Margin Positivity in Resectable Esophageal Cancer: Are there Modifiable Risk Factors?

Abstract

BACKGROUND:

Patients with esophageal cancer have poor overall survival, with positive resection margins worsening survival. Margin positivity rates are used as quality measures in other malignancies, but modifiable risk factors are necessary to develop actionable targets for improvement. Our objectives were to (1) evaluate trends in esophageal cancer margin positivity and (2) identify modifiable patient/hospital factors associated with margin positivity.

METHODS:

Patients who underwent esophagectomy from 2004–2015 were identified from the National Cancer Database. Trends in margin positivity by time and hospital volume were evaluated by Cochrane-Armitage tests. Associations between patient/hospital factors and margin positivity were assessed by multivariable logistic regression.

RESULTS:

Among 29,706 patients who underwent esophagectomy for cancer, 9.37% had positive margins. Margin positivity rates decreased over time (10.62% in 2004 to 8.61% in 2015; p<0.001). Older patients were more likely to have positive margins (≥75 OR 2.04 [95% CI 1.42–2.92]), as were patients with Charlson-Deyo Index ≥3 (OR 1.84 [1.08–3.12]). Patients who received neoadjuvant therapy were less likely to have positive margins (OR 0.37 [95% CI 0.290.47]), while laparoscopic surgical approach was associated with increased margin positivity (OR 1.70 [95% CI 1.40–2.06]). As hospital annual esophagectomy volume increased, margin positivity rates decreased (7.76% in 4^th quartile vs. 11.39% in 1^st; OR 0.70 [95% CI 0.49–0.99]).

CONCLUSIONS:

Use of neoadjuvant therapy, surgical approach, and hospital volume are modifiable risk factors for margin positivity in esophageal cancer. These factors should be considered in treatment planning, and margin positivity rates could be considered as a quality measure in esophageal cancer.

Introduction:

Esophageal cancer is among the top 10 leading causes of cancer related mortality worldwide, with an overall 5-year survival rate of less than 20%.¹ Recommended treatment algorithms include surgical resection for patients with locally advanced or early stage cancer, after which survival rates can reach roughly 50%.^2,3 However, incomplete surgical resection with positive resection margins portends a worse survival than that which can be achieved with complete surgical resection.^4–7

Margin negativity rates have been indicated as quality measures in other malignancies based on the same underlying principles. In rectal cancer, for example, total mesorectal excision (TME) with negative margins has been associated with decreased local recurrence and improved survival.⁸ The Optimizing Surgical Treatment of Rectal Cancer (OSTRiCh) Clinical Consortium developed a set of rectal cancer quality measures to improve the quality of rectal cancer care in the United States.⁹ After noting frequent margin positivity in rectal cancer patients, they advocated complete TME with standardized pathologic evaluation of margin positivity.^9,10 Negative resection margins following gastrectomy for malignancy are also used as quality measures in Europe, and have been advocated by some in the United States.^11,12

In esophageal cancer, R0 resection has been proposed as a quality measure based on data that both early and advanced stage patients who undergo esophagectomy have improved survival after R0 resection compared with those with R1 resections.¹³ However, clearly identified modifiable risk factors for margin positivity must be delineated before broad use of resection status as a quality measure can be advocated. Thus, the objectives of this study are to (1) evaluate for trends in margin positivity following surgical resection for esophageal cancer and (2) identify modifiable patient or hospital factors associated with margin positivity.

Methods:

Data Source and Study Population

The National Cancer Database (NCDB) is a national cancer data registry sponsored by the American College of Surgeons Commission on Cancer (CoC) and the American Cancer Society.¹⁴ The NCDB began abstracting clinical information on patients with malignancies in 1988, and since 1998 all CoC accredited hospitals have been required to submit all of their cancer cases.¹⁵ Currently, data are collected from over 1,500 hospitals, capturing 72% of annual cancer cases in the United States.¹⁶ Data are electronically evaluated for completeness and duplication as well as regularly audited for quality assurance and reliability.¹⁷

Of the 140,234 patients in the esophageal cancer participant use file (PUF) from 2004–2015, 42,925 underwent esophagectomy. Patient were excluded if they had stage IV disease (n=1,541), had greater than 1 primary malignancy (n=8,461), or had non-endothelial histology (n=150). All patients without documented margin status were excluded (n=3,067). Thus, the final analytic cohort included 29,706 patients with primary endothelial esophageal cancer in this retrospective cohort study. Patients with primary gastric cancer extending into the esophagus were not included in this study.

