Adjusted multivariate Cox proportional hazards regression model showing extent of lymph node resection.
Downing SR, Cadogan K, Ortega G, Oyetunji TA, Siram SM, Chang DC, Ahuja N, Leffall LD, Frederick WAI. Early-Stage Gallbladder Cancer in the Surveillance, Epidemiology, and End Results DatabaseEffect of Extended Surgical Resection. Arch Surg. 2011;146(6):734-738. doi:10.1001/archsurg.2011.128
Copyright 2011 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2011
Extended surgical resection (ESR) may improve survival in patients with early-stage primary gallbladder cancer.
Retrospective analysis of findings in the Surveillance, Epidemiology, and End Results (SEER) database.
Individuals with potentially surgically curable gallbladder cancer (Tis, T1, or T2) who underwent a surgical procedure.
Main Outcome Measures
Overall survival, number of lymph nodes (LNs) excised, and results of simple cholecystectomy vs ESR.
We identified 3209 patients with early-stage gallbladder cancer (11.7% Tis, 30.1% T1, and 58.2% T2). On multivariate analysis, decreased survival was noted among patients older than 60 years (hazard ratio, 1.57; 95% confidence interval, 1.30-1.90), among patients with more advanced cancer (1.99; 1.46-2.70 for T1; 3.29; 2.45-4.43 for T2), and among patients with disease-positive LNs (1.65; 1.39-1.95 for regional; 2.58; 1.54-4.34 for distant) (P < .001 for all), while increased survival was observed among female patients (0.82; 0.70-0.96; P = .02) and among patients undergoing ESR (0.59; 0.45-0.78; P < .001). The survival advantage of ESR was seen only in patients with T2 lesions (0.49; 0.35-0.68; P < .001). Lymph node excision data were available for a subset of 2507 patients, of whom 68.2% had no LN excised, 28.2% had 1 to 4 LNs excised, and 3.6% had 5 or more LNs excised. On multivariate analysis, patients with 1 to 4 LNs excised had a survival benefit over those with no LN excised (HR, 0.55; 95% CI, 0.46-0.66; P < .001), and patients with 5 or more LNs excised had a survival benefit over patients with 1 to 4 LNs removed (0.63; 0.40-0.96; P = .03). Lymph node excision improved survival in patients with T2 lesions (0.42; 0.33-0.53; P < .001 for patients with 1-4 LNs excised).
Extended surgical resection, LN excision, or both may improve survival in certain patients with incidentally discovered gallbladder cancer.
Gallbladder cancer (GBC) is the most common tumor of the biliary tract and represents more than three-fourths of extrahepatic biliary cancers diagnosed during 2008 in 9760 patients.1 Patients with GBC are often diagnosed as having later-stage tumors, which are more likely to cause symptoms of biliary obstruction and pain from invasion of surrounding structures. Patients with later-stage disease have 5-year survival of less than 5%, but those with earlier-stage tumors do not fare much better.2,3 Poor survival is seen even in patients with the earliest-stage GBC, believed to result in part from aggressive biologic characteristics of tumors.
Recent literature about GBC supports the performance of cholecystectomy plus at least nonanatomical resection of liver segments IV and V in contact with the gallbladder among patients with violation of the visceral peritoneal surface of the gallbladder by tumor cells (T2 disease).4- 7 Considered as having regional disease, patients with T2 tumors benefit from an effort to achieve a microscopically complete resection.6 However, considerable debate exists about optimal treatment for patients whose tumors have invaded the muscularis propria but not the visceral peritoneal surface of the gallbladder. There is no consensus for treatment of these T1b tumors, with some authors advocating simple cholecystectomy and others promoting the wider resection that patients with T2 tumors tend to receive.3,6,8- 10
In addition, no current guidelines exist for the extent of regional lymph node (LN) excision in patients with T1 or T2 tumors, despite recognition that LN metastases represent the most common means of progression in GBC.9,11 While some experts advocate wide resection of all LNs between the head of the pancreas and the porta hepatis and the periaortic nodes to the point of pancreatoduodenectomy, others recommend a more local resection of selected LNs immediately adjacent to the cystic duct and common hepatic duct or a more narrow resection of the cystic duct LN only.3,5- 7,12
We sought to delineate the survival benefit, if any, of extended surgical resection (ESR) of the gallbladder among patients with potentially surgically curable GBC using a population-based data set. We further examined the effects associated with the number of LNs excised, as well as the possible benefit of liver resection in addition to cholecystectomy.
