Abstract
Background and Objectives:
Tumor location (peritoneal vs hepatic) has been incorporated in the 8th edition of the American Joint Committee on Cancer Staging system for gallbladder cancer. However, larger studies are needed to confirm the prognostic impact of tumor location.
Methods:
Patients with pathologically-confirmed gallbladder cancer with information on primary tumor location were included from the National Cancer Database (2009–2012). We compared patients with hepatic-side tumors to those on the peritoneal side. Survival data were plotted using the Kaplan-Meier method. Prognostic factors were modeled with a multivariate Cox Proportional Hazards Model. Primary outcome was overall survival (OS).
Results:
A total of 1251 patients were included. In comparison to patients with peritoneal-sided tumors, patients with hepatic-sided tumors were more likely to: be of higher pT stage (pT3: 49% vs 24%; P < .001); node positive (31% vs 24%; P = .016); undergo liver resection (53% vs 25%; P < .001); or have positive margins (29% vs 16%; P < .001). However, on multivariate analysis, there was no difference in OS between the groups (HR, 0.97; 95% CI, 0.79–1.18; P = .753). Liver resection was associated with improved survival regardless of tumor location in pT2 tumors (peritoneal: HR, 0.57; P = .034; hepatic: HR, 0.67; P < .001).
Conclusions:
This study failed to demonstrate the independent prognostic value of primary tumor location in patients with gallbladder cancer.
Keywords: gallbladder cancer, liver resection, primary tumor location, prognostic value, staging
1 |. INTRODUCTION
Gallbladder cancer is the most common malignancy of the biliary tract and represents 40% of all biliary cancers.1 Even though it is rare (incidence 1.13 cases per 100 000 worldwide), it carries a dismal prognosis (5-year overall survival 13%).2 The majority of patients have either regional (43%) or distant (42%) metastases on diagnosis.2 For those with localized disease, surgical resection is potentially curative.
Tumor depth (T stage), lymph node involvement (N-stage), and distant metastases (M-stage) are well-established prognostic markers of gallbladder cancer.3 Shindoh et al4 first postulated that primary tumor location may impact the pattern of dissemination. This was based on a retrospective study of the combined experience among four institutions that demonstrated higher rates of vascular invasion, neural invasion, nodal metastases, and worse survival in patients with hepatic-sided primary tumor compared to those with peritoneal-sided primary tumors. Furthermore, hepatic-sided T2 tumors had a higher rate of liver metastases in resected specimens compared to those on the peritoneal side. Given the rarity of this disease, these data informed the reformulation of the American Joint Committee on Cancer (AJCC) staging. A validation study by Wang et al5 of the AJCC 8th edition staging system used an institutional data set, which supported substratification but did not test the independent prognostic utility of tumor location in T2 gallbladder cancers. Another recent study of 81 patients with T2 gallbladder cancers by Cho et al6 demonstrated that while overall survival significantly differed between the hepatic- and peritoneal-side tumors, this prognostic utility was lost on multivariate analysis.
The assertion that T2 tumors on the peritoneal side are biologically different from those on the hepatic side was based on the anatomic studies by Karlmark in 1932 which showed differences in venous drainage between the peritoneal and hepatic side of the gallbladder.7 However, Fahim et al7 demonstrated that only 14% of gallbladder cancers disseminate through vascular invasion whereas the vast majority (35%-75%) disseminate through lymphatic invasion. Given these observations, the independent prognostic impact of primary tumor location needs to be examined in a large number of patients.
The goal of the present study was to determine the prognostic impact of primary tumor location using a large national clinical oncology database. Based on prior literature, we hypothesized that patients with T2 gallbladder cancer located on the hepatic side will have a worse prognosis than those on the peritoneal side. Further, primary tumor location will have no impact on the outcome of T1 and T3 gallbladder cancers.
2 |. METHODS
2.1 |. National cancer database patient data
The National Cancer Database (NCDB), jointly sponsored by the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society, is a nationwide oncology outcomes database based on more than 1500 CoC programs, covering approximately 70% of new cancer cases in the United States.9,10 The CoC designates cancer programs based on ability to provide a wide range of oncological services and specialists. CoC-approved hospitals are larger, perform more operations, and provide more cancer-related services to patients than non-CoC hospitals.10
The NCDB shared files contain site-specific deidentified data on more than 80 variables comprising sociodemographic, tumor, treatment, and follow-up information. These data are abstracted by certified tumor registrars from medical records, even if the care extends to a non-CoC facility. The NCDB does not specify the frequency of follow-up, but sets the standard of 90% at 5 years. Quality is assured by the CoC by means of more than 600 electronic automated checks to maximize internal consistency and minimize missing data. In addition, the CoC also performs routine audits to ensure data quality and completeness.11 Institutional review board approval was not required for this study as patient deidentified data were analyzed.