Primary Outcome and Predictors

The primary outcome was surgical margin status. In the NCDB, margin status is documented based on national cancer registry standards utilizing the final pathology report. Patients were identified as having negative (R0), microscopically involved (R1), macroscopically involved (R2), or positive but not otherwise classified (R+ NOS) margin status. In the event that several margins were positive, the more extensive margin was reported. Final margin status was dichotomized into any margin positive (R1, R2, or R+ NOS) or margin negative (R0) for analysis.

Time and volume trends in margin positivity were evaluated using the Cochrane-Armitage test for trend. Mean annual hospital esophagectomy volume was calculated by tabulating the number of registered cases by facility ID divided by the number of years in which that institution reported cases to the NCDB.

Possible predictors of margin positivity were identified a priori based on a literature review and clinical face validity, including demographic, socioeconomic, tumor characteristics, and management factors. Variables included age, sex, race (White, Black, Hispanic, Other), insurance status (Medicaid/Medicare, Private, Other Government, Uninsured), income, Charlson-Deyo Index, neoadjuvant therapy, surgical approach, facility type, hospital location, distance to hospital, clinical stage, grade, histology, tumor location, tumor size, and hospital annual esophagectomy volume.

Income was reported as median ZIP code estimates based on census data, and subsequently divided in quartiles. Facility type was categorized based on CoC definitions of Academic/Research Program, Integrated Network Cancer Program, Comprehensive Community Cancer Program, and Community Cancer Program. Hospital location was determined by United States Department of Agriculture Economic Research Service rural-urban continuum codes and categorized into metro, urban, and rural locations. Distance to treating hospital was calculated by aerial distance between the center of the patient’s ZIP code and street address of the reporting hospital facility by the NCDB and subsequently divided into quartiles. Histologic subtypes were identified based on ICD-0–3 morphology codes and categorized as adenocarcinoma, squamous cell carcinoma, or carcinoma NOS. Clinical T and N stage was clinically determined based on American Joint Committee on Cancer (AJCC) definitions used in the year in which the case was diagnosed. Clinical stage was used rather than pathologic stage such that only pre-operatively available risk factors for margin positivity would be utilized for analysis. Tumor grade was based on histologic diagnosis and categorized into well differentiated, moderately differentiated, poorly differentiated, and undifferentiated. Tumor size was measured in centimeters and categorized into relatively equal sized groups. Hospital volume was separated into quartiles (≤2, 2.1–5.4, 5.5–13.3, >13.3 mean cases per year) for analysis.

Statistical Analysis:

Bivariate analyses of association between individual variables and positive resection margins were assessed by chi-squared tests. Statistically significant predictors with a predetermined P<0.05 were entered into a multivariable logistic regression model to estimate the association of patient and hospital specific factors and margin positivity. Model diagnostics included evaluation for both discrimination and calibration. The C-Statistic was calculated for discrimination, with a value closer to 1.0 indicating better discrimination, while the Hosmer and Lemeshow chi-square was calculated for calibration, with any P<0.05 indicating poor calibration.^18,19 Sensitivity analyses were performed to evaluate predictors of margin positivity based on early versus late T stage as well as predictors of R2 margin involvement.

Hospital volume thresholds were evaluated using restricted cubic splines on a regression model with knots placed at integers of 1, 5, 10, 15, and 20 cases per year. The restricted cubic splines technique allows for the evaluation of a non-linear continuous predictor of an outcome.²⁰ Integers were chosen based on clinical interpretability, and the number and position of knots were evaluated for model fit to optimize Akaike Information Criteria (AIC), as previously described.²¹

A two-sided pre-determined level of significance of 0.05 was used for all analyses. All statistical analyses were performed using Stata version 14.2 (Stata Corporation, College Station, TX). This study utilized deidentified patient data and was determined to be exempt from review by the Northwestern University Institution Review Board.

Results:

Of 29,706 analyzed patients, the mean age was 62.9 years, 83.12% were male, and 89.99% white (Table 1). The majority of patients were insured, with roughly equal numbers of patients having Medicaid/Medicare (48.74%) and private insurance (47.45%). 54.55% of patients were treated at academic hospitals, followed by 30.27% at comprehensive community cancer programs. Overall, 54.43% of patients received neoadjuvant therapy, with 66.52% of patients with locally advanced tumors receiving neoadjuvant therapy. Of the patients with a documented surgical approach, 57.96% underwent an open resection. Hospitals in the lowest volume quartile performed ≤2 esophagectomies per year on average, while those in the highest quartile performed >13.3. Additional demographic information can be found in Table 1.

