Patterns of Care in Neoadjuvant Chemoradiotherapy for Node Positive Esophageal Adenocarcinoma

Abstract

Background:

The aims of this study were to examine the factors associated with use of neoadjuvant chemoradiotherapy (NCR) for patients with locally advanced esophageal cancer, and to evaluate the effect of NCR on survival.

Methods:

The 2004–2015 National Cancer Database (NCDB) was used to identify patients with cT1–4aN1–3M0 (stage II-IVA) esophageal adenocarcinoma who underwent esophagectomy. Patients were stratified by receipt of NCR. A multivariable logistic regression was performed to examine factors associated with NCR, and survival between the two groups was compared using a multivariable Cox model.

Results:

Of 8076 patients meeting study criteria, 1616 (20%) did not receive NCR and 6460 (80%) did. In a multivariable regression, factors associated with receipt of neoadjuvant therapy included a later year of diagnosis, treatment in a high-volume center, and clinical stage III disease. Factors associated with non-receipt of NCR included increasing age, comorbidities, and treatment in a Middle Atlantic, South Central, or Pacific state. Receipt of trimodality therapy was associated with improved survival compared to other or no perioperative therapies (adjusted HR 0.80; 95%CI 0.74–0.87).

Conclusions:

Numerous personal, demographic, and treatment center related factors account for variability in neoadjuvant chemoradiotherapy for clinically node-positive esophageal adenocarcinoma, although neoadjuvant therapy was associated with a survival benefit. Further efforts are needed to identify reasons for these differences and design interventions to provide more equitable care for patients with esophageal cancer.

Introduction

Since the early 2000s, trimodality therapy, including induction chemoradiation followed by surgery, has been commonly offered for locally advanced esophageal cancer^1–6. The National Comprehensive Cancer Network (NCCN) guidelines recommend neoadjuvant chemotherapy and radiation (NCR) for patients with clinically node-positive esophageal adenocarcinoma, although the guidelines also allow pre- and perioperative chemotherapy regimens like 5-fluorouracil, leucovorin, oxaliplatin, and docetaxel (FLOT)⁷. However, the patterns of use of induction chemoradiotherapy have not been well described. There is increasing recognition that disparities in age, race, sex, socioeconomic status, insurance, education level, geography, and type of treatment center affect the care of patients with cancer across the world^8–11. Unlike the regulated setting of randomized controlled trials, dissemination of treatment is limited by variability. For instance, studies suggest that while black patients have a higher incidence rate of esophageal cancer in the United States, they, along with other non-white patients, are less likely to receive surgery for operable disease^12–14. While there are studies examining the disparities in surgery for esophageal cancer, and there is evidence that disparities limit receipt of neoadjuvant chemotherapy in bladder and gastric cancer^15,16, data are limited regarding the factors associated with use of induction chemoradiotherapy as part of the multimodality treatment of esophageal cancer.

We used a large, national registry to examine the patterns of use of neoadjuvant chemoradiation in patients with clinically node-positive esophageal adenocarcinoma, hypothesizing that differences in age, race, and socioeconomic status would be associated with non-receipt of NCR. Further, we compared the survival of patients receiving trimodality therapy and those who did not to evaluate the effect of NCR on outcomes in locally advanced esophageal cancer.

Methods

Data Source and Patient Selection

The study was deemed exempt by our Institutional Review Board. The 2004–2015 National Cancer Database (NCDB)¹⁷ was used to identify patients with American Joint Commission on Cancer (AJCC) Eighth Edition clinical T1–4aN1–3M0 (stage II-IVA) esophageal adenocarcinoma who underwent surgery. Patients with missing survival information and who were not offered chemotherapy or radiation because they were deemed unfit were excluded (Figure 1). Patients were stratified into two groups: no NCR or NCR. Patients were deemed to have been treated with NCR if they received neoadjuvant chemotherapy and radiation. All other patients were classified as non-NCR. While the NCCN recommends concurrent induction chemoradiation with radiation greater than 40.0 Gray (Gy), we did not apply these criteria to the definition of NCR because although the original intent of providers may have been to provide guideline-compliant care, some patients receive sequential chemoradiation if unable to tolerate concurrent therapy and others do not receive the intended dose of radiation if they suffer adverse events.

