Multi-institutional care is associated with improved survival in biliary tract cancer

Introduction

Biliary tract cancers (BTC), including intrahepatic cholangiocarcinoma (IHC), extrahepatic cholangiocarcinoma (EHC), and gallbladder cancer (GBC), are rare but aggressive malignancies with rising global incidence and poor survival (1-3). Curative-intent treatment requires complex, multidisciplinary management involving hepatobiliary surgery, medical oncology, interventional radiology, and specialized pathology (4). These resources are often concentrated at high-volume academic centers, where treatment is associated with lower perioperative mortality, higher rates of margin-negative resection, and improved long-term survival (5-8).

With cancer management becoming more centralized, care managed across multiple facilities is increasingly common (9,10). This pattern of multi-institutional care (MC) has been associated with worse outcomes in gastric, esophageal, and rectal cancers (11,12), often attributed to care fragmentation and delays. However, no large-scale studies have examined the impact of MC in BTC, where rarity and complexity may influence outcomes differently.

In this context, we conducted a retrospective cohort study using the National Cancer Database (NCDB) to identify factors associated with MC in BTC and to assess its relationship with overall survival (OS). Based on prior literature on gastrointestinal cancer, we hypothesized that MC would be associated with worse OS. We present this article in accordance with the STROBE reporting checklist (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-1-1062/rc) (13).

Methods

Data source and cohort selection

We performed a retrospective cohort study using the 2022 Participant User File (PUF) of the NCDB (14). We identified adults (aged 18 years or older) with a primary IHC, EHC, or GBC diagnosed between January 1, 2004, and December 31, 2022. We restricted the primary analysis to patients with stage I–III disease who underwent curative-intent resection at their reporting Commission on Cancer (CoC) center, denoted by site-specific surgery codes and primary site flags. Patients were excluded if they had T1a GBC or if their surgery was coded as local tumor destruction, local tumor excision, unspecified, or unknown. Those with missing key demographic or clinical data were also excluded, assuming that missingness occurred completely at random.

It was deemed exempt by the Institutional Review Board of University Hospitals Cleveland Medical Center as the NCDB contains only de-identified patient records. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Outcomes

Our primary outcome was MC, defined using the NCDB’s binary variable “puf_mult_source” to indicate whether a patient’s entire treatment course was recorded at a single CoC facility or at multiple facilities. The secondary outcome was OS, measured from the date of definitive surgical resection to the date of death. Patients still alive at the last follow-up were censored.

Covariates

Covariates spanned four domains. Patient demographics included age at diagnosis (18–49, 50–64, 65–74, ≥75 years), sex, and race/ethnicity (non-Hispanic White, non-Hispanic Black, non-Hispanic Asian, Hispanic, other), insurance payer (private, Medicare, Medicaid, not insured, other/unknown), and a composite socioeconomic status (SES) Index Score (1= lowest to 7= highest SES), calculated by summing ZIP code-area income and education quartiles minus one based on a previously described method (15). Clinical factors included Charlson-Deyo comorbidity score (0, 1, 2, ≥3), primary tumor site (gallbladder, intrahepatic bile duct, extrahepatic bile duct), American Joint Committee on Cancer (AJCC) analytic stage group (I–IV), margin status (negative vs. positive), and nodal status (negative vs. positive).

Treatment-geography factors were traveling distance to the reporting facility (0–10, 11–50, 51–200, >200 miles), rural-urban designation, and U.S. Census region. Facility characteristics included the type of CoC accreditation [academic/research program, community cancer program (CCP), integrated network cancer program, or comprehensive CCP] and volume. Annualized biliary resection volume was calculated for each center as its total number of operative biliary resections divided by the number of calendar years in which it performed ≥1 resection, dichotomized using the top quartile (≥8.7 resections/year) into high- and low-volume centers.