Table 1:

Characteristics of Patients Undergoing Esophagectomy for Malignancy

	Total
	N (%)
Total Cases	29,706
Age
<55	6,100 (20.53%)
55–64	10,286 (34.63%)
65–74	9,534 (32.09%)
≥75	3,786(12.74%)
Sex
Male	24,691 (83.12%)
Female	5,015(16.88%)
Race
White	26,731 (89.99%)
Black	1,237 (4.16%)
Hispanic	809 (2.72%)
Other	929(3.13%)
Insurance Status
Medicaid/Medicare	14,238 (48.74%)
Private	13,862 (47.45%)
Other Government	489(1.67%)
Uninsured	622 (2.13%)
Charlson-Deyo Index
0	21,734 (73.16%)
1	6,274 (21.12%)
2	1,315 (4.43%)
≥3	383(1.29%)
Median Zip Code Income
<38,000	4,403(15.05%)
38,000–48,000	7,218(24.67%)
48,000–63,000	8,349 (28.53%)
>63,000	9,290 (31.75%)
Quartile–of Distance to Hospital (miles)
1 (<1–7.3)	7,419(25.34%)
2(7.4–18.5)	7,256 (24.78%)
3(18.6–49.8)	7,283 (24.88%)
4 (≥49.9)	7,318(25.00%)
Facility Type
Academic/Research	15,974 (54.55%)
Integrated Network Cancer Program	2,986(10.20%)
Comprehensive Community Cancer Program	8,865 (30.27%)
Community Cancer Program	1,457 (4.98%)
Hospital Location
Metro	23,042 (80.48%)
Urban	4,955(17.31%)
Rural	635 (2.22%)
Histologic Morphology
Adenocarcinoma	23,899 (80.45%)
Squamous Cell Carcinoma	4,834(16.27%)
Carcinoma NOS	974 (3.28%)
Tumor Location
Upper Third	555(1.87%)
Middle Third	3,210(10.81%)
Lower Third	22,495 (75.73%)
Overlapping Sites/NOS	3,446(11.60%)
Clinical T Stage
Tis	934 (4.16%)
T1	5,523 (24.63%)
T2	4,271 (19.04%)
T3	11,104 (49.51%)
T4	596 (2.66%)
Clinical N Stage
NO	13,352 (56.39%)
N1	8,773 (37.05%)
N2	1,314(5.55%)
N3	239(1.01%)
Tumor Grade
Well Differentiated	2,233 (9.14%)
Moderately Differentiated	10,521 (43.06%)
Poorly Differentiated	11,255 (46.06%)
Undifferentiated	424(1.74%)
Tumor Size3
≤2.0	6,762(31.94%)
2.1–4.0	6,950(32.82%)
4.1–6.0	4,512(21.31%)
>6.0	2,950(13.93%)
Neoadjuvant Therapy
No	13,537 (45.57%)
Yes	16,169 (54.43%)
Surgical Approacha
Open	7,559(57.96%)
Laparoscopic	4,622 (35.44%)
Robotic	860 (6.59%)
Mean Annual Hospital Esophagectomy Volume Quartile (n/year)
1 (≤2)	7,544 (25.40%)
2 (2.1–5.4)	7,406 (24.94%)
3(5.5–13.3)	7,369 (24.81%)
4 (>13.3)	7,380 (24.85%)

cm = centimeter

^asurgical approach available beginning in 2010

Margin Positivity Rates and Trends by Year

Overall, 9.37% of patients in this study had positive resection margins after undergoing esophagectomy. Margin positivity rates decreased with time, with 10.62% (SE 0.66%) of patients having positive resection margins in 2004 vs. 8.61% in 2015 (SE 0.53%; p<0.001). The lowest rate was in 2013 with 8.24% margin positivity (SE 0.53%) while the highest rate was in 2005 with 10.88% (SE 0.64%; Figure 1).