2.1.1 |. Patient eligibility and exclusion criteria
Patients diagnosed with gallbladder cancer International Classifications of Disease for Oncology, 3rd edition (ICD-O-3) topographical code 23.9 and morphological codes consistent with adenocarcinoma histology between 2009 and 2012 were identified (Figure 1). Only patients who had surgery of the primary site were included. We excluded patients who had no documentation of primary tumor location. Further, patients were excluded if they had in situ disease, T4 disease, overlapping tumor location, or if they received neoadjuvant therapy. The resulting cohort was used for a secondary analysis to determine propensity of metastases and nodal dissemination by tumor location. For the primary analysis focused on determining the prognostic value of tumor location, patients with pathological or clinical distant metastases on diagnosis were excluded.
FIGURE 1.
CONSORT diagram
2.1.2 |. Patients' data variables and definitions
Variables included age, ethnicity, sex, Charlson-Deyo score12 for comorbid conditions, insurance type, hospital type, tumor extent, tumor size, node status, and receipt of adjuvant therapies. Primary tumor location was extracted from AJCC Collaborative Staging System (CSS) fields and was defined as peritoneal side or hepatic side. Liver resection was extracted from the CSS field and was defined as partial hepatectomy (one or more liver segments), wedge resections, formal lobectomy, or extended hepatectomy. Surgical resection was defined based on primary site codes, timing of first surgical procedure, and timing of most definitive surgical procedure with respect to the date of diagnosis into three categories: (1) primary cholecystectomy (surgery resection codes 20–50 for primary site 23.9); (2) upfront radical cholecystectomy (surgery resection code 60 for primary site 23.9); and (3) delayed radical resection, which was defined as patients who had undergone liver resection (CSS site-specific factor 10–50) separate from primary cholecystectomy (surgery resection codes 20–50 for primary site 23.9). We evaluated staging based on the AJCC staging manual, 8th edition.3 Missing data were reported as separate categories, as they may differ systematically between the two groups.
2.1.3 |. Outcomes
The primary outcome was overall survival. Secondary outcomes included rate of high-grade tumors as well as nodal, liver and, distant metastases on diagnosis.
2.2 |. Statistical analysis
Continuous variables are presented as median with interquartile range (IQR) and categorical variables as frequencies with percentages. Patient demographics, cancer-specific and hospital-level characteristics were analyzed using the Kruskal-Wallis test for nonparametric continuous data and the Pearson χ2 test or Fischer exact test for categorical data, as appropriate.
Overall survival (OS) was calculated from the date of diagnosis until the date of death. Survivors were censored at the date of last contact, whereas those who died were censored at the date of death. Kaplan-Meier curves were used to depict survival differences between the two groups, and the log-rank test was used to test these differences for statistical significance.
Univariate and multivariate analysis was performed using the Cox proportional hazards model. The proportional hazards assumption was confirmed by review of Schoenfeld residuals as well as graphically using methods recommended by Hosmer and Lemeshow.13 Hazard ratios (HRs) were reported with 95% confidence intervals (CI). Step-wise Cox regression was performed to identify prognostic factors from the data in the present study. In addition, well-established prognostic factors from the literature were incorporated in the final model. We tested several models and compared them on the basis of Akaike and Bayesian Information Criteria.14 For all statistical analyses, we used STATA/MP software (version 14.1; StataCorp) with assumption of 2-sided tests and a criterion for statistical significance set at α < .05.
3 |. RESULTS
3.1 |. Patient selection
Overall, 7406 patients had primary site surgery. Of these, 2079 (28%) had primary tumor location available. We evaluated if the cohort of patients who had primary tumor location available (n = 2079) were comparable to those who did not (n = 5327) across established prognostic covariates. The patient cohort with the primary tumor location available had a slightly higher T stage, node positivity, and rate of metastases (Table S1). Consequently, OS was slightly lower in patients who had the primary tumor location available (location available vs not available; median OS 17 vs 19 months; HR 1.09 [95% CI, 1.02, 1.17], P = .007). The results demonstrate small but significant differences between the two cohorts.
3.2 |. Prognostic impact of tumor location in metastatic disease
To determine the impact of tumor location on prognosis in metastatic disease, we compared overall survival of patients with metastases stratified by primary tumor location There was no difference in overall survival between these two groups (log-rank P = .321).
