| Factor | Odds ratio (adjusted) | 95% confidence interval |
|---|---|---|
| Gender | ||
| Women (vs. men) | 0.59 | 0.36–0.95 |
| Age groups (years) | ||
| 50–59 (vs. <50) | 3.43 | 1.02–11.48 |
| 60–69 (vs. <50) | 5.31 | 1.63–17.31 |
| >70 (vs. <50) | 6.12 | 1.85–20.22 |
| Diagnosis type | ||
| Intrahepatic ductal CA (vs. GBC) | 3.45 | 1.70–7.02 |
| Extrahepatic ductal CA (vs. GBC) | 3.87 | 1.68–8.93 |
| Contiguous sites (vs. GBC) | 3.98 | 1.69–9.41 |
| Procedure type | ||
| Duct resection (vs. hepatectomy) | 0.43 | 0.21–0.88 |
| PD (vs. hepatectomy) | 0.59 | 0.31–1.11 |
| Year of resection | ||
| 1998 (vs. 2006) | 0.76 | 0.35–1.62 |
| 1999 (vs.2006) | 0.57 | 0.25–1.27 |
| 2000 (vs. 2006) | 0.83 | 0.38–1.83 |
| 2001 (vs. 2006) | 0.67 | 0.28–1.63 |
| 2002 (vs. 2006) | 0.70 | 0.28–1.73 |
| 2003 (vs. 2006) | 0.75 | 0.32–1.73 |
| 2004 (vs.2006) | 0.66 | 0.30–1.49 |
| 2005 (vs.2006) | 0.58 | 0.27–1.26 |
| Comorbid conditions | ||
| Congestive heart failure | 7.30 | 3.56–14.94 |
| Renal failure | 9.98 | 3.79–26.25 |
| Peripheral vascular disease | 0.54 | 0.09–3.17 |
| Diabetes | 0.37 | 0.15–0.92 |
| Chronic lung disease | 1.09 | 0.53–2.25 |
| Hospital type | ||
| Non-teaching (vs. teaching) | 1.26 | 0.80–1.99 |
| Admission type | ||
| Emergency (vs. elective) | 1.54 | 1.01–2.37 |
Discussion
In the present study, the observed in-hospital mortality rate for all patients undergoing resection for biliary tract cancers in NIS was 5.6%. It was demonstrated that age greater than 50 and comorbidities such as congestive heart failure and renal failure were independently associated with increased in-hospital mortality. Finally, a significant mortality difference was discovered between patients undergoing emergent resection procedures compared with patients receiving elective procedures. After excluding all cholecystectomies, for which definitions of emergent vs. elective admission type may differ from major hepatobiliary operations, the survival disadvantage of patients undergoing emergent procedures persisted.
The role of elective vs. emergent operation for major operations has been discussed in the context of other diseases. Poulose et al. describe a six- to sevenfold adjusted increase in in-hospital mortality associated with non-elective paraesophageal repair in a population-based examination of octogenarians where nearly half of the cohort underwent non-elective repairs.12 Khuri et al. also identified ‘emergency status’ as one of five pre-operative variables predictive of 30-day mortality in their analysis of eight major operations performed in the Veterans Administration.13 McKay et al. discuss ‘urgency of admission’ as a contributing predictor of peri-operative mortality in their evaluation of surgeon training on outcomes after hepatic surgery.14 The literature specific to biliary tract cancer regarding the effect of admission type on mortality is, however, sparse. Single institution data from multiple centres do not examine admission status as a factor in their reviews of biliary tract cancer mortality.15–21 Rather, evaluation of outcome in these studies is focused on patient demographics, type of operation, resection status and tumour grade. Of these characteristics, few have the potential to correlate well with emergent admission; jaundice, for instance, has been described as an adjusted risk factor for poor outcome, but not in the context of emergent admission.20,21
This analysis provides a current national view of surgical resection for biliary tract cancers, and provides the first US population-based evidence that patients incur a significant increase in mortality when undergoing emergent surgery for their biliary tract cancers. Single-institutional experiences with biliary tract cancers provide a wealth of data on a smaller number of patients, being limited at most to a few hundred patients given that biliary tract cancers are a rare disease. In contrast, this review of the NIS data provides a significant cross-section of the US experience with biliary tract cancers, enabling a robust analysis that supports queries into subgroups of the original cohort. Although recently published population-based investigations address intrahepatic cholangiocarcinoma, this study is the first to address in-hospital mortality as an outcome for the larger group of all resected biliary tract cancers.22,23
This study has several limitations. Administrative databases, while broad in their nature, cannot provide relevant clinical detail to the degree of that provided by single-institution databases or clinician chart reviews; in particular, information such as tumour size, staging and vascular invasion is unavailable in the NIS. The NIS reports only in-patient mortality, preventing examination of data pertaining to patients who were discharged, readmitted and died subsequently in the hospital. Furthermore, additional endpoints other than hospital stay (e.g. 30-day and 60-day mortality) are not available in the NIS, whereas they are used routinely in Surveillance Epidemiology End Results (SEER)-based studies.23 Lastly, data that is longitudinal in nature, such as that pertaining to pre- and post-hospital care, is excluded from the NIS, making inquiries into referral and follow-up patterns impossible.
These results have several important implications for future investigation. The finding of increased mortality for emergent resection of biliary tract cancers may reflect the effects of a delay in diagnosis, thus increasing morbidity and mortality for such patients who do undergo resection. Future studies should be directed at elucidating reasons for operating emergently, and at examining safer alternatives for temporizing patients until elective resection. In the case that emergent operation may not be avoided, this data may have important ramifications for pre-operative patient and family discussions. Ultimately, these results warrant further investigation into the various means by which biliary tract cancers could be detected and treated earlier.
In conclusion, this study queried a large nationally-representative database to investigate in-hospital mortality for resection for biliary tract cancer. Several factors were identified that increased the odds of mortality, including congestive heart failure, renal failure and emergent admission. These results suggest that patients should be medically optimized prior to surgery whenever possible. When this is not feasible, the mortality risk this may confer should be discussed with patients.
Acknowledgments
The contributions of Dr Fred Anderson in database provision and statistical analysis are greatly appreciated.