Statistical analysis

The overall goal of this study was to identify factors associated with MC and to determine its impact on long-term survival in BTC. We first compared baseline patient, tumor, treatment, and facility characteristics between those treated at a single facility versus those treated at multiple facilities. Continuous variables are reported as medians with interquartile ranges and compared using the Wilcoxon rank-sum test. Categorical variables are presented as counts and percentages and compared by Chi-squared tests. We also summarized distributions of treatment type by stage, categorizing patients as receiving surgery alone, surgery plus chemotherapy, or surgery, chemotherapy, and radiation.

Predictors of MC were evaluated with a multivariable logistic regression model that included all covariates. Adjusted odds ratios (ORs) with 95% confidence intervals (CIs) were calculated. A multivariable Cox proportional-hazards model incorporating the same covariates, plus MC care, was then used to estimate adjusted hazard ratios (HRs) for mortality, with proportional-hazards assumptions verified using Schoenfeld residuals.

OS was visualized with Kaplan-Meier (KM) curves stratified by whether patients received all care at a single facility versus multiple facilities, with differences assessed using log-rank tests. To examine the influence of surgical volume, we also generated KM curves for four subgroups defined by care setting and annualized biliary resection volume, reporting median OS and comparing survival distributions using log-rank tests. Pairwise log-rank tests with Holm adjustment were additionally performed across the four subgroups. All KM curves were truncated when the number at risk in any stratum fell below 10% of the original cohort to comply with the NCDB’s cell-count reporting guidelines.

All statistical tests were two-sided, with P<0.05 considered statistically significant. Analyses were performed in RStudio (version 2025.05.0+496).

Results

Patient characteristics

A total of 13,250 patients met the inclusion criteria (Figure 1), including 10,426 (78.7%) treated at a single CoC-accredited facility and 2,824 (21.3%) who received MC. Patients receiving MC were younger (39.5% vs. 34.8% aged 18–64), more often non-Hispanic White (75.0% vs. 66.9%), more likely to have private insurance (35.2% vs. 30.5%), and more frequently in the highest SES category (19.7% vs. 17.7%) (all P<0.001) (Table 1).

Figure 1 STROBE diagram of patient selection. BTC, biliary tract cancer.

Most patients in both cohorts lived in metropolitan areas, though this was slightly less common among MC patients (83.8% vs. 87.1%, P<0.001). MC patients lived farther from their treating facility, most commonly 11–50 miles away (42.2% vs. 37.8%), whereas single-facility patients more often lived less than 10 miles away (44.6% vs. 33.0%) (P<0.001). Both groups most often received care at academic programs (52.3% of single-facility vs. 48.6% of MC patients) and at low-volume centers (66.3% vs. 62.0%), though the distributions of facility type and volume differed significantly (both P<0.001).

Stage distribution varied modestly (P<0.001), with slightly more Stage III disease among MC patients (38.9% vs. 36.0%). Treatment patterns also varied by stage, with surgery alone most common in Stage I and multimodality therapy increasingly used in Stages II and III (Table S1).

Primary tumor site also differed (P<0.001): patients with GBC were more likely to receive single-facility care (43.1% vs. 40.9%), whereas those with IHC were more likely to receive MC care (19.6% vs. 17.3%). MC patients had slightly lower comorbidity burden (Charlson-Deyo score of 0: 67.5% vs. 65.2%, P=0.002) and marginally higher lymph node involvement [median 0 (IQR 0–1.0) vs. 0 (IQR 0–1.0), P=0.001], though margin status did not differ between groups (P=0.18).

Predictors of MC

In multivariable analysis, several demographic, clinical, and facility factors independently predicted receipt of MC care (Figure 2). Older age was associated with lower odds of MC, with patients 75 and older less likely than those aged 18–49 to receive care across multiple centers (OR 0.56, 95% CI: 0.45–0.71). Compared with non-Hispanic White patients, non-Hispanic Black and Hispanic patients had lower odds of MC (ORs 0.82 and 0.64, respectively; both P<0.001). Uninsured patients were also less likely to receive MC (OR 0.625, 95% CI: 0.46–0.90), whereas Medicaid and Medicare beneficiaries did not differ from privately insured patients. Higher SES was associated with increased MC (OR 1.28, 95% CI: 1.06–1.55) compared to the lowest SES.