The National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) database (January 1988 to December 2005) was queried for patients diagnosed as having early-stage GBC, defined by the AJCC Staging Manual, 6th edition, as Tis, T1, or T2 cancers. SEER contains information on cancer incidence and survival in 17 states and metropolitan regions comprising approximately 26% of the American population and requires an error rate of less than 5% for all entries (http://www.seer.cancer.gov/about).
Patients were identified by SEER recode 21080 for gallbladder tumors, corresponding to International Classification of Diseases for Oncology, Third Edition, code C23.9. Included were patients having the following histologic tumor types: adenocarcinoma (codes 814.0-814.7), papillary carcinoma or papillary adenocarcinoma (codes 805.0-805.2 and 826.0-826.3), adenosquamous carcinoma (codes 856.0-856.2 and 857.0), or other cancers (849.0 signet ring, 804.0-804.6 small cell, 801.0-801.5 carcinoma not otherwise specified, or 802.0-802.2 undifferentiated carcinoma). Patients with any other histologic type or with more than 1 cancer diagnosis were excluded from analysis.
Also excluded from analysis were patients who did not undergo definitive surgical resection as treatment for their cancer (eg, patients with biopsy specimens or cytologic brushings only). Patients who were recorded as undergoing surgery not otherwise specified were also excluded, requiring the exclusion of patients diagnosed before 1988. Patients who had a surgical intervention were further categorized as undergoing simple cholecystectomy or ESR, with the latter including patients who underwent resection of the liver synchronously. Liver resection was defined as any resection of liver parenchymal tissue, ranging in extent from nonanatomical resection of the gallbladder bed to formal lobectomy. Because SEER codes surgical interventions by the first procedure only, patients who had an early-stage GBC diagnosed at cholecystectomy for another indication and later had a metachronous liver resection would not be identified and were included in the simple cholecystectomy group.
All statistics were calculated using commercially available software (STATA/MP version 10; StataCorp LP, College Station, Texas). Bivariate analyses were performed using t test for continuous variables (eg, age) and χ2 test for categorical variables (eg, race/ethnicity). Cox proportional hazards regression models were used for all survival analyses. All multivariate regressions were adjusted for patient, tumor, and intervention characteristics. Patient characteristics included age, sex, marital status, and race/ethnicity. Tumor characteristics included the following: grade, depth of invasion (T status), number of disease-positive LNs identified, and histologic subtype as already described. Intervention characteristics included extent of surgical resection, number of LNs examined, and year of diagnosis. Patients with missing data in any evaluated category were excluded. Subsequent Cox proportional hazards regression models performed among patients delineated by T classification (Tis, T1a, T1b, or T2) of disease included all previously mentioned characteristics.
We identified 3209 patients with early-stage GBC (11.7% Tis, 30.1% T1, and 58.2% T2). On multivariate analysis, decreased survival was noted among patients older than 60 years (hazard ratio [HR], 1.57; 95% confidence interval [CI], 1.30-1.90), among patients with more advanced cancer (1.99; 1.46-2.70 for T1; 3.29; 2.45-4.43 for T2), and among patients with disease-positive LNs (1.65; 1.39-1.95 for regional; 2.58; 1.54-4.34 for distant) (P < .001 for all), while increased survival was observed among female patients (0.82; 0.70-0.96; P = .02) and among patients undergoing ESR (0.59; 0.45-0.78; P < .001). The survival advantage of ESR was seen only in patients with T2 lesions (0.49; 0.35-0.68; P < .001). Lymph node excision data were available for a subset of 2507 patients, of whom 68.2% had no LN excised, 28.2% had 1 to 4 LNs excised, and 3.6% had 5 or more LNs excised.