Figure 1.

Patient selection scheme

Statistical Design

Background characteristics between the two groups were compared using the Wilcoxon rank sum and Pearson’s chi-squared tests for continuous and categorical variables, respectively. A multivariable logistic regression was performed to examine variables independently associated with use of guideline-concordant neoadjuvant chemoradiotherapy. Variables for the logistic regression were selected a priori based on prior literature and availability in the NCDB: age, sex, race, insurance status, Charlson-Deyo comorbidity index (CDCC) score, urban or rural treatment location, academic center treatment, education level by zip code, median income by zip code, geographic location of treatment center, center volume for esophagectomy by quartile, tumor size, and clinical stage. Center volume was defined as a categorical variable based on quartiles from the distribution of annualized volume for esophagectomy during the study period in the NCDB.

The effect of use of NCR on survival was then evaluated using the Kaplan-Meier method and a multivariable Cox Proportional Hazards regression using the same variables as in the logistic regression. Variables for the Cox model were again selected a priori based on availability in the NCDB and known prognostic association from prior studies in esophageal cancer patients. Because of the size of the cohort, parsimony was not a concern and variable selection techniques were not applied. This model was evaluated by concordance, likelihood ratio, and degrees of freedom. To mitigate the immortal time bias experienced by patients in the NCR group (Figure 3), we further performed a multivariable Cox regression examining survival from surgery rather than from diagnosis.

Figure 3.

Stacked bar graph showing the proportion of patients treated with neoadjuvant chemoradiation by year of diagnosis

Because of the heterogeneity of the non-NCR group and the lack of knowledge about intended therapies in this group, we performed a subgroup analysis including only patients who received NCR and those who received no perioperative therapies. A multivariable logistic regression to identify factors associated with receipt of NCR and a multivariable Cox regression were both performed in this subgroup employing identical variables as in the main models. Missing data were handled with complete case analysis given the high degree of completeness in the NCDB. A summary of baseline characteristics for patients excluded from the study because they were catalogued as medically unfit for chemotherapy or radiation is provided in Supplemental Table 1; this cohort’s five-year OS was 24% (95%CI 13–42). A p value less than or equal to 0.05 was considered statistically significant. All statistical analyses were performed using R version 3.5.1 for Mac (Vienna, Austria).

Results

Overall, 8076 patients met study criteria: 1616 (20%) received non-NCR therapy and 6460 (80%) NCR. The background characteristics of patients in each group are summarized in Table 1. Therapies received by non-NCR patients are summarized in Figure 2. Patients who received induction chemoradiation were more likely to be diagnosed at a later year, be privately insured, be treated in New England, South Atlantic, East North Central, or West North Central states, be treated at highest volume surgery centers, and have stage III disease compared to patients not receiving NCR (Supplemental Table 2). Patients not receiving NCR underwent a variety of neoadjuvant and adjuvant therapies, as demonstrated in Table 1. The distribution of patients, by year of diagnosis, is demonstrated in Figure 3. In a multivariable logistic regression, factors independently associated with NCR included a later year of diagnosis, treatment in a high-volume (Q4) esophagectomy center, and clinical stage III disease (Table 2). Factors associated with non-receipt of NCR included increasing age, a CDCC score equal to or greater than two, and treatment in a Middle Atlantic, South Central, or Pacific state. In a subgroup of patients who either received NCR or no perioperative therapy, factors associated with non-use of NCR included increasing age, treatment at an academic center, and treatment in centers outside New England, while factors associated with use of NCR included later year of diagnosis, comorbidities, and clinical stage III disease (Supplemental Table 3).

Table 1.

