Key Points

Question听 Among patients with esophageal cancer, does traveling to high-volume centers for an esophagectomy outweigh the benefits of more accessible, local care?

Findings听 In this cohort study of data for 17鈥�970 patients from the National Cancer Database, traveling longer distances to receive care at high-volume centers was linked with significantly improved overall survival at 1 and 5 years compared with treatment at local low-volume institutions.

Meaning听 Centralization of surgical management for esophageal cancer to high-volume esophagectomy centers may be linked with improved outcomes, particularly for patients with locoregionally advanced disease.

Importance听 Ongoing efforts have encouraged the regionalization of esophageal adenocarcinoma treatment to high-volume centers (HVCs). Yet such centralization has been linked with increased patient travel burden and reduced postoperative continuity of care.

Objective听 To determine whether traveling to undergo esophagectomy at HVCs is linked with superior overall survival compared with receiving care locally at low-volume centers (LVC).

Design, Setting, and Participants听 This cohort study considered data for all patients diagnosed with stage I through III esophageal adenocarcinoma in the 2010-2021 National Cancer Database. Patients were stratified based on distance traveled to receive care and the annual esophagectomy volume at the treating hospital: the travel-HVC cohort included patients in the top 25th percentile of travel burden who received care at centers in the top volume quartile, and the local-LVC cohort represented those in the bottom 25th percentile of travel burden who were treated at centers in the lowest volume quartile. Data were analyzed from July 2023 to January 2024.

Main Outcomes and Measures听 The primary end points were overall survival at 1 year and 5 years. Secondary end points included perioperative outcomes and factors linked with traveling to receive care.

Results听 Of 17鈥�970 patients, 2342 (13%) comprised the travel-HVC cohort, and 1969 (11%), the local-LVC cohort. The median (IQR) age was 65 (58-71) years; 3748 (87%) were male and 563 (13%) were female. After risk adjustment and with care at local LVCs as the reference, traveling to HVC was associated with superior survival at 1 year (hazard ratio for mortality [HR], 0.69; 95% CI, 0.58-0.83) and 5 years (HR, 0.80; 95% CI, 0.70-0.90). Stratifying by stage, traveling to HVCs was associated with comparable outcomes for stage I disease but reduced mortality for stage III (1-year HR, 0.72; 95% CI, 0.60-0.87; 5-year HR, 0.83; 95% CI, 0.74-0.93). Further, traveling to HVC was associated with greater lymph node harvest (尾, 5.08 nodes; 95% CI, 3.78-6.37) and likelihood of margin-negative resection (adjusted odds ratio, 1.83; 95% CI, 1.29-2.60).

Conclusions and Relevance听 Traveling to HVCs for esophagectomy was associated with improved 1-year and 5-year survival compared with receiving care locally at LVCs, particularly among patients with locoregionally advanced disease. Future studies are needed to ascertain barriers to care and develop novel targeted pathways to ensure equitable access to high-volume facilities and high-quality oncologic care.

With approximately 21鈥�000 new diagnoses each year and 5-year survival rates of only 15% to 25%, esophageal cancer stands as the sixth-leading cause of cancer-related deaths in the US.^1,2 Esophagectomy represents the standard of care for select patients with early stage or locally advanced disease.^3,4 Despite improvements in surgical technique, this operation remains high risk, with mortality ranging from 3% to 10%.^5-7 Notably, mortality rates after esophagectomy appear to vary across hospitals, with high-volume centers (HVCs) conferring significant reductions in the odds of death compared with low-volume institutions.^8,9 Thus, a large body of evidence has supported the regionalization of complex oncologic operations.^10-17 Importantly, the Leapfrog Hospital Survey established minimum volume standards and has encouraged hospital systems to take 鈥渧olume pledges鈥� for esophagectomy and other high-risk procedures.¹⁸

However, centralization to specialized centers has also been associated with increases in the distance patients must travel to receive care, in particular for esophageal cancer.¹⁹ Indeed, Diaz et al²⁰ found only 5 of 154 hospitals performing esophagectomy to have met Leapfrog standards. While greater travel distance is expected to limit access to appropriate oncologic treatment, prior literature has reported conflicting findings.^21-27 A study by Speicher et al²⁸ of patients with esophageal cancer diagnosed from 2006 to 2011 noted improved survival at 5 years among those who traveled to HVCs. Yet others have linked increased travel distance with lower likelihood of receiving chemoradiotherapy or adequate postoperative surveillance screening.^29-32 Moreover, given the highly morbid nature and prolonged recovery after esophagectomy, many patients may prefer to receive care locally in order to benefit from family and social support.³³ With continued efforts toward centralization of complex oncologic care, a contemporary understanding of the relationship between travel distance and hospital esophagectomy volume is warranted.