3.3 |. Prognostic impact of tumor location in non-metastatic disease
To explore the prognostic impact of tumor location in non-metastatic disease, we restricted our analysis to our primary cohort (pT1–3, any N, M0) as shown in Figure 1. Because pTis tumors (n = 39, 2%) are not considered to have metastatic potential and there were few pT4 cancers (n = 65, 3%), we excluded these from primary analysis. In addition, we excluded tumors in which the T stage was unknown (n = 135, 7%). A total of 1251 patients comprised the primary analysis cohort, of whom 907 (72.5%) had tumors on the hepatic side and 344 (27.5%) had tumors on the peritoneal side. Analysis of patient, tumor, and procedure information for the two groups is shown in Table 1. These data demonstrated that patients with tumors on the hepatic side were younger and more likely to be treated at an academic facility. Patients with hepatic-sided tumors also had larger tumors, higher pT stage, higher pN-stage, and were more likely to undergo a radical resection. Similarly, hepatic-sided tumors were more likely to have positive margins and receive adjuvant therapy. However, there was no difference in sex, race, insurance status, or time between first and definitive operation in the two analysis groups.
TABLE 1.
Sociodemographic and clinicopathologic characteristics stratified by adjuvant regimen
| Characteristic | N = 1251 | Primary location |
P value | ||
|---|---|---|---|---|---|
| All | Peritoneal (n = 344) | Hepatic (n = 907) | |||
| Age, y, n (%) | <60 | 266 (21) | 72 (21) | 194 (21) | <.001 |
| 60–79 | 676 (54) | 160 (46) | 516 (57) | ||
| 80+ | 309 (35) | 112 (33) | 197 (22) | ||
| Sex, n (%) | Male | 426 (34) | 105 (30) | 321 (35) | .15 |
| Female | 825 (66) | 239 (69) | 586 (65) | ||
| Charlson-Deyo score, n (%) | 0 | 802 (64) | 214 (62) | 588 (65) | .669 |
| 1 | 335 (27) | 98 (28) | 237 (26) | ||
| 2+ | 114 (9) | 32 (9) | 82 (9) | ||
| Race, n (%) | Non-Hispanic White | 858 (69) | 236 (69) | 622 (69) | .805 |
| Black | 170 (14) | 45 (13) | 125 (14) | ||
| Hispanic White | 131 (10) | 40 (12) | 91 (10) | ||
| NA/PI/Asian/other | 92 (7) | 23 (7) | 69 (8) | ||
| Insurance, n (%) | Private insurance | 349 (28) | 1,557 (38) | 1,979 (48) | .827 |
| Government | 836 (67) | 2,409 (59) | 2,071 (50) | ||
| Not insured | 46 (4) | 97 (2) | 105 (2) | ||
| Unknown | 20 (2) | 4(1) | 16 (2) | ||
| Facility type, n (%) | Non-academic A:F | 752 (60) | 227 (66) | 525 (58) | .022 |
| Academic | 484 (39) | 115 (33) | 369 (41) | ||
| Unknown | 15 (1) | 2 (<1) | 13 (1) | ||
| Tumor size, mm | Median (IQR) | 30 (20–45) | 25 (15–40) | 30 (20–50) | <.001 |
| AJCC pT stage, n (%) | T1 | 157 (12) | 54 (16) | 103 (11) | <.001 |
| T2 | 561 (45) | 206 (60) | 355 (39) | ||
| T3 | 533 (43) | 84 (24) | 449 (49) | ||
| AJCC pN-stage, n (%) | N0 | 588 (47) | 160 (46) | 428 (47) | .016 |
| N+ | 362 (29) | 84 (24) | 278 (31) | ||
| Nx | 301 (24) | 100 (29) | 201 (22) | ||
| Lymph nodes examined | Median (IQR) | 1 (0–3) | 1 (0–2) | 1 (0–4) | <.001 |
| Lymph nodes positive | Median (IQR) | 0 (0–1) | 0 (0–1) | 0 (0–1) | .363 |
| Grade, n (%) | Low | 184 (15) | 64 (19) | 120 (13) | .068 |
| Intermediate | 527 (42) | 146 (42) | 381 (42) | ||
| High | 468 (37) | 118 (34) | 350 (39) | ||
| Unknown | 72 (6) | 16 (5) | 56 (6) | ||
| Type of surgery, n (%) | Simple cholecystectomy | 635 (51) | 239 (69) | 396 (44) | <.001 |
| Radical cholecystectomy | 170 (14) | 23 (7) | 147 (16) | ||
| Delayed radical resection | 418 (33) | 71 (21) | 347 (38) | ||
| Not specified | 28 (2) | 11 (3) | 17 (2) | ||
| Liver resection, n (%) | No | 673 (54) | 257 (75) | 416 (46) | <.001 |
| Yes | 567 (45) | 85 (25) | 482 (53) | ||
| Not specified | 11 (1) | 2 (<1) | 9(1) | ||
| Time between first and definitive operation, d | Median (IQR) | 40 (28–56) | 43 (34–55) | 40 (25–57) | .386 |
| Final margin, n (%) | R0 | 870 (69) | 272 (79) | 589 (66) | <.001 |
| R+ | 319 (25) | 54 (16) | 265 (29) | ||
| Unknown | 62 (5) | 18 (5) | 44 (5) | ||
| Adjuvant therapy, n (%) | None | 763 (61) | 238 (69) | 525 (58) | <.001 |
| Radiation | 35 (3) | 12 (3) | 23 (3) | ||
| Chemotherapy | 198 (16) | 41 (12) | 157 (17) | ||
| Chemoradiation | 255 (20) | 53 (15) | 202 (22) | ||
Note: X2 test for categorical variables. Kruskal-Wallis rank test for continuous variables.