Background characteristics of study patients stratified receipt of induction chemoradiation

Variable	No Neoadjuvant Chemoradiation (n=1616) (%)	Neoadjuvant Chemoradiation (n=6460) (%)	p-value
Age (median [IQR])	63(56–71)	62(55–68)	<0.001
Sex (female)	185(11)	703(11)	0.55
Race (Caucasian)	1558(97)	6245(98)	0.83
Year of diagnosis	2009(2006–2011)	2011(2008–2013)
CDCC Score			0.03
0	1170(72)	4831(75)
1	349(22)	1337(21)
≥2	97(6)	292(5)
Insurance			<0.001
Private	767(48)	3431(54)
Government	802(50)	2762(44)
None	27(2)	125(2)
Treatment location			0.48
Metro	1226(80)	4929(79)
Urban	268(17)	1136(18)
Rural	43(3)	147(2)
Academic center	864(55)	3589(57)	0.17
Education quartile			0.07
Q1 (≥21% without high school diploma)	191(12)	655(10)
Q2 (13–21%)	383(24)	1614(25)
Q3 (7–12.9%)	566(36)	2426(38)
Q4 (<7%)	437(28)	1676(26)
Income quartile (median)			0.24
Q1 (<$38,000)	213(14)	835(13)
Q2 (<$48,000)	382(24)	1562(25)
Q3 (<$63,000)	421(27)	1843(29)
Q4 (≥$63,000)	561(36)	2128(33)
Facility location			<0.001
New England (CT,MA,ME,NH,RI,VT)	89(6)	520(8)
Middle Atlantic (NJ,NY,PA)	373(24)	1039(16)
South Atlantic (DC,DE,FL,GA,MD,NC,SC,VA,WV)	272(17)	1255(20)
East North Central (IL,IN,MI,OH,WI)	267(17)	1355(21)
East South Central (AL,KY,MS,TN)	82(5)	313(5)
West North Central (IA,KS,MN,MO,ND,NE,SD)	159(10)	795(13)
West South Central (AR,LA,OK,TX)	116(7)	265(4)
Mountain (AZ,CO,ID,MT,NM,NV,UT,WY)	50(3)	249(4)
Pacific (AK,CA,HI,OR,WA)	175(11)	557(9)
Center volume (esophagectomies per year)			<0.001
Q1 (0–1 a year)	187(12)	581(9)
Q2 (>1–3)	327(20)	1367(21)
Q3 (>3–8)	411(25)	1439(22)
Q4 (>8)	691(43)	3073(48)
Clinical stage			<0.001
II	348(22)	1067(17)
III	1178(73)	5068(79)
IVA	90(6)	325(5)
Number of radiation doses (IQR)	28(25–30)	28(25–28)	N/A
Radiation dose (median Gy)(IQR)	46(45–50)	45(45–50)	N/A
Radiation dose (range, Gy)	7–57	11–56	N/A

Figure 2.

Pie chart summarizing therapies received by non-neoadjuvant chemotherapy patients

Table 2.

Multivariable logistic regression of independent predictors of receipt of induction chemoradiation for patients with clinically node-positive esophageal adenocarcinoma

Variable	Odds Ratio	95% Confidence Interval	p-value
Age (per year)	0.98	0.97–0.99	<0.001
Sex (reference: male)	0.90	0.73–1.12	0.34
Race (reference: Caucasian)
Black	0.84	0.48–1.48	0.55
Year of diagnosis (per year)	1.19	1.16–1.22	<0.001
CDCC Score (reference: 0)
1	0.93	0.79–1.10	0.39
≥2	0.60	0.45–0.81	<0.001
Insurance (reference: private)
Government	0.87	0.74–1.02	0.09
None	1.20	0.69–2.09	0.52
Treatment location (reference: metro)
Urban	0.88	0.73–1.07	0.20
Rural	0.74	0.48–1.14	0.18
Academic center	0.87	0.72–1.05	0.13
Education quartile (reference: Q1)
Q2 (13–21%)	1.04	0.81–1.34	0.75
Q3 (7–12.9%)	1.20	0.92–1.57	0.17
Q4 (<7%)	1.04	0.77–1.42	0.78
Income quartile (reference: Q1)
Q2 (<$48,000)	0.94	0.74–1.20	0.63
Q3 (<$63,000)	1.07	0.83–1.38	0.62
Q4 (≥$63,000)	1.03	0.77–1.37	0.86
Facility location (reference: New England)
Middle Atlantic	0.43	0.31–0.59	<0.001
South Atlantic	0.79	0.56–1.11	0.17
East North Central	0.77	0.56–1.08	0.13
East South Central	0.65	0.42–0.99	0.04
West North Central	0.84	0.59–1.21	0.35
West South Central	0.31	0.20–0.46	<0.001
Mountain	0.98	0.61–1.59	0.93
Pacific	0.59	0.41–0.84	0.003
Center volume (reference: Q1 0–1/year)
Q2 (>1–3)	1.41	1.09–1.82	0.008
Q3 (>3–8)	1.06	0.82–1.37	0.67
Q4 (>8)	1.69	1.27–2.25	<0.001
Tumor size (per mm)	1.00	1.00–1.00	0.14
Clinical stage (reference: II)
III	1.44	1.22–1.71	<0.001
IVA	1.08	0.79–1.49	0.63