Patient, tumor, and intervention characteristics of 2495 patients with early-stage GBC identified in the SEER database are summarized in Table 1. The median age at diagnosis was 72 years, and most patients were female (74.9%) and of white race/ethnicity (60.0%) and were diagnosed as having adenocarcinoma (81.0%). More patients were identified in recent years (32.0% during 2003 to 2005), but because the population base for SEER grew during that period, no increase was noted in the diagnosis of early-stage GBC during later years of the study.
The median survival for all patients was 27 months. By T classification, the median survival was 156 months for patients with Tis, 93 months for patients with T1a, 33 months for patients with T1b, and 17 months for patients with T2. In unadjusted analyses, the rates of simple cholecystectomy vs ESR did not vary by patient sex or race/ethnicity or by tumor grade, histologic subtype, grade, or T classification. However, there was a nonsignificant trend of higher percentages of patients undergoing ESR with increasing depth of invasion (3.9% for Tis vs 6.0% for T1 vs 7.2% for T2, P = .10). The type of intervention received varied significantly by marital status, with single (never married) patients being more likely to undergo ESR than their divorced or widowed counterparts (8.3% single vs 3.4% divorced and 4.5% widowed, P = .01). Patients who underwent ESR were significantly younger than patients who underwent simple cholecystectomy (64.6 vs 70.8 years, P < .001). Patients with disease-positive regional LNs were more likely to undergo ESR than patients with disease-negative or disease-positive distant LNs (12.9% vs 6.2% vs 8.0%, P < .001).
Table 2 gives the results of a Cox proportional hazards regression model among the entire cohort. Patients who received ESR had a significant survival benefit over patients who underwent simple cholecystectomy (HR, 0.75; 95% CI, 0.57-1.00; P = .05). On multivariate analysis, additional survival benefit was obtained by LN excision, shown in the Figure as adjusted Kaplan-Meier survival curves (HR, 0.55; 95% CI, 0.46-0.66 for patients with 1-4 LNs excised; 0.31; 0.20-0.49 for patients with ≥5 LNs excised) (P < .001 for both). Further investigation revealed a survival benefit for patients with 5 or more LNs excised over patients who had 1 to 4 LNs excised (0.63; 0.40-0.96; P = .03). No survival difference was seen by patient race/ethnicity or by year of intervention.
We next examined the effect of surgery and LN excision, stratifying patients by extent of disease and using a similar Cox proportional hazards regression model. Patients with Tis or T1a tumors showed no survival difference by type of surgical intervention or by extent of LN excision. However, patients with T1a tumors demonstrated a trend toward survival benefit among those who underwent excision of 5 or more LNs over those who underwent excision of 1 to 4 LNs (HR, 0.80; 95% CI, 0.55-1.16; P = .25). Table 3 summarizes results of a Cox proportional hazards regression model among patients diagnosed as having T1b GBC. While a trend toward worse survival was seen for patients who underwent ESR (HR, 1.51; 95% CI, 0.78-2.90; P = .22), a trend toward better survival was seen for patients who underwent LN excision (0.82; 0.56-1.18; P = .29 for 1-4 LNs excised; 0.42; 0.10-1.85; P = .25 for ≥5 LNs excised). In contrast, among patients diagnosed as having T2 GBC (Table 4), better survival was seen in those who underwent ESR (0.64; 0.46-0.90; P = .01). Furthermore, LN excision significantly improved survival in patients with T2 lesions (0.42; 0.33-0.53 for 1-4 LNs excised; 0.26; 0.16-0.42 for ≥5 LNs excised) (P < .001 for both).
We find that patients with T2 GBC benefit greatly from ESR of the gallbladder in addition to excision of 5 or more LNs. This supports a previous suggestion by Coburn et al,8 who showed a trend toward survival benefit for ESR and for LN excision in patients with T2 GBC. However, in contrast to Coburn et al, we find no survival benefit for ESR in patients with T1a or T1b GBC, a result supported by previous investigations.6,8,10
Furthermore, patients having T1b GBC in our study showed no adjusted survival difference by extent of surgical resection or by number of LNs excised. However, the fewer patients in that subgroup (n = 462) (Table 3), especially compared with patients having T2 GBC (n = 1533) (Table 4), may limit the statistical power of this portion of the study analysis. Therefore, we believe that the demonstrated trend toward survival benefit for LN excision and trend toward survival detriment for ESR among these patients with T1b GBC represent true between-group differences. Further investigation of these findings is warranted because, while ESR of the liver may not affect survival in a patient with T1b GBC, more extensive excision of regional LNs than is practiced may improve survival. Because preoperative diagnosis of T1 lesions is uncommon at best, with one series showing 28% of these early cancers diagnosed by ultrasonography before surgical resection,13 patients incidentally found to have T1b GBC may benefit from surgical reexploration for appropriate LN evaluation.