Abbreviations: AJCC, American Joint Committee on Cancer; IQR, interquartile range.
Next, we assessed the prognostic impact of tumor location in the primary analysis cohort (pT1–3, any N, M0). Median follow-up for alive patients was 28 months (IQR, 20–39 months). This was not significantly different (P = .388) between the hepatic-sided tumors (median, 28; IQR, 20–39 months) and peritoneal-sided tumors (median, 27; IQR, 20–38 months). At the time of analysis, there were 516 (57%) deaths in the hepatic group as compared to 169 (49%) deaths in the peritoneal group (log-rank P < .06). Median overall survival for the hepatic group was 24 months (IQR, 10–58) and that for the peritoneal group was 29 months (IQR, 10—not reached). Kaplan-Meier overall survival estimates for the primary cohort are shown in Figures 2 and S1.
FIGURE 2.
Kaplan-Meier overall survival estimates of patients with (T1–3, any N, M0) gallbladder cancer stratified by primary tumor location. Also see Figure S3
Univariate and multivariate survival analyses for the primary cohort (pT1–3, any N, M0) patients are presented in Table 2. These analyses demonstrated that tumor location is not an independent prognostic factor for early gallbladder cancer.
TABLE 2.
Univariate and multivariate analysis of prognostic factors in T1–3, any N, M0 gallbladder cancer
| Characteristic | N = 1251 | Univariate analysis |
Multivariate analysis |
||
|---|---|---|---|---|---|
| HR (95% CI) | P value | HR (95% CI) | P value | ||
| Age (y), n (%) | <60 | 1 <referent> | 1 <referent> | ||
| 60–79 | 1.61 (1.30–2.00) | <.001 | 1.28 (1.02–1.60) | .036 | |
| 80+ | 2.32 (1.83, 2.93) | <.001 | 1.68 (1.29–2.19) | <.001 | |
| Sex, n (%) | Male | 1 <referent> | |||
| Female | 0.91 (0.77–1.06) | .219 | |||
| Charlson-Deyo score, n (%) | 0 | 1 <referent> | 1 <referent> | ||
| 1 | 1.33 (1.12–1.57) | .001 | 1.17 (0.98–1.40) | .084 | |
| 2+ | 1.40 (1.09–1.80) | .009 | 1.60 (1.22–2.10) | .001 | |
| Race, n (%) | Non-Hispanic White | 1 <referent> | |||
| Black | 0.83 (0.66–1.05) | .123 | |||
| Hispanic White | 0.61 (0.46–0.82) | .001 | |||
| NA/PI/Asian/Other | 0.75 (0.55–1.02) | .069 | |||
| Insurance, n (%) | Private insurance | 1 <referent> | |||
| Government | 1.38 (1.16–1.65) | <.001 | |||
| Not insured | 1.04 (0.64–1.67) | .867 | |||
| Facility type, n (%) | Non-academic A:F | 1 <referent> | 1 <referent> | ||
| Academic | 0.80 (0.68–0.93) | .005 | 0.86 (0.72–1.02) | .082 | |
| Tumor size, cm | 1.02 (1.01–1.03) | .003 | |||
| AJCC pT stage, n (%) | T1 | 1 <referent> | 1 <referent> | ||
| T2 | 1.82 (1.33–2.49) | <.001 | 1.49 (1.07–2.08) | .018 | |
| T3 | 4.05 (2.98–5.51) | <.001 | 2.90 (2.04–4.10) | <.001 | |
| AJCC pN-stage, n (%) | N0 | 1 <referent> | 1 <referent> | ||
| N+ | 2.07 (1.73–2.47) | <.001 | 1.68 (1.39–2.04) | <.001 | |
| Nx | 1.93 (1.60–2.33) | <.001 | 1.34 (1.09–1.65) | .006 | |
| Lymph nodes examined | 0.97 (0.95–0.99) | .004 | |||
| Lymph nodes positive | 1.25 (1.18–1.31) | <.001 | |||
| Grade, n (%) | Low | 1 <referent> | 1 <referent> | ||
| Intermediate | 1.59 (1.22–2.08) | .001 | 1.31 (0.98–1.74) | .064 | |
| High | 2.62 (2.01–3.42) | <.001 | 1.97 (1.48–2.62) | <.001 | |
| Unknown | 2.11 (1.43--3.10) | <.001 | 1.80 (1.18–2.73) | .006 | |
| Location, n (%) | Peritoneal | 1 <referent> | 1 <referent> | ||
| Hepatic | 1.18 (0.99–1.41) | .059 | 0.97 (0.79–1.18) | .753 | |