Figure 2 Forest plot of multivariable logistic regression showing adjusted ORs and 95% CIs for predictors of receiving treatment at more than one CoC center. CCP, community cancer program; CI, confidence interval; CoC, Commission on Cancer; EHC, extrahepatic cholangiocarcinoma; IHC, intrahepatic cholangiocarcinoma; OR, odds ratio; SES, socioeconomic status.

Patients with ≥3 comorbidities were less likely to receive MC than those with none (OR 0.70, 95% CI: 0.56–0.86). Stage was an independent predictor: Stage II (OR 1.24, 95% CI: 1.10–1.41) and Stage III (OR 1.35, 95% CI: 1.19–1.54) were both associated with higher odds of MC than Stage I disease.

Geographic and facility-level factors demonstrated strong associations. Compared with the Northeast, patients in the Midwest, South, and West were less likely to receive MC (ORs 0.85, 0.74, and 0.76, respectively). Odds of MC increased steadily with travel distance: 11–50 miles (OR 1.50, 95% CI: 1.35–1.66), 51–200 miles (OR 1.95, 95% CI: 1.69–2.26), and >200 miles (OR 2.51, 95% CI: 1.93–3.25) relative to <10 miles (all P<0.001). Facility type was the strongest predictor, with patients treated at CCPs having more than twice the odds of MC compared with those at academic centers (OR 2.25, 95% CI: 1.79–2.83). Annual surgical volume was not associated with MC (OR 0.93, 95% CI: 0.84–1.03).

Survival analyses

KM analysis demonstrated superior OS among patients treated at multiple facilities compared with those treated at a single center (median OS of 43.7 vs. 35.7 months; log-rank P<0.0001; Figure 3). When stratified by both care pattern and annualized facility volume, multiple low-volume centers had the highest median OS (44.1 months), followed by multiple high-volume (43.2 months), single high-volume (38.4 months), and single low-volume centers (34.3 months) (log-rank P<0.0001; Figure 4).

Figure 3 Overall survival by single vs. multiple-facility care.

Figure 4 Overall survival by facility volume and single vs. multiple-facility care.

Pairwise comparisons confirmed that patients treated at multiple low-volume centers had significantly better survival than those treated at single low- or high-volume centers (all P<0.05). In contrast, survival did not differ between multiple low- and multiple high-volume centers (P=0.51), nor between single low- and single high-volume centers (P=0.056).

In a multivariable Cox hazards model, MC remained independently associated with lower mortality (HR 0.85, 95% CI: 0.80–0.89) (Table 2). Mortality increased with advancing age, reaching HR 1.66 for patients over 75 (P<0.001), and comorbidity burden (Charlson-Deyo score ≥3: HR 1.46, 95% CI: 1.32–1.60). Female sex was protective (HR 0.89, 95% CI: 0.86–0.93). Compared with non-Hispanic White patients, Asian (HR 0.80, 95% CI: 0.72–0.88) and non-Hispanic Black patients (HR 0.82, 95% CI: 0.75–0.89) had lower mortality. Medicare insurance was associated with worse survival (HR 1.14, 95% CI: 1.06–1.21), while Medicaid and uninsured status were not. Higher SES showed modest protection (SES 7: HR 0.90, 95% CI: 0.82–0.99).

Tumor-level factors were among the strongest predictors: both intrahepatic (HR 1.18, 95% CI: 1.10–1.27) and extrahepatic cholangiocarcinoma (HR 1.63, 95% CI: 1.54–1.72) had higher mortality than GBC, and mortality rose sharply with advancing stage (Stage III: HR 2.72, 95% CI: 2.53–2.92).

Regional differences persisted, with patients in the Midwest (HR 1.17, 95% CI: 1.09–1.25) and the South (HR 1.10, 95% CI: 1.03–1.17) experiencing higher mortality than those in the Northeast. Facility characteristics were also significant. Compared with academic centers, CCPs, comprehensive CCPs, and integrated network programs all had higher mortality (HRs 1.15–1.38). Surgical volume and travel distance were not associated with OS.