Patients not receiving NCR experienced a median survival of 25 months (95% confidence interval [CI] 24–27) and a five-year survival of 26% (95%CI 24–29) while patients receiving NCR had a median and five-year survival of 33 months (95%CI 32–35) and 35% (95%CI 34–36), respectively (Figure 4). In multivariable analysis, receipt of NCR was associated with an overall survival benefit (HR 0.80; 95%CI 0.74–0.87) compared to non-NCR (Table 3). When survival was computed from surgery rather than diagnosis, induction chemoradiation was still associated with a survival benefit (HR 0.85; 95%CI 0.79–0.93; p<0.001) compared to not receiving it. In a subgroup analysis comparing NCR patients with those who received no perioperative therapies, NCR was still associated with a survival benefit following surgery (HR 0.70; 95%CI 0.61–0.81; p<0.001).

Figure 4.

Kaplan-Meier survival curves for patients stratified by receipt of neoadjuvant chemoradiation

Table 3.

Multivariable Cox Proportional Hazards regression of independent predictors of mortality for patients with clinically node-positive esophageal adenocarcinoma

Variable	Hazard Ratio	95% Confidence Interval	p-value
Age (per year)	1.01	1.01–1.02	<0.001
Sex (reference: male)	0.85	0.75–0.95	0.004
Race (reference: Caucasian)
Black	1.03	0.76–1.39	0.85
Year of diagnosis (per year)	0.98	0.97–0.99	<0.001
CDCC Score (reference: 0)
1	1.13	1.04–1.22	0.005
≥2	1.26	1.08–1.47	0.003
Insurance (reference: private)
Government	1.13	1.04–1.23	0.004
None	1.21	0.95–1.56	0.13
Treatment location (reference: metro)
Urban	1.06	0.96–1.16	0.26
Rural	1.00	0.80–1.25	0.98
Academic center	0.97	0.88–1.07	0.53
Education quartile (reference: Q1)
Q2 (13–21%)	1.10	0.96–1.25	0.17
Q3 (7–12.9%)	1.08	0.94–1.24	0.27
Q4 (<7%)	1.00	0.86–1.17	0.96
Income quartile (reference: Q1)
Q2 (<$48,000)	0.93	0.83–1.05	0.26
Q3 (<$63,000)	0.96	0.85–1.09	0.52
Q4 (≥$63,000)	0.84	0.73–0.97	0.02
Facility location (reference: New England)
Middle Atlantic	0.93	0.79–1.09	0.37
South Atlantic	1.05	0.90–1.23	0.55
East North Central	1.08	0.93–1.26	0.33
East South Central	1.20	0.98–1.47	0.08
West North Central	0.97	0.82–1.15	0.77
West South Central	1.04	0.84–1.29	0.72
Mountain	1.00	0.80–1.24	0.98
Pacific	0.88	0.74–1.05	0.16
Center volume (reference: Q1 0–1/year)
Q2 (>1–3)	0.87	0.76–0.99	0.03
Q3 (>3–8)	0.89	0.78–1.01	0.07
Q4 (>8)	0.76	0.66–0.88	<0.001
Tumor size (per mm)	1.00	1.00–1.00	<0.001
Clinical stage (reference: II)
III	1.31	1.19–1.44	<0.001
IVA	1.79	1.52–2.10	<0.001
Standard of care therapy (reference: No neoadjuvant chemoradiation)
Neoadjuvant chemoradiation	0.80	0.74–0.87	<0.001

Discussion

We used the NCDB to identify factors associated with use of induction chemoradiotherapy in locally advanced esophageal adenocarcinoma. We found that several patient, treatment center, and tumor related factors were associated with receipt NCR, including age, comorbidities, clinical stage, treatment at a high-volume center, and geographic location of treatment center. Patients receiving trimodality therapy experienced improved overall survival compared to those not receiving it, suggesting that these differences in treatment affect survival in patients with esophageal cancer.