Therefore, we examined the association of travel distance and hospital volume with survival after esophagectomy for cancer. We compared patients undergoing operative management at local, low-volume centers (LVCs), with those traveling to distant HVCs. We hypothesized traveling to receive care at HVCs to be associated with superior survival at 1 and 5 years compared with local treatment at LVCs.

This was a retrospective cohort study of the National Cancer Database (NCDB). Because the data used in this work were derived from a deidentified NCDB file, this study was deemed exempt from full review by the institutional review board of the University of California, Los Angeles (No. 24-000294). This study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology () reporting guideline.

All adults (age 鈮�18 years) diagnosed with American Joint Committee on Cancer stage I through III esophageal adenocarcinoma were tabulated from the NCDB using relevant codes from the International Classification of Diseases for Oncology, Third Edition.³⁴ Only patients who underwent partial or total esophagectomy, with or without laryngectomy or gastrectomy, between January 2010 and December 2021 were considered for analysis.

To limit cohort heterogeneity, we excluded patients with cervical disease (22; 0.1%), as well as those undergoing palliative procedures (262; 1.2%). While records missing data regarding distance traveled were excluded from our main analysis (3069; 14.4%), we conducted a sensitivity analysis using multiple imputation, as reported below.

Patient, disease, and hospital factors were described in accordance with the NCDB Data Dictionary. Esophageal cancer staging was defined using guidelines from the National Comprehensive Cancer Network Tumor, Node, Metastasis, 8th Edition.³ The modified Charlson/Deyo comorbidity index (CDI) was used to quantify patient burden of chronic illness. Minimally invasive surgery was defined as both thoracoscopic and robotic-assisted esophagectomy. Academic or community hospital status was determined by the Commission on Cancer Accreditation program.

The distance between the centroid of each patient鈥檚 residential zip code and the address of the operating facility was calculated using the Haversine formula within the NCDB.³⁵ Travel distance was divided into quartiles; lowest travel distance was defined as 7.6 miles or less, while highest travel distance was set as 47.1 miles or more.

Annual hospital volume was computed for each center. We subsequently stratified hospitals into quartiles by volume. Centers with more than 19 cases per year were considered HVCs,³⁶ while those with fewer than 4 cases annually were classified as LVCs.

In our primary analysis, patients in the highest quartile of travel distance who received care at HVCs comprised the travel-HVC cohort, while those in the lowest quartile of travel distance who were treated at LVCs were considered the local-LVC group (Figure 1). As an additional sensitivity analysis, we considered the impact of care at HVC or LVC among both local and travel cohorts. Patients who were treated locally at HVCs were considered local-HVC, while those who traveled to LVCs were grouped as travel-LVC.

The primary outcome of this study was overall survival at 1 and 5 years. We secondarily considered several perioperative outcomes, including time to esophagectomy, lymph node harvest, and likelihood of complete resection, as well as duration of hospitalization, nonelective readmission, and mortality within 30 days or 90 days of esophagectomy. We also evaluated patient, disease, and hospital factors associated with traveling to receive care at HVC.

Normally distributed continuous data are reported as mean (SD), while non鈥搉ormally distributed data are detailed as median (IQR). Categorical variables are presented as cohort proportion (%). The significance of intergroup differences was evaluated using the Mann-Whitney U, adjusted Wald, or Pearson 蠂² tests, as appropriate. Standardized mean differences (SMDs) were used to elucidate true effect size, given the large sample size.