Discussion

In this national cohort of patients undergoing curative-intent resection for BTC, approximately 1 in 5 received MC. In adjusted analysis, MC was more likely among younger, non-Hispanic White, privately insured patients with higher SES, those living farther from their treating facility, and those treated at CCPs. Conversely, it was less likely among older, uninsured, and minority patients and among those with more comorbidities. MC was independently associated with improved survival for resected BTCs, and stratified analyses showed the poorest outcomes at single low-volume centers.

These findings suggest that the relationship between MC and outcomes in BTC may differ from patterns observed in other gastrointestinal cancers, in which MC has often been associated with poorer survival (16,17). One possible explanation is that in BTC, MC may partially reflect planned referral to hepatobiliary centers for surgery, an approach specifically recommended by the American Association for the Study of Liver Diseases (10), followed by adjuvant therapy closer to home. However, because the NCDB does not capture these granular data, MC likely encompasses heterogeneous pathways that cannot be distinguished in this dataset. Large cohort studies link access to curative care and improved survival to treatment at high-volume and academic centers, which often require care across multiple institutions (8). Although MC in our study was most common among patients treated at community or integrated network cancer programs, this may reflect a pathway in which diagnosis or non-surgical therapy occurs locally, with referral to specialized centers for surgery. Furthermore, the increasing use of molecular profiling and targeted therapies, as well as the integration of clinical trials, further drives multi-institutional collaboration (18).

The sociodemographic and clinical patterns associated with MC highlight potential disparities in access to care. Younger, non-Hispanic White, privately insured, and higher-SES patients were more likely to receive MC, whereas older, uninsured, and minority patients were less likely. Similar patterns have been reported in retrospective studies of rectal, gastric, and esophageal cancer (19-21). Patients treated at community cancer centers were twice as likely to seek MC, likely reflecting a care pathway in which diagnosis or non-surgical therapy occurs locally, with subsequent referral to specialized centers for resection. Although surgical volume was not a predictor of receiving MC, comparisons of high- versus low-volume centers within the MC subgroup should be interpreted cautiously, given the selected nature of this population and potential differences in case-mix and referral thresholds. The absence of a detectable volume effect within MC should not be construed as evidence that operative volume is unimportant or that MC offsets volume-related differences. Rather, the observed associations may reflect structural and process-level advantages more common in certain facility types—such as multidisciplinary teams, specialized perioperative care, and access to advanced therapies—that can shape referral patterns and outcomes (22). Evidence from other complex fields, such as transplantation, demonstrates that multidisciplinary oversight and attention to social determinants of health improve long-term outcomes (23,24). Similar mechanisms may explain the improved outcomes we observed among BTC patients receiving MC.

Beyond these structural factors, patient-level characteristics likely influence both access to MC and outcomes. An important unmeasured confounder in the NCDB is patient activation, defined as patients’ knowledge, skills, and confidence in engaging in complex disease management, which has been linked to smoother perioperative processes, enhanced recovery, and more effective healthcare utilization (25). Studies show that even when surgery is technically feasible, demographic differences influence who undergoes operative management. A recent national analysis found that Black patients with resectable cancers had the highest rates of declined surgery compared to White patients across cancer types (26), and another demonstrated that refusal was also associated with older age, lack of insurance, unmarried status, and advanced stage (27).

These patterns parallel our observation that older, minority, and underinsured patients were less likely to receive MC, suggesting that unmeasured factors such as trust in the healthcare system, health literacy, and willingness to navigate complex referral pathways influence both access to surgery and patterns of care delivery. Residual confounding by patient-level factors may contribute to the observed association between MC and survival. In this context, the survival advantage associated with MC may partly reflect the characteristics of patients who are “most activated” and resourced to seek specialized care. Future qualitative and mixed-methods research should clarify how patient- and system-level factors shape access to MC and distinguish coordinated shared-care from unplanned transfers, particularly among the vulnerable populations identified.