Figure 1:

Esophagectomy margin positivity rates by year of diagnosis

Factors Associated with Margin Positivity

Positive resection margins increased with patient age, with patients 65–74 more frequently experiencing positive margins than patients under 55 (9.05% vs. 8.56%; OR 1.38 [95% CI 1.00–1.90]). Similarly, patients ≥75 had positive resection margins even more frequently than patients under 55 (14.10% vs. 8.56%; OR 2.04 [95% CI 1.42–2.92]). Although Black patients were more likely to have positive resection margins than Whites on univariate analysis, these findings were not statistically significant after controlling for other patient and hospital factors on multivariable analysis (11.72% vs. 9.18%; OR 1.23 [95% CI 0.79–1.92]). A similar trend was found for Hispanic patients (11.50% vs. 9.18%; OR 1.25 [95% CI 0.75–2.10]). Although uninsured patients and those with Medicaid or Medicare coverage more frequently had positive resection margins compared to patients with private insurance, no statistically significant association was found after controlling for other factors (12.06% margin positivity for uninsured patients, OR 1.05 [95% CI 0.53–2.08]; 10.67% for Medicaid/Medicare, OR 1.04 [95% CI 0.82–1.31] vs. 7.86% for private insurance). Patients with a Charlson-Deyo Index ≥3 were more likely to have positive resection margins than those with an index of 0 (13.84% vs. 9.25%; OR 1.84 [95% CI 1.08–3.12]).

Patients who reside in more affluent areas were less likely to have positive resection margins (8.81% in patients with median zip code income > $63,000 vs. 10.31% with median zip code income < $38,000; OR 0.67 [95% CI 0.50–0.91]). On bivariate analysis, there was a step wise decrease in margin positivity rates as the distance to treating hospital increased, but on multivariable analysis, only the third quartile (18.6–49.8 miles to treating hospital) was associated with an increased likelihood of positive margins (9.43% for third quartile vs. 10.08% for first; OR 1.41 [95% CI 1.08–1.85]). Patients treated at academic or research institutions less frequently had positive resection margins on bivariate analysis, but no association was found on multivariable analysis (see Tables 2 and 3).

Table 2:

Frequency of Esophaaectomv Marqin Positivity bv Sample Characteristics

	Negative Margin	Positive Margin	p-value*
	N (%)
Total Cases	26,924 (90.63%)	2,782 (9.37%)
Age
<55	5,578(91.44%)	522 (8.56%)	<0.001
55–64	9,423(91.61%)	863 (8.39%)
65–74	8,671 (90.95%)	863 (9.05%)
≥75	3,252 (85.90%)	534(14.10%)
Sex
Male	22,386 (90.66%)	2,305 (9.34%)	0.696
Female	4,538(90.49%)	477(9.51%)
Race
White	24,276 (90.82%)	2,455 (9.18%)	0.004
Black	1,092(88.28%)	145(11.72%)
Hispanic	716 (88.50%)	93 (11.50%)
Other	840 (90.42%)	89 (9.58%)
Insurance Status
Medicaid/Medicare	12,719(89.33%)	1,519(10.67%)	<0.001
Private	12,772 (92.14%)	1,090 (7.86%)
Other Government	442 (90.39%)	47 (9.61%)
Uninsured	547 (87.94%)	75 (12.06%)
Charlson-Deyo Index
0	19,724 (90.75%)	2,010 (9.25%)	0.001
1	5,706 (90.95%)	568 (9.05%)
2	1,164(88.52%)	151 (11.48%)
≥3	330 (86.16%)	53 (13.84%)
Median Zip Code Income
<38,000	3,949(89.69%)	454(10.31%)	0.039
38,000–47,999	6,557(90.84%)	661 (9.16%)
48,000–62,999	7,763 (90.59%)	786(9.41%)
>63,000	8,472(91.19%)	818(8.81%)
Quartile of Distance to Hospital (miles)
1 (<1–7.3)	6,671 (89.92%)	748(10.08%)	<0.001
2(7.4–18.5)	6,564 (90.46%)	692 (9.54%)
3(18.6–49.8)	6,596(90.57%)	687 (9.43%)
4 (≥49.9)	6,723(91.87%)	595(8.13%)
Facility Type
Academic/Research	14,601 (91.40%)	1,373 (8.60%)	<0.001
Integrated Network Cancer Program	2,710(90.76%)	276 (9.24%)
Comprehensive Community Cancer Program	7,965(89.85%)	900(10.15%)
Community Cancer Program	1,275 (87.51%)	182 (12.49%)
Hospital Location
Metro	20,910(90.75%)	2,132 (9.25%)	0.444
Urban	4,475(90.31%)	480 (9.69%)
Rural	582 (91.65%)	53 (8.35%)
Histologic Morphology
Adenocarcinoma	21,741 (90.97%)	2,158 (9.03%)	<0.001
Squamous Cell Carcinoma	4,293(88.81%)	541 (11.19%)
Carcinoma NOS	890 (91.47%)	83 (8.53%)
Tumor Location
Upper Third	451 (81.26%)	104(18.74%)	<0.001
Middle Third	2,877 (89.63%)	333(10.37%)
Lower Third	20,554 (91.37%)	1,941 (8.63%)
Overlapping Sites/NOS	3,042 (88.28%)	404(11.72%)
Clinical T Stage
Tis	849 (90.90%)	85 (9.10%)	<0.001
T1	5,021 (90.91%)	502 (9.09%)
T2	3,966 (92.86%)	305 (7.14%)
T3	10,224 (92.07%)	880 (7.93%)
T4	507 (85.07%)	89 (14.93%)
Clinical N Stage
NO	12,147 (90.98%)	1,205 (9.02%)	<0.001
N1	8,084(92.15%)	689 (7.85%)
N2	1,227(93.38%)	87 (6.62%)
N3	212 (88.70%)	27 (11.30%)
Tumor Grade
Well Differentiated	2,078(93.06%)	155 (6.94%)	<0.001
Moderately Differentiated	9,647(91.69%)	874(8.31%)
Poorly Differentiated	9,929 (88.22%)	1,326(11.78%)
Undifferentiated	377 (88.92%)	47 (11.08%)
Tumor Size3
≤2.0	6,244 (92.34%)	518(7.66%)	<0.001
2.1–4.0	6,289 (90.49%)	661 (9.51%)
4.1–6.0	4,016(89.01%)	496(10.99%)
>6.0	2,545 (86.27%)	405(13.73%)
Neoadjuvant Therapy
No	11,768 (86.93%)	1,769(13.07%)	<0.001
Yes	15,156 (93.73%)	1,013 (6.27%)
Surgical Approacha
Open	7,021 (92.88%)	538 (7.12%)	<0.001
Laparoscopic	4,064 (87.93%)	558(12.07%)
Robotic	828 (96.28%)	32 (3.72%)
Mean Annual Hospital Esophagectomy Volume Quartile (n/year)
1 (≤2)	6,684(88.61%)	859(11.39%)	<0.001
2 (2.1–5.4)	6,679(90.18%)	727 (9.82%)
3(5.5–13.3)	6,746(91.55%)	623 (8.45%)
4 (>13.3)	6,807 (92.24%)	573.76%)