| Type of surgery, n (%) | Simple Cholecystectomy | 1 <referent> | 1 <referent> | ||
| Radical Cholecystectomy | 0.69 (0.54–0.87) | .002 | |||
| Delayed Radical Resection | 0.65 (0.55–0.77) | <.001 | |||
| Liver resection, n (%) | No | 1 <referent> | 1 <referent> | ||
| Yes | 0.66 (0.56–0.76) | <.001 | 0.78 (0.65–0.94) | .009 | |
| Time between first and definitive operation, days | 0.98 (0.97–1.00) | .044 | |||
| Final margin, n (%) | R0 | 1 <referent> | 1 <referent> | ||
| R+ | 2.94 (2.50–3.45) | <.001 | 1.92 (1.59–2.32) | <.001 | |
| Adjuvant therapy, n (%) | None | 1 <referent> | 1 <referent> | ||
| Radiation | 1.05 (0.68–1.62) | .831 | 0.82 (0.52–1.29) | .385 | |
| Chemotherapy | 1.03 (0.84–1.26) | .773 | 0.86 (0.69–1.07) | .185 | |
| Chemoradiation | 0.70 (0.57–0.86) | .001 | 0.61 (0.48–0.76) | <.001 | |
Abbreviations: AJCC, American Joint Committee on Cancer; HR, hazard ratio; 95% CI, 95% confidence interval.
P value calculated using cox regression model.
3.4 |. Prognostic impact of tumor location in pT2 tumors
Prior studies4,5 have suggested that the prognostic impact of tumor location is only evident for T2 gallbladder cancers. Therefore, we performed additional survival analyses on this subgroup of patients (pT2, any N, M0). Multivariable analyses demonstrate that tumor location was not independently associated with prognosis, regardless of nodal status (Table 3).
TABLE 3.
Multivariate analysis of prognostic factors in T2, M0 gallbladder cancer stratified by nodal status
| Characteristic | pT2, Any N, M0 N = 561 |
pT2, N0, M0 N = 286 |
|||
|---|---|---|---|---|---|
| HR (95% CI) | P value | HR (95% CI) | P value | ||
| Age (y), n (%) | <60 | 1 <referent> | 1 <referent> | ||
| 60–79 | 1.10 (0.75–1.61) | .62 | 0.85 (0.45–1.58) | .6 | |
| 80+ | 1.13 (0.73–1.75) | .572 | 1.09 (0.55–2.15) | .801 | |
| Charlson-Deyo score, n (%) | 0 | 1 <referent> | 1 <referent> | ||
| 1 | 1.04 (0.76–1.41) | .822 | 1.08 (0.65–1.78) | .766 | |
| 2+ | 1.44 (0.99–2.10) | .055 | 1.65 (0.88–3.10) | .121 | |
| Facility type, n (%) | Non-academic A:F | 1 <referent> | 1 <referent> | ||
| Academic | 1.03 (0.77–1.36) | .841 | 0.94 (0.60–1.49) | .805 | |
| AJCC pN-stage, n (%) | N0 | 1 <referent> | |||
| N+ | 2.14 (1.56–2.95) | <.001 | |||
| Nx | 1.28 (0.91–1.80) | .154 | |||
| Grade, n (%) | Low | 1 <referent> | 1 <referent> | ||
| Intermediate | 1.05 (0.69–1.60) | .812 | 1.01 (0.53–1.93) | .962 | |
| High | 1.82 (1.20–2.77) | .005 | 2.40 (1.29–4.46) | .006 | |
| Unknown | 1.64 (0.83–3.22) | .153 | 1.40 (0.49–3.94) | .537 | |
| Location, n (%) | Peritoneal | 1 <referent> | 1 <referent> | ||
| Hepatic | 1.00 (0.76–1.33) | .968 | 1.15 (0.73–1.80) | .554 | |
| Liver resection, n (%) | No | 1 <referent> | 1 <referent> | ||
| Yes | 0.66 (0.48–0.92) | .013 | 0.50 (0.30–0.82) | .007 | |
| Final margin, n (%) | R0 | 1 <referent> | 1 <referent> | ||
| R+ | 2.96 (2.16–4.07) | <.001 | 2.85 (1.59–5.09) | <.001 | |
| Adjuvant therapy, n (%) | None | 1 <referent> | 1 <referent> | ||
| Radiation | 0.45 (0.18–1.14) | .092 | 0.82 (0.52–1.29) | .385 | |
| Chemotherapy | 0.61 (0.39–0.95) | .028 | 0.86 (0.69–1.07) | .185 | |
| Chemoradiation | 0.57 (0.40–0.83) | .003 | 0.61 (0.48–0.76) | <.001 | |
Abbreviations: AJCC, American Joint Committee on Cancer; HR, hazard ratio; 95% CI, 95% confidence interval.
