Association of Treatment Facility Characteristics With Overall Survival After Mohs Micrographic Surgery for T1a-T2a Invasive Melanoma

Key Points

Question

Are facility-level factors associated with long-term survival after Mohs micrographic surgery for early-stage invasive melanoma (T1a-T2a category)?

Findings

In this cohort study of 4062 adults with nonmetastatic, T1a-T2a melanoma treated at 462 centers, facility academic affiliation was associated with significantly improved survival for patients with T1a-T2a melanoma. Patients treated at higher-volume facilities also had improved survival compared with those treated at other centers.

Meaning

This study noted significant interfacility variation in overall survival after Mohs micrographic surgery for early-stage invasive melanoma; strategies aimed at creating more uniform treatment practices may help to improve overall patient survival for this procedure across centers.

Abstract

Importance

Early-stage melanoma, among the most common cancers in the US, is typically treated with wide local excision. However, recent advances in immunohistochemistry have led to an increasing number of these cases being excised via Mohs micrographic surgery (MMS). Although studies of resections for other cancers have reported that facility-level factors are associated with patient outcomes, it is not yet established how these factors may affect outcomes for patients treated with Mohs micrographic surgery.

Objective

To evaluate the association of treatment center academic affiliation and case volume with long-term patient survival after MMS for T1a-T2a invasive melanoma.

Design, Setting, and Participants

In a retrospective cohort study, 4062 adults with nonmetastatic, T1a-T2a melanoma diagnosed from 2004 to 2014 and treated with MMS in the National Cancer Database (NCDB) were identified. The NCDB includes all reportable cases from Commission on Cancer–accredited facilities and is estimated to capture approximately 50% of all incident melanomas in the US. Multivariable survival analyses were conducted using Cox proportional hazards models. Data analysis was conducted from February 27 to August 18, 2020.

Exposures

Treatment facility characteristics.

Main Outcomes and Measures

Overall survival.

Results

The study population included 4062 patients (2213 [54.5%] men; median [SD] age, 60 [16.3] years) treated at 462 centers. Sixty-two centers were top decile–volume facilities (TDVFs), which treated 1757 patients (61.9%). Most TDVFs were academic institutions (37 of 62 [59.7%]). On multivariable analysis, treatment at an academic center was associated with a nearly 30% reduction in hazard of death (hazard ratio, 0.730; 95% CI, 0.596-0.895). In a separate analysis, treatment at TDVFs was also associated with improved survival (hazard ratio, 0.795; 95% CI, 0.648-0.977).

Conclusions and Relevance

In this cohort study, treatment of patients with T1a-T2a invasive melanoma excised with MMS at academic and top decile–volume (≥8 cases per year) facilities was associated with improved long-term survival compared with those excised by MMS at nonacademic and low-volume facilities. Identification and protocolization of the practices of these facilities may help to reduce survival differences between centers.

This cohort study compares survival rates among differing facility types and volumes in patients who undergo Mohs micrographic surgery for nonmetastatic melanoma.

Introduction

Melanoma is one of the most common cancers in the US, with 106 110 incident cases estimated to be diagnosed in 2021.¹ Although the standard therapy for localized invasive melanoma is wide local excision, Mohs micrographic surgery (MMS) has been increasingly used as an alternative excisional approach,^2,3,4 with a recent study reporting that 7.9% of surgical melanoma cases in 2016 were excised using MMS.⁵ Mohs micrographic surgery offers the advantage of complete margin evaluation during the procedure while potentially limiting healthy tissue removal. Such complete margin evaluation is especially helpful in the case of facial tumors, for which National Comprehensive Cancer Network guidelines state that margins may be reduced for anatomic considerations, or tumors with clinically indistinct borders, such as lentigo maligna melanoma, for which appropriate margins may be difficult to determine clinically.^2,6 Recent work suggests that treatment of early-stage (T1a-T2a category) melanoma with MMS is associated with a modest survival advantage compared with traditional wide local excision.⁷ An additional consideration for the more technically complex MMS for melanoma procedure is how effectively it can be applied across centers. Although studies of resections for other cancers have reported that facility-level factors are associated with patient outcomes, it is not yet established how such factors may affect outcomes for patients treated with MMS for invasive melanoma.^8,9,10