This is the first study, to our knowledge, that examines differences associated with receipt of neoadjuvant chemoradiation in locally advanced esophageal cancer, but the variations we identified are largely corroborated by existing literature in esophageal and other cancers. We found, for instance, that older patients were less likely to undergo trimodality therapy. This is consistent with Surveillance, Epidemiology, and End Results (SEER) registry analyses showing that older patients with locoregional esophageal cancer were less likely to be seen by a surgeon or medical oncologist, and were in turn less likely to receive any treatment, with only about 6% receiving trimodality therapy^18,19. We also found that patients treated in high volume esophagectomy centers were more likely to be treated with trimodality therapy, which corroborates findings from many studies showing that patients treated at high volume centers experience improved outcomes^20–23. Our study suggests that patients treated at high volume centers may have a survival benefit from not only better perioperative care but also increased use of NCR. Unlike other studies^{12–14,24–28}, we also did not find that race affected receipt of NCR, though less than 3% of patients in our cohort were non-white. The existing literature suggests, however, that racial disparities may affect patients with esophageal cancer at different levels; for instance, non-white patients are less likely to be offered surgery at all for esophageal cancer, let alone be eligible for induction therapy^{12,13,24,27,28}. Our study also did not find a difference in economic status, unlike other studies reporting that esophageal cancer patients with a lower socioeconomic status were less likely to receive surgery, obtain a second opinion, and were more likely to lose their jobs after diagnosis²⁹. However, our study was limited by the NCDB’s reporting of education and income at the level of the zip code rather than the census tract or individual patient. We did find that the probability of neoadjuvant chemoradiation increased with time (Figure 2), likely representing a reaction to the CROSS trial published in 2012¹.

Our study has limitations in addition to those mentioned above. As a retrospective cohort analysis, it is limited by selection bias that cannot be entirely adjusted for. For instance, we do not know the reasons that some patients were offered trimodality therapy while others received no perioperative therapy at all. It is possible that patients who were generally more healthy in ways not measured by the NCDB were offered neoadjuvant therapy and in turn experienced improved survival compared to those who received other therapies. For instance, the NCDB does not contain information about the frailty and nutritional status of patients, which may have influenced their receipt of induction therapy. We attempted to adjust for this limitation by excluding from our cohort all patients who were catalogued as medically unfit for chemotherapy or radiation. The NCDB does not contain information on how patients were staged before surgery. It also does not catalogue information about the chemotherapeutic regimens offered to patients. Another significant limitation of this study is that the cohort identified was overwhelmingly Caucasian, which decreased our ability to detect differences in care in non-Caucasian patients. In the context of this study’s primary objective, another major limitation of the NCDB is that it does not contain granular information about patient and provider preferences and beliefs that may contribute to the differences observed, including quality of life metrics. Finally, while this study specifically compared neoadjuvant chemoradiation to other regimens based on the preferred regimen recommended by the NCCN, other perioperative therapies like the FLOT regimen have been adopted with good outcomes^30–32. Therefore, the findings described in our study may not necessarily represent disparities in care but variations in clinical practice.

In this NCDB analysis, we found that 80% of patients with locally advanced esophageal adenocarcinoma received trimodality therapy, and it was associated with improved survival compared to non-receipt of neoadjuvant chemoradiation. We found that numerous patient, treatment center, and tumor related factors were associated with receipt of NCR. Further studies are necessary to establish the reasons for these differences and design interventions to mitigate them in an effort to improve the care and outcomes of patients with locally advanced esophageal cancer.