P value calculated using cox regression model.
3.5 |. Nodal and distant metastatic potential of pT1–3 tumors
The analyses thus far suggested that tumor location was not independently prognostic. To determine if there are biological differences between the peritoneal- and hepatic-side tumors in their ability to metastasize, we compared the rate of high-grade histology and nodal, liver, and all distant metastases on diagnosis in our secondary analysis cohort. These are shown in Figure 3. While there was an increase in the rates of high-grade tumors and nodal, liver, and all distant metastases with increasing tumor depth (pT stage), there was no association with primary tumor location within any of the pT stage strata. Since the extent of nodal examination is variable between the two categories and that may influence the rate of nodal metastases, we performed additional sensitivity analyses stratified by the extent of lymphadenectomy. The results confirmed the lack of association between primary tumor location and rate of nodal metastases, regardless of the definition used for adequate lymphadenectomy.
FIGURE 3.
Invasive potential of pT1-T3 gallbladder cancers. Rates of high-grade tumors, nodal metastases, all distant metastases and liver metastases at diagnosis are compared between peritoneal and hepatic sided tumor stratified by pT stage
3.6 |. Impact of tumor location on the value of liver resection
Findings shown in Table 3 demonstrate that negative margins, liver resection, and adjuvant chemotherapy with or without radiation are independently associated with improved overall survival in pT2 gallbladder cancers. We therefore analyzed the impact of tumor location on the value of radical resection in a subset analysis (Figure 4). These results demonstrate that liver resection was associated with improved overall survival for peritoneal-sided tumors (resection vs no resection; HR, 0.57; 95% CI, 0.34–0.96; P = .034) as well as hepatic-sided tumors (resection vs no resection; HR 0.67, 95% CI, 0.55–0.82; P <.001).
FIGURE 4.
Kaplan-Meier overall survival estimates of patients with (T2, any N, M0) gallbladder cancer stratified by primary tumor location and whether or not they had liver resection (also see Figure S1)
4 |. DISCUSSION
This is a national cohort analysis that uses data from multiple Commission on Cancer-accredited institutions to evaluate the prognostic impact of primary tumor location in gallbladder cancer. As such, this comprises the largest analysis to date on the subject and provides some valuable insights. In this study, we found that in unadjusted analyses, patients with non-metastatic gallbladder cancers originating on the peritoneal side have a slightly better survival compared to those that originate on the hepatic side, although that was not statistically significant (P = .061). Furthermore, through a series of multivariate models, we determined that this difference cannot be independently attributed to the primary tumor location. Since the most profound impact of tumor location has been suggested to be among pT2 tumors, we specifically evaluated this subgroup and again found no evidence of prognostic significance of primary tumor location. Additionally, our analysis demonstrates that peritoneal- and hepatic-side tumors have similar propensities to be of high-grade and to have nodal and distant metastases. This suggests that these tumors are biologically similar. Lastly, we demonstrate that liver resection is valuable for all pT2 tumors regardless of the primary tumor location and should continue to be standard of care.
Gallbladder cancer originates in the mucosa of the gallbladder and progressively invades: lamina propria (T1a); muscular layer (T1b); perimuscular subserosal connective tissue (T2); serosa and/or one adjacent oragan, for example, liver, bile ducts, stomach, duodenum, pancreas or colon (T3); and portal vessels or more than one adjacent organ (T4). Other than direct extension to other organs (58%-89%), gallbladder cancer can disseminate through lymphatic invasion (up to 75%), vascular invasion (14%), peritoneal spread (10%), and neural spread (22%-35%).15,16 Most gallbladder cancers demonstrate multiple modes of dissemination.