Among facility-level factors that may impact patient survival, the most established are academic affiliation and facility case volume. Data from other novel oncologic surgical procedures suggest that facilities with higher case volumes are more likely to achieve clear surgical margins.^11,12 Based on the complexity of these techniques and the need for specialized training to optimize care, it has been reasoned that centralization of such procedures would result in improved outcomes.^{13,14,15,16,17} In addition, while facility-level influences on outcomes of MMS for melanoma are unknown, academic and high-volume centers have been reported to have improved long-term survival for nonmetastatic melanoma in general as well as for Merkel cell carcinoma.^18,19

In the present study, we aimed to characterize the association between facility characteristics, specifically academic status and case volume, and overall survival after MMS for T1a-T2a invasive melanoma. To accomplish this, we studied a national cohort of 4062 cases captured in the National Cancer Database (NCDB) with diagnostic years 2004-2014. We limited the analysis to T1a-T2a category tumors because this is the disease subset that was recently reported to have improved outcomes with MMS vs traditional wide local excision.⁷

Methods

Data Source

Data originated from the NCDB with diagnosis years from 2004 to 2015 for analysis of yearly case numbers and 2004 to 2014 for the remainder of the analysis. Data analysis was conducted from February 27 to August 18, 2020. The NCDB is a nationwide clinical surveillance resource data set that includes approximately 70% of all newly diagnosed cancers in the US, with coverage for incident melanoma estimated at 50% from over 1500 cancer programs as previously described.^20,21 Data from the NCDB have been used in numerous studies of treatment-related cancer outcomes.^{22,23,24,25,26} This study was determined to be exempt from institutional review by the Yale Human Investigation Committee for use of existing data with deidentified patients. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.

Study Population

Study inclusion and exclusion criteria are presented in Figure 1. We identified melanoma cases with a primary site in the skin using the International Classification of Diseases for Oncology Third Edition histology codes 8720 to 8780 and primary site codes C44.0-44.9 (n = 525 271). Analysis was limited to cases treated with MMS (n = 22 975), defined as patients who received MMS as either the primary treatment or as definitive surgery after shave biopsy, regardless of the size of the margins. Yearly MMS for invasive melanoma case volumes were calculated before additional exclusions. Additional inclusion criteria for survival analysis were American Joint Committee on Cancer Cancer Staging Manual, Eighth Edition pathologic T category 1a-2a (<1.0 mm with or without ulceration or 1.1-2.0 mm without ulceration), nonmetastatic (staged as cM0 and not pM1), and cases from institutions with more than 1 year reporting MMS for melanoma procedures that were not treated with any adjuvant therapy (n = 6505). Owing to incomplete survival data for patients in the 2015 diagnosis year, these cases were excluded (n = 838). We also excluded patients if they had previous or other cancer diagnoses (n = 1590), had clinical evidence of nodal disease (cN1-3) (n = 13), or had missing or incomplete follow-up data (n = 2). For a sensitivity analysis, we excluded all cases that did not have confirmed clinically negative lymph node disease status (n = 873). All inclusion and exclusion criteria were determined a priori.

Figure 1. Case Selection and Exclusion Criteria.

NCDB indicates National Cancer Database; MMS, Mohs micrographic surgery.

Statistical Analysis

Patients were defined as not having received adjuvant therapy if they were coded as not having received any form of radiotherapy, systemic chemotherapy, or immunotherapy. Comorbidities were measured by the Charlson/Deyo comorbidity index score, a validated method that assigns scores to a patient’s chronic medical conditions and estimates long-term mortality based on the sum of these scores.^27,28 Average annual case volume was determined by dividing the total number of melanoma cases of any stage treated with MMS by the number of reporting years of the center, as previously described.^29,30 After excluding facilities with only 1 year reporting on MMS for melanoma cases, facilities were defined as top decile–volume facilities (TDVFs) if their average case volume was 8 or more cases per year, corresponding to the 90th percentile of facilities; the remainder of the institutions were considered low-volume facilities (LVFs).^31,32,33 A sensitivity analysis was also performed using the median case volume (1.71) to divide TDVFs and LVFs. Our primary outcome was overall survival. Per staging guidelines, Breslow depths were rounded to the nearest tenth of a millimeter.³⁴