This study has limitations. First, our analytic cohort was restricted to patients with stage I–III BTC who underwent curative-intent resection at CoC-accredited facilities; therefore, our findings reflect associations conditional on receipt of surgery and should not be interpreted as the causal effect of MC. Patients who were not surgical candidates, declined surgery, or were unable to access referral networks are underrepresented, and referral/selection mechanisms may contribute to observed survival differences. Second, MC in the NCDB is captured by “puf_mult_source”, a binary indicator of whether more than one CoC-accredited facility submitted a report for the case. The dataset does not provide the number of reporting facilities, nor does it capture the sequence, timing, or intent of multi-facility involvement. Consequently, we cannot distinguish coordinated referral-based MC from unplanned transfers. It also raises the possibility of immortal time bias when MC is modeled as a fixed exposure. Third, because MC is derived from reporting across CoC-accredited facilities, care delivered in outpatient or non-CoC settings is not captured, potentially underestimating the prevalence of MC. NCDB also lacks recurrence, cause-specific mortality, and patient-reported outcomes. Finally, IHC, EHC, and GBC were analyzed as a single cohort; although clinically distinct, they share overlapping surgical principles, referral pathways, and multidisciplinary care needs, supporting a combined analysis.

Conclusions

In this national cohort of resected stage I–III BTC, MC was associated with improved survival, particularly when surgery was performed at high-volume centers. These findings are consistent with a coordinated care model that centralizes specialized surgery while allowing other treatments to be administered locally. Observed differences in outcomes may be influenced by a complex interplay of referral patterns, facility resources, and patient engagement that shape access to MC, and should not be interpreted causally. Future work should characterize referral pathways and barriers to accessing support to inform the development of equitable, streamlined care models for patients most likely to benefit from specialized surgical care.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-1-1062/rc