The idea that primary tumor location is a predictor of tumor progression and survival was discussed by Shindoh et al.4 They speculated that anatomic differences in vascular drainage may explain the greater propensity to metastasize in hepatic-side tumors compared to peritoneal-side tumors. While there are anatomic differences in the number of cystic veins that drain the peritoneal and hepatic side of the gallbladder, they all drain into the liver and communicate with the hepatic veins. Further, Fahim et al7 demonstrated that only 14% of gallbladder cancers disseminate through vascular invasion whereas the vast majority (35%-75%) disseminate through lymphatic invasion. Recently, Wakai et al17 employed immunohistochemistry to demonstrate that lymphatic invasion and not vascular invasion is the predominant histologic feature of the hepatic spread of gallbladder cancer. Anatomically, the lymphatic plexus within the gallbladder wall is a continuous network without any notable differences between peritoneal or hepatic side as described by Clermont8 and more recently by Nagahashi et al.18 Therefore, the anatomic pattern of lymphovascular distribution in the gallbladder wall is unlikely to lead to a differential propensity to metastasize between hepatic- and peritoneal-sided tumors.
While the multi-institutional analysis of 252 patients by Shindoh et al. found that tumor location stratified the biologic behavior of pT2 tumors,4 the results of the present study looking at 561 patients stand in contrast. A closer look at the analysis in the study by Shindoh et al indicates that metastasis was eliminated in the backwards step-wise elimination approach and were not incorporated as a predictor of survival in the final multivariate model. Macroscopic liver metastases were present in 8.7% of the hepatic-side tumors and 2.5% of the peritoneal-side T2 tumors in their cohort. Because metastatic disease is the strongest predictor of survival in gallbladder cancer,3 failure to incorporate this in the final model could bias these results. Therefore, it is possible that the survival difference between hepatic- and peritoneal-side pT2 gallbladder cancers could be attributed to the differing rate of metastases in the two groups. Recently, Wang et al attempted to validate the eighth edition of the AJCC staging system.5 They found that categorization of the T2 tumors based on location improved stratification. However, they did not test the independent prognostic significance of this stratification in their multivariate model. Lee et al compared the 7th and 8th edition of AJCC staging using patients from the NCDB and found similar c-statistics and little prognostic improvement over the existing system.19 Of note, tumor location was not included in the analysis and therefore this was not directly addressed in this validation study. A similar study by Oweira et al using the Surveillance, Epidemiology, and End Results (SEER) database reached similar conclusions.20 Finally, a more recent study of 81 patient with T2 gallbladder cancers by Cho et al demonstrated that while overall survival significantly differed between the hepatic- and peritoneal-side tumors, this prognostic utility was lost on multivariate analysis.6
The present study addresses this question using a larger clinical oncology database. We found granular information regarding tumor location and resection in a significant number of patients, allowing us to perform a robust analysis in this cohort. We found small differences between the cohort for which detailed pathologic information was available on tumor location compared to the cohort that did not have that information. Because the multivariable models presented in this study adjusted for these differences, the results are generalizable to the larger US population. Another strength of the study is that the large sample size allowed us to build a robust multivariable model without risking over-fitting, which can lead to erroneous associations.21 This was especially important when performing subset analyses.
The main limitation of the study is its retrospective observational design, and therefore the results should be interpreted as correlative rather than causative. The information regarding tumor location was based on the pathology report and was used as abstracted by the NCDB. Therefore, it was not possible to set standards in the reporting of primary tumor location. We excluded overlapping tumors from our analysis for this reason. It is possible that our selection criteria allowed some patients to be over-represented. To address this potential bias, we compared the prognostic variables of included and excluded patients and found minor differences. Further, the multivariate model further reduces this bias by adjusting for relevant covariates. The NCDB does not record information on location and timing of recurrence. While this information is valuable, it is unlikely to change our overall conclusions because overall survival remains the most meaningful endpoint for patients with gallbladder cancer.
In summary, the results of this large national cohort analysis fail to demonstrate the independent prognostic utility of tumor location. The results argue for re-examination of the AJCC 8th edition staging system which incorporates tumor location. These data affirm that radical resection that includes part of the liver with the goal of negative margins followed by adjuvant therapy should be a standard clinical practice regardless of the primary tumor location for pT2 tumors. Future studies should focus on incorporating multiomic signatures for improved prognostication in gallbladder cancers.
Supplementary Material
Footnotes
SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section.
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the National Cancer Database. Restrictions apply to the availability of these data, which were used under license for this study. Data are available with the permission of the National Cancer Database.