Academic affiliation was determined based on the Commission on Cancer accreditation category for the cancer center reporting the case.³⁵ Group differences were tested using the Wilcoxon rank sum test for continuous variables and the χ² test for categorical variables. The Somers D statistic was used to compare differences in the magnitude of association between facility type and volume and patient age. Multivariable survival analyses were conducted using Cox proportional hazards regression models using either facility case volume or facility type in the model due to their collinearity. The appropriateness of the proportional hazards assumption was confirmed using Schoenfeld residuals.³⁶ All nonstaging variables outlined in Table 1 were tested for appropriateness for inclusion as covariates in these models using Akaike information criterion minimization to ensure parsimony of the multivariable model.³⁷ Subgroup analyses were performed for melanomas with a primary site in the head and neck. Sensitivity analyses excluding cases without confirmed clinically negative lymph node disease status were conducted. Statistical significance was determined at the P < .05 level; all testing was 2-tailed and unpaired. Data analysis was performed using Stata, version 13 (StataCorp LLP).

Table 1. Characteristics of the Analytic Sample Including 4062 Patients.

Variable	No. (%)
Age, median (SD), y	60 (16.3)
Sex
Female	1849 (45.5)
Male	2213 (54.5)
Race/ethnicity
Non-Hispanic White	3913 (96.3)
Black	16 (0.4)
Hispanic White	29 (0.7)
Asian/Pacific Islander	4 (0.1)
Other/unknowna	100 (2.5)
Charlson/Deyo score
0	3739 (92.0)
1	274 (6.8)
2	37 (0.9)
≥3	12 (0.3)
Insurance
Private	2346 (57.7)
Medicare	1424 (35.1)
Other governmental	119 (2.9)
None	91 (2.2)
Unknown	82 (2.0)
Primary site
Head and neck	1591 (39.2)
Trunk	980 (24.1)
Extremities	1461 (36.0)
Overlapping/NOS	30 (0.7)
Histologic characteristics
Superficial spreading	1310 (32.3)
Lentigo maligna melanoma	618 (15.2)
Nodular	85 (2.1)
Malignant acral lentiginous	27 (0.7)
Malignant desmoplastic	30 (0.7)
Spindle cell	19 (0.5)
Other/NOS	1973 (48.6)
Ulceration
Not present	3927 (96.7)
Present	135 (3.3)
Breslow depth, mm
<0.8	3144 (77.4)
0.8-1.0	486 (12.0)
1.1-2.0	432 (10.6)
Clinical node stage reported
Yes	3189 (78.5)
No	873 (21.5)
Facility volume
Low	1549 (38.1)
Top decile	2513 (61.9)
Facility type
Nonacademic	1757 (43.2)
Academic	2305 (56.8)

Abbreviations: NA, not applicable; NOS, not otherwise specified.

^aThis group includes American Indian, Aleutian, or Eskimo as well as those for whom race was coded as Other or Unknown. The Asian/Pacific Islander group includes a large number of subgroups: Chinese, Japanese, Filipino, Hawaiian, Korean, Vietnamese, Laotian, Hmong, Kampuchean, Thai, Asian Indian or Pakistani, not otherwise specified (NOS), Asian Indian, Pakistani, Micronesian, NOS, Chamorran, Guamanian, NOS, Polynesian, NOS, Tahitian, Samoan, Tongan, Melanesian, NOS, Fiji Islander, New Guinean, Other Asian, including Asian, NOS and Oriental, NOS, and Pacific Islander, NOS.

Results

The characteristics of the study sample are presented in Table 1. Of the 4062 patients, 1849 were women (45.5%) and 2213 (54.5%) were men. The median (SD) age of patients was 60 (16.3) years, with most being non-Hispanic White (3913 [96.3%]), having no comorbidities (3739 [92.0%]), and carrying private insurance (2346 [57.7%]). The most common tumor primary site was on the head and neck (1591 [39.2%]), followed by the extremities (1461 [36.0%]) and the trunk (980 [24.1%]). Among tumors with specified histologic patterns, the most common subtype was superficial spreading (1310 [32.3%]). The majority of tumors were less than 0.8 mm (3144 [77.4%]) and without ulceration (3927 [96.7%]). A total of 3189 patients (78.5%) had confirmed clinically negative lymph node disease status. Patients were treated at a total of 462 facilities. Of these, 127 (27.5%) were defined as academic institutions, which treated 2305 patients (56.8%) of patients. Sixty-two centers (13.4%) were TDVFs, which treated 61.9% (2513 of 4062) of the patients. Most of the 62 TDVFs were academic institutions (37 [59.7%]). Median facility case volume was 3.71 (interquartile range [IQR], 2.00-9.87) at academic centers and 1.75 (IQR, 1.29-2.90) at nonacademic centers (Wilcoxon rank-sum test, P < .001). The distribution of case volume by facility academic affiliation is presented in Figure 2. Among LVFs, annual case volume ranged from 1.00 to 7.82, with a median of 1.80 (IQR, 1.33-3.00). Among TDVFs, the annual case volume ranged from 8.00 to 127.29, with a median of 15.7 (IQR, 10.5-26.25). Median follow-up was 4.45 years with a maximum follow-up of 12.90 years (IQR, 2.48-7.04 years).