Survival was assessed using Kaplan-Meier time-to-event analyses. To adjust for patient clustering and underlying variation in hospital quality, we evaluated risk-adjusted survival using mixed-effects Cox regression analyses, following a gamma frailty distribution. Patient factors were considered as the first level, while hospital-level factors comprised the second.³⁷ Multilevel, mixed-effects multivariable regression models were developed to assess independent associations. Covariate selection was guided by elastic net regularization to minimize model overfitting and bias.³⁸ Selected covariates included patient age, sex, CDI, disease stage, tumor size and location, income, insurance coverage, receipt of neoadjuvant or adjuvant therapy, and year of diagnosis. To avoid attributing the time between diagnosis and surgery to postoperative survival, time zero was set as the date of resection. We then defined survival as the period between time zero and last contact or death. Patients alive at last follow-up were censored.

Following the main analysis, we performed 2 subgroup analyses, stratifying patients by disease stage and by care fragmentation status. We additionally conducted 2 sensitivity analyses. First, we evaluated outcomes across the travel-HVC, local-HVC, travel-LVC, and local-LVC groups. Second, to address the number of records missing data regarding distance traveled, we performed a sensitivity analysis using multiple imputation with chained equations.³⁹ Through this approach, missing variables are imputed conditional on other relevant variables, using regression models. Following creation of 20 imputed datasets, we repeated our survival analyses, and then calculated a pooled HR with CI.

Outputs of Cox survival models are presented as hazard ratios for mortality (HR). Mixed-effects logistic regression outputs are detailed as adjusted odds ratios, while linear regression outputs are reported as 尾, both with 95% CI. Statistical significance was set at 伪鈥�=鈥�.05. All statistical analyses were performed using Stata 16.1 (StataCorp). Data were analyzed from July 2023 to January 2024.

A total of 17鈥�970 patients with esophageal adenocarcinoma were included for analysis. The median (IQR) age was 65 (58-71) years; 15鈥�645 (87%) were male and 2325 (13%) were female. These patients traveled a median (IQR) of 18.7 miles (7.6-47.0 miles) to receive treatment (eFigure 1 in Supplement 1). Among these patients, 2342 (13%) traveled to HVCs and comprised the travel-HVC cohort, while 1969 (11%) were treated locally at LVCs and were classified as local-LVC.

Patients in the travel-HVC group were of similar age, sex, and comorbidity burden compared with those in local-LVC. Travel-HVC patients were more frequently of lowest income and from rural areas. Further, the travel-HVC group received care exclusively at academic hospitals. Relative to local-LVC patients, travel-HVC patients more commonly presented with stage I disease and more often received minimally invasive surgery (Table 1).

After risk adjustment, traveling to HVCs was associated with reduced hazards of mortality at 1 year (HR, 0.69; 95% CI, 0.58-0.83; P鈥�<鈥�.001) and 5 years (HR, 0.80; 95% CI, 0.70-0.90; P鈥�<鈥�.001), relative to care at local LVCs (Figure 2). Notably, other factors associated with greater mortality included increasing age, greater CDI, Medicaid insurance, and stage III disease (eTable 1 in Supplement 1).

Considering distance traveled and hospital case volume as continuous elements, increasing yearly esophagectomy volume was associated with reduced mortality hazard at 1 year (HR, 0.99/additional case; 95% CI, 0.99-1.00; P鈥�<鈥�.001) and 5 years (HR, 0.99/additional case; 95% CI, 0.99-1.00; P鈥�&濒迟;鈥�.001). Incremental 10-mile increases in distance traveled to receive care were associated with comparable survival at 1 year (HR, 1.00; 95% CI, 0.99-1.01; P鈥�=鈥�.45) but reduced mortality hazard at 5 years (HR, 0.99; 95% CI, 0.99-1.00; P鈥�=鈥�.045). The interaction between distance traveled and hospital case volume was not significant at 1 year, and at 5 years, the HR was 1.00 (95% CI, 1.00-1.00; P鈥�=鈥�.06).

After risk adjustment and with the local-LVC group as reference, the travel-HVC cohort experienced a 7.72-day delay (95% CI, 2.51-12.93 days; P鈥�=鈥�.004) in receiving an esophagectomy. However, traveling to receive care at HVCs was associated with increased lymph node harvest and greater likelihood of complete resection with negative margins.