chi2 test

cm = centimeter

^asurgical approach available beginning in 2010

Table 3:

Association of Patient and Hospital Factors with Esophagectomy Margin Positivity

	Adjusted Odds Ratio (95% Confidence Interval)	p-value*
Age
<55	1.00	REF
55–64	1.30 (0.97–1.75)	0.078
65–74	1.38(1.00–1.90)	0.048
≥75	2.04(1.42–2.92)	<0.001
Race
White	1.00	REF
Black	1.23 (0.79–1.92)	0.360
Hispanic	1.25 (0.75–2.10)	0.389
Other	1.07 (0.62–1.86)	0.805
Insurance Status
Medicaid/Medicare	1.04 (0.82–1.31)	0.750
Private	1.00	REF
Other Government	0.65 (0.30–1.44)	0.289
Uninsured	1.05 (0.53–2.08)	0.887
Charlson-Deyo Index
0	1.00	REF
1	0.90 (0.73–1.12)	0.367
2	1.22 (0.84–1.78)	0.293
>3	1.84(1.08–3.12)	0.024
Median Zip Code Income
<38,000	1.00	REF
38,000–48,000	0.96 (0.72–1.28)	0.785
48,000–63,000	0.95 (0.72–1.25)	0.695
>63,000	0.67 (0.50–0.91)	0.009
Quartile of Distance to Hospital (miles)
1 (<1–7.3)	1.00	REF
2(7.4–18.5)	1.24 (0.94–1.63)	0.130
3(18.6–49.8)	1.41 (1.08–1.85)	0.013
4 (>49.9)	1.12(0.84–1.50)	0.424
Facility Type
Academic/Research	1.00	REF
Integrated Network Cancer Program	1.32 (0.97–1.79)	0.080
Comprehensive Community Cancer Program	0.98 (0.74–1.28)	0.864
Community Cancer Program	0.97 (0.48–1.95)	0.927
Histologic Morphology
Adenocarcinoma	1.00	REF
Squamous Cell Carcinoma	0.99 (0.75–1.31)	0.964
Carcinoma NOS	0.65 (0.25–1.67)	0.373
Tumor Location
Upper Third	2.07(1.16–3.68)	0.014
Middle Third	1.13(0.83–1.54)	0.444
Lower Third	1.00	REF
Overlapping Sites/NOS	1.40(1.05–1.86)	0.023
Clinical T Stage
Tis	1.02 (0.50–2.11)	0.948
T1	1.00	REF
T2	1.00 (0.74–1.35)	0.993
T3	1.70(1.24–2.33)	0.001
T4	3.42(1.89–6.17)	<0.001
Clinical N Stage
NO	1.00	REF
N1	0.96 (0.76–1.22)	0.758
N2	0.73 (0.50–1.07)	0.107
N3	1.27 (0.71–2.28)	0.422
Tumor Grade
Well Differentiated	1.00	REF
Moderately Differentiated	1.59(1.11–2.28)	0.012
Poorly Differentiated	2.27(1.57–3.27)	<0.001
Undifferentiated	1.83 (0.84–4.00)	0.130
Tumor SizeJ
<2.0	1.00	REF
2.1–4.0	0.98 (0.77–1.25)	0.881
4.1–6.0	1.01 (0.76–1.35)	0.932
>6.0	1.50(1.10–2.04)	0.010
Neoadjuvant Therapy
No	1.00	REF
Yes	0.37 (0.29–0.47)	<0.001
Surgical Approach
Open	1.00	REF
Laparoscopic	1.70(1.40–2.06)	<0.001
Robotic	0.68 (0.44–1.06)	0.086
Mean Annual Hospital Esophagectomy Volume Quartile (n/year)
1 (<2)	1.00	REF
2 (2.1–5.4)	0.90 (0.67–1.20)	0.461
3(5.5–13.3)	0.72 (0.52–1.00)	0.049
4 (>13.3)	0.70 (0.49–0.99)	0.047