Peer Review File: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-1-1062/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-1-1062/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Valle JW, Kelley RK, Nervi B, et al. Biliary tract cancer. Lancet 2021;397:428-44. [Crossref] [PubMed]
2. Razumilava N, Gores GJ. Classification, diagnosis, and management of cholangiocarcinoma. Clin Gastroenterol Hepatol 2013;11:13-21.e1; quiz e3-4. [Crossref] [PubMed]
3. Baria K, De Toni EN, Yu B, et al. Worldwide Incidence and Mortality of Biliary Tract Cancer. Gastro Hep Adv 2022;1:618-26. [Crossref] [PubMed]
4. Saha SK, Zhu AX, Fuchs CS, et al. Forty-Year Trends in Cholangiocarcinoma Incidence in the U.S.: Intrahepatic Disease on the Rise. Oncologist 2016;21:594-9. [Crossref] [PubMed]
5. Henley SJ, Weir HK, Jim MA, et al. Gallbladder Cancer Incidence and Mortality, United States 1999-2011. Cancer Epidemiol Biomarkers Prev 2015;24:1319-26. [Crossref] [PubMed]
6. Zhu X, Zhang X, Hu X, et al. Survival analysis of patients with primary gallbladder cancer from 2010 to 2015: A retrospective study based on SEER data. Medicine (Baltimore) 2020;99:e22292. [Crossref] [PubMed]
7. Elgenidy A, Afifi AM, Jalal PK. Survival and Causes of Death among Patients with Intrahepatic Cholangiocarcinoma in the United States from 2000 to 2018. Cancer Epidemiol Biomarkers Prev 2022;31:2169-76. [Crossref] [PubMed]
8. Tzedakis S, Challine A, Katsahian S, et al. Clinical care pathways of patients with biliary tract cancer: A French nationwide longitudinal cohort study. Eur J Cancer 2024;202:114018. [Crossref] [PubMed]
9. Shroff RT, Kennedy EB, Bachini M, et al. Adjuvant Therapy for Resected Biliary Tract Cancer: ASCO Clinical Practice Guideline. J Clin Oncol 2019;37:1015-27. [Crossref] [PubMed]
10. Bowlus CL, Arrivé L, Bergquist A, et al. AASLD practice guidance on primary sclerosing cholangitis and cholangiocarcinoma. Hepatology 2023;77:659-702. [Crossref] [PubMed]
11. Tay E, Gambhir S, Stopenski S, et al. Outcomes of Complex Gastrointestinal Cancer Resection at US News & World Report Top-Ranked vs Non-Ranked Hospitals. J Am Coll Surg 2021;233:21-7.e1. [Crossref] [PubMed]
12. Lee GC, Gamblin TC, Fong ZV, et al. Facility Type is Associated with Margin Status and Overall Survival of Patients with Resected Intrahepatic Cholangiocarcinoma. Ann Surg Oncol 2019;26:4091-9. [Crossref] [PubMed]
13. von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol 2008;61:344-9. [Crossref] [PubMed]
14. American College of Surgeons. National Cancer Database. [Internet]. [cited 2025 Jun 3]. Available online: https://www.facs.org/quality-programs/cancer-programs/national-cancer-database/
15. Swords DS, Scaife CL. Decompositions of the Contribution of Treatment Disparities to Survival Disparities in Stage I-II Pancreatic Adenocarcinoma. Ann Surg Oncol 2021;28:3157-68. [Crossref] [PubMed]
16. Rhodin KE, Raman V, Eckhoff A, et al. Patterns and Impact of Fragmented Care in Stage II and III Gastric Cancer. Ann Surg Oncol 2022;29:5422-31. [Crossref] [PubMed]
17. Ngongoni RF, Timmerhuis HC, Li AY, et al. Association of care fragmentation and hospital cancer designation with survival in gastroesophageal junction cancer: a statewide study. J Gastrointest Surg 2025;29:101962. [Crossref] [PubMed]
18. Merters J, Lamarca A. Integrating cytotoxic, targeted and immune therapies for cholangiocarcinoma. J Hepatol 2023;78:652-7. [Crossref] [PubMed]
19. Shannon AB, Mo J, Song Y, et al. Does multicenter care impact the outcomes of surgical patients with gastrointestinal malignancies requiring complex multimodality therapy? J Surg Oncol 2020;122:729-38. [Crossref] [PubMed]
20. Freischlag K, Olivere L, Turner M, et al. Does Fragmentation of Care in Locally Advanced Rectal Cancer Increase Patient Mortality? J Gastrointest Surg 2021;25:1287-96. [Crossref] [PubMed]
21. Rhodin KE, Raman V, Jensen CW, et al. Multi-institutional Care in Clinical Stage II and III Esophageal Cancer. Ann Thorac Surg 2023;115:370-7. [Crossref] [PubMed]
22. Gunasekaran G, Bekki Y, Lourdusamy V, et al. Surgical Treatments of Hepatobiliary Cancers. Hepatology 2021;73:128-36. [Crossref] [PubMed]
23. Khera N, Martin P, Edsall K, et al. Patient-centered care coordination in hematopoietic cell transplantation. Blood Adv 2017;1:1617-27. [Crossref] [PubMed]
24. George RP, Winterberg PD, Garro R. Multidisciplinary and multidimensional approaches to transplantation in children with rare genetic kidney diseases. Pediatr Transplant 2023;27:e14567. [Crossref] [PubMed]
25. Yin Y, Zhang J, Long X, et al. Application and advances of patient activation in surgical patients. World J Surg Oncol 2025;23:261. [Crossref] [PubMed]
26. Patel VR, Liu M, Snyder RA, et al. Trends in Racial and Ethnic Differences in Declined Surgery for Resectable Malignancies in the United States. Ann Surg 2025;281:711-9. [Crossref] [PubMed]
27. Rapp J, Tuminello S, Alpert N, et al. Disparities in surgery for early-stage cancer: the impact of refusal. Cancer Causes Control 2019;30:1389-97. [Crossref] [PubMed]