Importantly, care at distant HVCs was similarly associated with an incremental increase in duration of hospitalization. While odds of mortality within 30 days were comparable between groups, the travel-HVC group demonstrated significantly reduced likelihood of mortality within 90 days (Table 2).

We repeated the primary analysis including an interaction term between travel to HVC and disease stage. On inclusion of this interaction term, traveling to HVC was associated with significantly improved survival at 1 year (HR, 0.56; 95% CI, 0.35-0.91; P鈥�=鈥�.02) and 5 years (HR, 0.62; 95% CI, 0.46-0.85; P鈥�=鈥�.003) (Figure 3).

Among patients with stage I adenocarcinoma, traveling to HVCs was associated with similar survival at both 1 and 5 years. Considering those with stage II adenocarcinoma, traveling to HVCs was associated with reduced survival at 1 year (HR, 0.54; 95% CI, 0.30-0.98; P鈥�=鈥�.04) and improved survival at 5 years (HR, 0.75; 95% CI, 0.55-1.02; P鈥�=鈥�.07). Evaluating patients with stage III disease, traveling to HVCs remained associated with reduced hazard of death at 1 year (HR, 0.72; 95% CI, 0.60-0.87; P鈥�=鈥�.001) and 5 years (HR, 0.83; 95% CI, 0.74-0.93; P鈥�=鈥�.002) (eTable 2 in Supplement 1).

Among the study cohort, 5663 patients (32%) received treatment at more than 1 facility while 12鈥�307 (68%) underwent their complete treatment course at a single institution. Considering those treated at a single institution, traveling to HVCs remained associated with improved survival at 1 year (HR, 0.70; 95% CI, 0.57-0.85; P鈥�<鈥�.001) and 5 years (HR, 0.80; 95% CI, 0.70-0.92; P鈥�=鈥�.002). Of those who received care at more than 1 center, travel to HVCs was also associated with improved survival at 1 year (HR, 0.65; 95% CI, 0.48-0.87; P鈥�=鈥�.004) and 5 years (HR, 0.79; 95% CI, 0.68-0.92; P鈥�=鈥�.003).

After risk adjustment, several factors remained associated with decreased likelihood of travel to HVCs, including Black race, Medicaid insurance, and lowest-quartile education (eFigure 2 and eTable 3 in Supplement 1). Increasing stage was also associated with decreased odds of travel to HVCs. Meanwhile, residence in rural areas was associated with dramatically elevated odds of traveling to HVCs, relative to living in metropolitan areas. Decreasing income quartile was similarly associated with increased likelihood of traveling to HVCs. Relative to care in the northeast, care in the west was associated with reduced odds of traveling to HVC, while treatment in the south was associated with increased travel likelihood.

Among all patients considered for analysis, 2342 (13%) comprised the travel-HVC cohort, 376 (2%) local-HVC, 1969 (11%) local-LVC, and 289 (2%) travel-LVC. After risk adjustment and with care at local LVCs as reference, traveling to distant LVCs was associated with comparable survival at 1 year (HR, 1.06; 95% CI, 0.80-1.40; P鈥�=鈥�.71) and 5 years (HR, 1.12; 95% CI, 0.93-1.35; P鈥�=鈥�.24). Meanwhile, the local-HVC cohort demonstrated similar 1-year survival as the local-LVC group (HR, 0.85; 95% CI, 0.64-1.13; P鈥�=鈥�.27) but significantly reduced mortality hazard at 5 years (HR, 0.71; 95% CI, 0.60-0.85; P鈥�&濒迟;鈥�.001). Finally, traveling to receive care at HVCs was associated with greater survival at 1 year (HR, 0.68; 95% CI, 0.57-0.82; P鈥�<鈥�.001) and 5 years (HR, 0.78; 95% CI, 0.71-0.86; P鈥�&濒迟;鈥�.001).

Subsequently, we evaluated the impact of distance traveled among patients treated at HVCs. Relative to the local-HVC group, the travel-HVC group demonstrated comparable outcomes at both 1 year (HR, 0.80; 95% CI, 0.60-1.07; P鈥�=鈥�.13) and 5 years (HR, 1.11; 95% CI, 0.92-1.34; P鈥�=鈥�.26). Lastly, considering only patients treated at LVCs, the travel-LVC patients faced comparable survival at 1 year (HR, 1.14; 95% CI, 0.86-1.50; P鈥�=鈥�.38) but increased mortality hazard at 5 years (HR, 1.53; 95% CI, 1.13-2.07; P鈥�=鈥�.007), with the local-LVC group as reference (eTable 4 in Supplement 1).