1 cm = centimeter

Several tumor characteristics are associated with margin positivity and were controlled for in our multivariable model. Clinical T3 and T4 tumors are associated with increased likelihood of positive margins (T3 OR 1.70 [95% CI 1.24–2.33], T4 OR 3.42 [95% CI 1.89–6.17], vs. T1). After controlling for other variables, clinical N stage was not associated with margin positivity, nor was histologic morphology. However, moderately and poorly differentiated tumors were associated with margin positivity (moderately differentiated OR 1.59 [95% CI 1.11–2.28], poorly differentiated OR 2.27 [95% CI 1.57–3.27], vs. well differentiated). Tumors located in the upper third of the esophagus and those classified as locating over overlapping sites were also more likely to have margin positivity than tumors in the distal third of the esophagus (upper third OR 2.07 [95% CI 1.16–3.68]; overlapping sites OR 1.40 [95% CI 1.05–1.86]). Tumors larger than 6.0 cm in size were also more likely to have positive margins (OR 1.50 [95% CI 1.10–2.04]).

Modifiable Factors Associated with Margin Positivity

Several directly modifiable risk factors were found to be associated with margin positivity in esophageal cancer. Neoadjuvant therapy was associated with a decreased likelihood of having positive resection margins (6.27% vs. 13.07%; OR 0.37 [95% CI 0.29–0.47]). Patients who underwent a laparoscopic surgical resection were more likely to have positive margins than those who underwent open surgery (12.07% of laparoscopic cases vs. 7.12% of open; OR 1.70 [95% CI 1.40–2.05]), while there was no statistically significant difference between margin positivity rates following robotic esophagectomy as compared to open (3.72% of robotic cases; OR 0.68 [95% CI 0.44–1.06]). Finally, as mean hospital volume increased, margin positivity rates decreased in a step-wise manner (7.76% margin positivity in the fourth quartile vs. 11.39% in first; OR 0.70 [95% CI 0.49–0.99] and 8.45% margin positivity in the third quartile; OR 0.72 [95% CI 0.52–1.00]).

Evaluating for trends in margin positivity rates by mean hospital esophagectomy volume reveals the highest margin positivity rates of 12.70% (SE 0.61%) at hospitals that perform a mean of one or fewer esophagectomies for cancer per year while the rate is 7.75% (SE 0.38%) at hospitals that perform on average 20 or more cases per year (p<0.001; Figure 2). Based on a restricted cubic spline analysis, risk of margin positivity levels off when hospitals perform 10 or more cases for malignancy per year (p=0.025).

Figure 2:

Esophagectomy margin positivity rates by hospital mean annual esophagectomy volume

In evaluating the multivariable logistic regression model, good discrimination (C-Statistic 0.682) and calibration (Hosmer-Lemeshow chi-square p=0.471) was identified. Qualitatively similar results were found when predicting R2 margin positivity and when T4 patients were excluded from analysis.