REFERENCES
- 1.Lazcano-Ponce EC, Miquel JF, Munoz N, et al. Epidemiology and molecular pathology of gallbladder cancer. CA Cancer J Clin. 2001; 51(6):349–364. [DOI] [PubMed] [Google Scholar]
- 2.Rahman R, Simoes EJ, Schmaltz C, Jackson CS, Ibdah JA. Trend analysis and survival of primary gallbladder cancer in the United States: a 1973–2009 population-based study. Cancer Med 2017. Apr;6(4): 874–880. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Amin MB, Edge S, Greene F, et al. , eds. AJCC Cancer Staging Manual. 8th ed. Springer International Publishing: American Joint Commission on Cancer; 2017. [Google Scholar]
- 4.Shindoh J, de Aretxabala X, Aloia TA, et al. Tumor location is a strong predictor of tumor progression and survival in T2 gallbladder cancer: an international multicenter study. Ann Surg. 2015;261(4):733–739. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Wang L, Dong P, Zhang Y, et al. Prognostic validation of the updated 8th edition Tumor-Node-Metastasis classification by the Union for International Cancer Control: survival analyses of 307 patients with surgically treated gallbladder carcinoma. Oncol Lett 2018;16(4):4427–4433. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Cho JK, Lee W, Jang JY, et al. Validation of the oncologic effect of hepatic resection for T2 gallbladder cancer: a retrospective study. World J Surg Oncol. 2019;17(1):8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Fahim RB, Mc DJ, Richards JC, Ferris DO. Carcinoma of the gallbladder: a study of its modes of spread. Ann Surg. 1962. Jul;156: 114–124. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Clermont D Lymphatiques des voies biliaires cancer de la vésicule: anatomie pathologique et traitement chirurgical. Toulouse. 1909:814. [Google Scholar]
- 9.Lerro CC, Robbins AS, Phillips JL, Stewart AK. Comparison of cases captured in the national cancer data base with those in population-based central cancer registries. Ann Surg Oncol. 2013;20(6): 1759–1765. [DOI] [PubMed] [Google Scholar]
- 10.Bilimoria KY, Bentrem DJ, Stewart AK, Winchester DP, Ko CY. Comparison of commission on cancer-approved and -nonapproved hospitals in the United States: implications for studies that use the National Cancer Data Base. J Clin Oncol. 2009;27(25):4177–4181. [DOI] [PubMed] [Google Scholar]
- 11.Boffa DJ, Rosen JE, Mallin K, et al. Using the National Cancer Database for Outcomes Research: a review. JAMA Oncol. 2017;3(12): 1722–1728. [DOI] [PubMed] [Google Scholar]
- 12.Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45(6):613–619. [DOI] [PubMed] [Google Scholar]
- 13.Hosmer D Jr, Lemeshow S. Applied Survival Analysis: Time-to-Event. Hoboken, NJ: Wiley-Interscience; 1999. [Google Scholar]
- 14.Lancelot R, Lesnoff M, McDermott JJ. Use of akaike information criteria for model selection and inference. An application to assess prevention of gastrointestinal parasitism and respiratory mortality of Guinean goats in Kolda, Senegal. Prev Vet Med 2002;55(4):217–240. [DOI] [PubMed] [Google Scholar]
- 15.Lin HT, Liu GJ, Wu D, Lou JY. Metastasis of primary gallbladder carcinoma in lymph node and liver. World J Gastroenterol. 2005;11(5): 748–751. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Ito H, Ito K, D'Angelica M, et al. Accurate staging for gallbladder cancer: implications for surgical therapy and pathological assessment. Ann Surg. 2011;254(2):320–325. [DOI] [PubMed] [Google Scholar]
- 17.Wakai T, Shirai Y, Sakata J, Nagahashi M, Ajioka Y, Hatakeyama K. Mode of hepatic spread from gallbladder carcinoma: an immunohistochemical analysis of 42 hepatectomized specimens. Am J Surg Pathol. 2010;34(1):65–74. [DOI] [PubMed] [Google Scholar]
- 18.Nagahashi M Perimuscular connective tissue contains more and larger lymphatic vessels than the shallower layers in human gallbladders. World J Gastroenterol. 2007;13(33):4480–4483. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Lee AJ, Chiang YJ, Lee JE, et al. Validation of American Joint Committee on Cancer eighth staging system for gallbladder cancer and its lymphadenectomy guidelines. J Surg Res. 2018;230:148–154. [DOI] [PubMed] [Google Scholar]
- 20.Oweira H, Mehrabi A, Giryes A, Tekbas A, Abdel-Rahman O. External validation of the 8(th) American Joint Committee on Cancer staging system for gall bladder carcinoma. J Gastrointest Oncol. 2018;9(6): 1084–1090. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Ogundimu EO, Altman DG, Collins GS. Adequate sample size for developing prediction models is not simply related to events per variable. J Clin Epidemiol. 2016;76:175–182. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are available from the National Cancer Database. Restrictions apply to the availability of these data, which were used under license for this study. Data are available with the permission of the National Cancer Database.