A sensitivity analysis using imputation for records missing the distance traveled yielded comparable findings as our main analysis. Namely, travel to HVCs was associated with improved survival at 1 year (HR, 0.66; 95% CI, 0.57-0.77; P鈥�<鈥�.001) and 5 years (HR, 0.79; 95% CI, 0.72-0.86; P鈥�&濒迟;鈥�.001).

In an era defined by ongoing centralization of complex oncologic care,^13,40 careful analysis of the consequences of regionalization remains paramount. In this work, we evaluated the impact of travel distance and hospital volume on survival after esophagectomy for cancer. Traveling longer distances to HVCs was associated with greater survival at 1 and 5 years relative to care at local low-volume hospitals. Such survival benefit appeared to be most significant for patients with locoregionally advanced, stage II through III adenocarcinoma. With implications toward national policy and practice, as well as the broader centralization of complex oncologic care, several of these findings merit further discussion.

A large body of work has linked increasing hospital volume with superior outcomes after oncologic operations and considered it a surrogate for quality.^8,10-17 Indeed, centralization of esophagectomy to select higher-volume hospitals has been proposed to dramatically reduce postoperative mortality by more than 12%.⁴¹ Yet regionalization has also been suggested to negatively impact postoperative continuity of care,^21,30 yielding ambiguous benefits. In our study, we found traveling to undergo esophagectomy at HVCs to be associated with significantly improved overall survival at both 1 and 5 years. This finding persisted after comprehensive multivariable risk adjustment and consideration of hospital clustering and remained true even among patients who experienced fragmented care. Altogether, our work suggests greater travel burden does not undermine the benefit of care at HVCs and assuages existing concerns surrounding the downsides of care centralization. Importantly, we found traveling to HVCs to be associated with the greatest survival advantage for patients with stage II or III disease. These institutions may have established, multidisciplinary care pathways, greater experience perioperatively managing advanced tumor burdens, and access to advanced therapies or clinical trials, which could prove life-saving for such patients. Therefore, our findings support regionalization of care for patients with esophageal adenocarcinoma and in particular those who require more complex or multimodal management approaches.

We noted patients traveling to HVCs to have an increased number of lymph nodes harvested, experience greater likelihood of negative surgical margins, and face reduced odds of death within 90 days. These findings echo prior literature reporting improved oncologic efficacy at high-volume institutions and may underlie the noted survival differences in the longer term. Further, patients who traveled farther distances experienced an incremental increase in duration of hospitalization but comparable nonelective readmission within 30 days of discharge. Discharge planning may be expectedly more complex for this cohort, given the need to coordinate transportation. Yet patients who travel for oncologic treatment may not have access to local outpatient care and could face rehospitalization on incidence of postoperative sequelae, including dehydration, uncontrolled pain, or feeding tube dislodgement. As the present work could only capture readmission to the treating hospital, future studies should consider whether increased travel distance may contribute to postoperative care fragmentation and readmission to centers closer to the patient鈥檚 residence. Recognizing the challenges of caring for a more geographically widespread population, HVCs should continue to expand their armamentarium of telehealth resources and other approaches to enhance continuity of care.