Discussion:

Positive resection margins are associated with worse overall survival in patients with various malignancies, including esophageal cancer. Although several other studies have suggested that positive resection margins should be a quality measure in esophageal cancer, modifiable risk factors for margin positivity must be identified for facilities to focus improvement efforts. In this study, 9.37% of patients undergoing esophagectomy for cancer had positive resection margins. Although margin positivity rates decreased over time, both unmodifiable and modifiable risk factors were found to be associated with margin positivity, including older age, patient comorbidities, and tumor specific factors (clinical T stage, tumor location, tumor grade and tumor size). Modifiable risk factors included use of neoadjuvant therapy which was associated with a decreased risk of margin positivity, laparoscopic approach which was associated with an increased risk of margin positivity, and hospital esophagectomy volume, wherein higher hospital volume quartiles were associated with lower margin positivity rates. Restricted cubic splines regression analysis revealed that the risk of margin positivity plateaued when hospitals performed at least 10 esophagectomies for cancer per year. Taken together, these findings indicate that it would be reasonable to utilize margin positivity rates as a quality measure in esophageal cancer management.

Margin Positivity Rates and Trends by Year

The margin positivity rate in this study was 9.37%, which is concordant with prior studies. In a study of the association between overall survival and adherence to quality measures in esophageal cancer using the NCDB, Samson et al. report an 8.2% margin positivity rate in Stage IIB-IIIB patients.¹³ Several other studies have reported much higher rates of margin positivity: up to 31% in patients who do not undergo neoadjuvant therapy in a Dutch study and 21.6% in a French study that included some patients who underwent neoadjuvant therapy.^22,23 The use of neoadjuvant therapy for locally advanced esophageal cancer has increased over time, particularly with the progressive publication of the Walsh trial, CALGB 9781, and the Chemotherapy for Oesophageal Cancer Followed by Surgery Study.^22,24,25 It is plausible that increased neoadjuvant therapy usage is a primary driver of the lower overall margin positivity rates seen in this study, as well as the trend toward lower margin positivity rates over time.

Factors Associated with Margin Positivity

Previous studies have identified socioeconomic disparities in the incidence, treatment, and survival of esophageal cancer, including an increased time between symptom onset and treatment initiation in patients of low socioeconomic status and a decreased likelihood to undergo surgical resection.^26–29 On bivariate analysis in this study, racial minorities and uninsured or government insured patients were more likely to have positive resection margins after undoing esophagectomy. However, these findings did not hold after controlling for other factors on multivariable logistic regression, after which only increased median zip code income was associated with decreased margin positivity rates. This lack of statically significant association between race or insurance status and margin positivity could be due to controlling for advanced tumor characteristics, which is a strong contributor to margin positivity, and more common within these groups.

Advanced patient age is associated with positive margins in esophageal cancer, with patients 75 years of age or older being twice as likely as patients less than 55 to have positive margins. Prior studies have indicated that older patients are more likely to have positive margins after undergoing resection with a curative intent for a variety of malignancies.^30,31 These age-based differences could be due to surgeons treating elderly patients less aggressively in an attempt to spare them operative morbidity. Additionally, patients in this study with more medical comorbidities (CDI ≥3) had an 84% increased odds of margin positivity than patients with CDI of 0. A previous single institution study of 658 patients with esophageal cancer found no statistically significant difference in margin positivity rates in patients with and without comorbid conditions.³² However, that study may not have been adequately powered to detect such a difference. It is reasonable to hypothesize that the same logic by which surgeons are less apt to perform radical resections on elderly patients would apply to those with multiple comorbidities, which could be contributing to the elevated positive margin rate. Additionally, older and more comorbid patients are less likely to receive neoadjuvant therapy, thereby increasing the likelihood of margin positivity.³³

It has previously been shown that more aggressive tumor characteristics, such as advanced T stage, N stage, grade, or size, are associated with increased rates of positive resection margins.^4,34,35 This study supports these findings with T3, T4, moderately and poorly differentiated tumors, proximal tumors, and tumors greater than 6.0 cm in size being associated with margin positivity. Although these tumor characteristics are not directly modifiable, we controlled for these tumor specific variables in our multivariable logistic regression model. Thus, we have accounted for all known predictors associated with margin positivity in identifying modifiable risk factors for positive margins.