Surprisingly, despite the increased performance of cholecystectomy for gallstone disease with the adoption of minimally invasive techniques, we found no heightened incidence of early-stage GBC over the past 2 decades. Coupled with the findings by Kiran et al,2 who showed that the increase in cholecystectomies during the 1980s coincided with a decrease in diagnosis of all stages of GBC, this implies that aggressive treatment of benign gallbladder disease has decreased the diagnosis of late-stage cancers but has not increased our discovery of early-stage cancers. The continued rarity of early-stage GBC may have contributed to the haphazard use of ESR. Notably, we found no difference in the receipt of ESR by T classification of tumor, indicating that American surgeons are no more likely to perform ESR on a patient with a T2 tumor than on a patient with in situ disease. Although no good chemotherapy or irradiation interventions had been developed during the course of this study, the rare and slightly better-than-random application of ESR may have been contributory to the lack of survival improvement among these patients with early-stage GBC.
As with any retrospective study, including database studies with prospectively collected data (as with SEER), there are several limitations to our findings. As noted in the “Methods” section, SEER lists only initial surgical treatment; therefore, any patient who underwent subsequent ESR would be included in our simple cholecystectomy group. This bias would make it more difficult to commit a type I error (ie, show a difference between surgical treatment groups when there is none) because some of the better-performing subsequent patients undergoing ESR would improve the mean outcome of the poorer-performing patients undergoing simple cholecystectomy. Therefore, this supports our assertion that a type II error (or a trend that does not reach significance) is in play for T1b tumors. Furthermore, this study is limited by the exclusion of data points not included in the SEER data set, such as chemotherapy use or completeness of initial resection. Although the former is not likely to alter our data because there are few successful chemotherapeutic agents available for GBC, the latter may influence our results because obtaining a complete R0 resection is paramount in cancer treatment.
In conclusion, we demonstrate a significant survival benefit for extended cholecystectomy in patients with early-stage GBC. Furthermore, such patients benefit from LN excision, especially those who have 5 or more LNs excised. The survival benefit for both of these interventions was seen primarily in patients with T2 disease. However, patients having less invasive tumors, especially those having T1b tumors, showed a trend toward increased survival with excision of 5 or more LNs and a trend toward decreased survival with ESR of the gallbladder. These findings suggest that T1b GBC is not a localized disease and may benefit from more extensive LN excision and reinforce the current recommendations for treatment of T2 cancers.
Correspondence: Wayne A. I. Frederick, MD, Department of Surgery, Howard University College of Medicine, 2041 Georgia Ave NW, Ste 4000, Washington, DC 20060 (firstname.lastname@example.org).
Accepted for Publication: May 22, 2010.
Author Contributions:Study concept and design: Downing, Oyetunji, Siram, Chang, and Frederick. Acquisition of data: Downing, Cadogan, Ortega, Oyetunji, and Chang. Analysis and interpretation of data: Downing, Cadogan, Ortega, Oyetunji, Chang, Ahuja, Leffall, and Frederick. Drafting of the manuscript: Downing, Cadogan, Ortega, and Oyetunji. Critical revision of the manuscript for important intellectual content: Downing, Siram, Chang, Ahuja, Leffall, and Frederick. Statistical analysis: Downing and Chang. Administrative, technical, and material support: Chang and Frederick. Study supervision: Siram, Chang, Ahuja, and Frederick.
Financial Disclosure: None reported.
Previous Presentation: This study was presented as a poster at the 90th Annual Meeting of the New England Surgical Society; September 12, 2009; Newport, Rhode Island.