While we could not ascertain the precise reasons patients elected to receive local or distant care, we hypothesize the presence of both socioeconomic and health network effects. Patients of White race, lowest socioeconomic status, and rural residency demonstrated significantly greater adjusted likelihood of traveling to HVCs. Therefore, our findings align with a growing body of evidence linking low-income, largely rural areas with increased barriers to appropriate oncologic care.^26,42-46 Importantly, these disparities may only be exacerbated by contemporary trends in the surgical workforce, with a disproportionate share of specialists choosing to live in metropolitan cities over rural regions.⁴⁷ This increasing geographic isolation may most significantly affect access to care among already vulnerable patient populations. Indeed, prior literature has reported that racial and ethnic minority, uninsured, and undereducated populations less frequently receive care at HVCs.^34,48 In our study, these patients were also substantially less likely to travel to distant HVCs. Given the known volume-outcome relationship for esophagectomy, clinicians and hospitals must seek to build and expand referral networks across communities, counties, and even states, to ensure that all patients have access to optimal surgical treatment. Finally, proximity to family and community may underlie facility choice, particularly given the high morbidity associated with esophagectomy. A landmark study by Finlayson et al³³ revealed a strong patient bias to undergo surgery locally, largely due to the accessibility of support systems, relationships with clinicians, and reduced travel burden. Notably, patients were willing to accept nearly double the mortality risk to remain close to home for their care. Therefore, in addition to ensuring equitable access to high-quality oncologic care, clinical teams must consider socioeconomic vulnerability, exposure to adverse social determinants of health, and access to psychosocial support. Alongside continued advocacy for centralization of care,⁴⁹ future studies are needed to elucidate the precise barriers preventing certain patients from receiving care at HVCs and develop appropriate interventions to improve their perioperative and long-term outcomes.

This present work has several limitations. Timing in disease onset or severity at presentation was not assessed. While we adjusted for CDI, we could not ascertain patient frailty, smoking history, or functional status. Our analysis was limited to patients who received an esophagectomy; we could not evaluate those deemed not surgical candidates. Broadening the study to all patients diagnosed with esophageal adenocarcinoma may reveal significant short-term differences across centers, with low-volume hospitals potentially treating greater numbers of patients who do not meet surgical candidacy. The reason for choice of operating facility was not provided. The NCDB does not report whether patients completed their neoadjuvant or adjuvant treatment courses. While we could identify those who were treated at multiple hospitals, we could not determine if patients received chemoradiation at the same institution at which they underwent surgery. The NCDB does not detail individual surgeon experience or case volume, but these factors may be of significance for future study. Further, the NCDB reports straight-line distance, which is accepted as a proxy for travel time but may underestimate driving distance by 20% to 30%.⁵⁰ We could not account for patient access to or use of transportation. Yet we recognize the same distance may represent different burdens based on the mode of transportation available. We additionally could not access data regarding reoperation rates. While we performed several subgroup and sensitivity analyses, they were limited by available sample size. Finally, the NCDB does not report disease-free survival or cause of death.

This study found that traveling longer distances to HVCs was associated with improved survival and perioperative outcomes after esophagectomy for esophageal cancer. Increasing center volume and travel distance were independently associated with improved outcomes over 5 years. Notably, patients with locoregionally advanced adenocarcinoma appear to experience the greatest impact of receiving care at these high-volume facilities. Patients of White race, lowest income, and rural residency faced greater travel burden relative to others. Future studies are needed to ascertain barriers to treatment and develop novel targeted pathways to ensure equitable access to high-volume facilities and high-quality oncologic care.

Accepted for Publication: August 24, 2024.

Published Online: November 13, 2024. doi:10.1001/jamasurg.2024.5009

Corresponding Author: Peyman Benharash, MD, MS, UCLA Division of Cardiac Surgery, University of California, Los Angeles, 10833 Le Conte Ave, 64-249 Center for Health Sciences, Los Angeles, CA 90095 (pbenharash@mednet.ucla.edu).

Author Contributions: Dr Benharash had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: All authors.

Acquisition, analysis, or interpretation of data: Sakowitz, Bakhtiyar, Mallick, Benharash.

Drafting of the manuscript: Sakowitz, Bakhtiyar, Benharash.

Critical review of the manuscript for important intellectual content: All authors.

Statistical analysis: Sakowitz, Bakhtiyar, Mallick, Benharash.

Administrative, technical, or material support: Sakowitz, Benharash.

Supervision: All authors.

Conflict of Interest Disclosures: Dr Yanagawa reported being an owner of ICONA, being an advisory board attendee and steering committee member for Astra Zeneca and lung cancer workshop attendee for OncLive, personal fees from Ideology, and grants from Lungevity outside the submitted work. Dr Benharash reported being a proctor for AtriCure outside the submitted work. No other disclosures were reported.

Meeting Presentation: This work was presented at the 95th Annual Meeting of the Pacific Coast Surgical Association; February 9, 2024; Rancho Mirage, California.

Data Sharing Statement: See Supplement 2.