Modifiable Factors Associated with Margin Positivity

This study identifies three primarily modifiable risk factors for margin positivity in patients undergoing resection of esophageal cancer: use of neoadjuvant therapy, surgical approach, and hospital volume. Neoadjuvant therapy was first advocated in esophageal cancer after Walsh et al. reported pathologic response, decreased nodal involvement, and improved survival with neoadjuvant treatment.²⁴ This evidence base grew when the Medical Research Council Oesophageal Cancer Working Group and the CROSS trial both revealed improvements in curative resection rates and overall survival following neoadjuvant therapy.^22,36,37 Our finding that neoadjuvant therapy is associated with a 63% reduced odds of margin positivity is concordant with these randomized trials. Current NCCN guidelines recommend neoadjuvant therapy for advanced stage esophageal cancer, defined as T3 or greater or node positive tumors.³ A sensitivity analysis limited to advanced stage patients by this definition yielded similar results, with neoadjuvant therapy associated with decreased odds of margin positivity. However, based on the finding that neoadjuvant therapy is associated with decreased odds of margin positivity, even after controlling for clinical stage, providers could consider neoadjuvant therapy for early stage esophageal tumors if margin positivity is a concern based upon tumor location or other factors.

Laparoscopic surgery was associated with a 70% increased odds of margin positivity, while no statistically significant association was found with robotic surgery and margin positivity. Minimally invasive esophagectomy has been performed since the early 2000s, and has been found to be feasible for esophageal cancer resection, with similar oncologic outcomes compared with open resection.^38–40 Several benefits of laparoscopic esophagectomy have been reported, most notably a reduction in pulmonary complications compared with open surgery.⁴¹ However, this is the first study, to our knowledge, to reveal an association between laparoscopic surgery and positive resection margins in esophageal surgery. A potential explanation of these findings could be related to surgeon experience and operator learning curve in esophagectomy, which is among the most complex laparoscopic operations performed.⁴² Recent studies have revealed that complication rates have decreased over time with adoption of laparoscopic surgery for complex operations.^43,44 Similarly, increased surgeon experience may reduce the likelihood of margin positivity, which may continue to improve over time.

Finally, increasing quartile of annual esophagectomy volume at the hospital level was associated with lower margin positivity rates. A prior NCDB analysis from our institution identified decreased use of surgery and multimodal therapy at low volume hospitals for esophageal cancers.³³ This analysis reveals that margin positivity risk levels off when hospitals perform at least 10 annual esophagectomies for cancer. The notion of localizing complex surgical care to specialized, higher volume centers is not novel. There is a correlation between hospital volume and post-operative morbidity and mortality in various high risk surgeries.⁴⁵ In esophageal cancer specifically, improved cancer specific survival and lower complication rates have been reported at high volume institutions.^46,47 Based on this body of work, the Leapfrog group has identified both hospital and surgeon volume standards for high risk operative procedures, including esophagectomy. They recommend a minimum of 20 annual hospital-wide esophagectomies to decrease post-operative morbidity and mortality.⁴⁸ Our finding of a minimum of 10 esophagectomies for cancer in order to minimize margin positivity rates may have resulted in a different volume threshold because this analysis was focused on one specific potential quality measure, rather than holistic post-operative care. Nonetheless, several prior studies have questioned whether Leapfrog provides meaningful volume standards.^49,50

Limitations

This study has several limitations. As a retrospective cohort study, we can identify association but not causation. Thus, although we have identified factors associated with margin positivity in esophageal cancer, we are not able to draw conclusions regarding why these factors as associated with margin positivity. Second, the NCDB does not collect data on all patients with malignancy, which may limit the generalizability of our findings. However, because all CoC hospitals report to the NCDB, the hospitals most likely to be missing are small, unaccredited hospitals. Thus, our findings, particularly those surrounding hospital volume, would likely be magnified rather than diminished by this bias. Third, margin positivity rates are dependent upon pathologist training, and can vary based upon experience. We were unable to standardize pathologic evaluation in this study, which could alter margin positivity rates. However, our overall margin positivity rates are concordant with other studies in the literature. Despite these limitations, this paper has many benefits, including a large sample size representing a diverse patient population treated across the US.

Conclusion

Margin positivity rates are 9.37% following esophagectomy for malignancy, which have trended down over time. Several modifiable risk factors were found to be associated with margin status, including use of neoadjuvant therapy, surgical approach, and hospital resection volume. These data support surgical care at hospitals that perform an average of at least 10 esophagectomies for cancer per year to minimize risk of margin positivity.

SYNOPSIS.

Among 29,706 identified patients with esophageal cancer in the NCDB, 9.37% had positive margins, which decreased over time. Neoadjuvant therapy, open surgical approach, and treatment at a high-volume center were associated with lower margin positivity rates.