Figure 2. Annual Case Volume Stratified by Facility Type.

Circles indicate the median values; boxes indicate interquartile ranges, and error bars represent the upper and lower adjacent values.

Differences in the demographic and tumor characteristics of cases treated at academic vs nonacademic institutions as well as TDVFs vs LVFs are presented in Table 2. Patients treated at academic centers were significantly older than those treated at nonacademic centers (median [SD], 64 [12.8] vs 53 [18.1] years; Wilcoxon rank sum test, P < .001), with a more than 10-year difference in median age (Somers D, 0.36; 95% CI, 0.33-0.39). This pattern was similarly present in TDVFs vs LVFs, although the difference was less pronounced (62 [16.0] vs 58 [16.4] years; Somers D, 0.12; 95% CI, 0.09-0.16). In addition, men constituted a larger proportion compared with women treated at academic centers (1353 [58.7%] vs 952 [41.3%]; P < .001) and TDVFs (1424 [56.7%] vs 1089 [43.3%]; P < .001). Tumors on the head and neck composed a significantly larger proportion of the cases at academic vs nonacademic centers (1119 [48.6%] vs 472 [26.9%]; P < .001) and TDVFs vs LVFs (1182 [47.0%] vs 409 [26.4%]; P < .001). The most common tumor primary site at nonacademic centers (752 [42.8%]) and LVFs (666 [43.0%]) was the extremities. Furthermore, superficial spreading tumors composed a larger proportion of those treated at nonacademic vs academic centers (656 [37.3%] vs 654 [28.4%]; P < .001) and LVFs vs TDVFs (596 [38.5%] vs 714 [28.4%]; P < .001). Conversely, lentigo maligna melanoma tumors composed slightly, albeit significantly, more of the cases at academic vs nonacademic centers (464 [20.1%] vs 154 [8.8%]; P < .001) and TDVFs vs LVFs (497 [19.8%] vs 121 [7.8%]; P < .001). In addition, nodal staging was more likely to be reported by academic vs nonacademic centers (1837 [79.7%] vs 1352 [77.0%]; P = .04) and TDVFs vs LVFs (2042 [81.3%] vs 1147 [74.0%]; P < .001). Tumors treated at academic centers vs nonacademic centers (1857 [80.6%] vs 1287 [73.2%]; χ²: P <.001) and TDVFs vs LVFs (2048 [81.5%] vs 1096 [70.8%]; χ²: P <.001) were also shallower in Breslow depth. Histopathologic ulceration was less likely at academic vs nonacademic centers (63 [2.7%] vs 72 [4.1%]; χ²: P = .02) and TDVFs vs LVFs (64 [2.6%] vs 71 [4.6%]; χ²: P < .001).

Table 2. Characteristics of the Analytic Sample by Facility Type and Volume.

Variable	Facility type, No. (%)		χ2 P value	Facility volume, No. (%)		χ2 P value
	Nonacademic	Academic		Low	Top decile
Age, median (SD), y	53 (18.1)	64 (12.8)	<.001a	58 (16.4)	62 (16.0)	<.001a
Sex
Female	897 (51.0)	952 (41.3)	<.001	760 (49.1)	1089 (43.3)	<.001
Male	860 (49.0)	1353 (58.7)		789 (50.9)	1424 (56.7)
Race/ethnicity
Non-Hispanic White	1694 (96.4)	2219 (96.3)	.63	1495 (96.5)	2418 (96.2)	.004
Black	9 (0.5)	7 (0.3)		11 (0.7)	5 (0.2)
Hispanic White	14 (0.8)	15 (0.6)		15 (1.0)	14 (0.6)
Asian/Pacific Islander	2 (0.1)	2 (0.1)		2 (0.1)	2 (0.1)
Other/unknownb	38 (2.2)	62 (2.7)		26 (1.7)	74 (2.9)
Charlson/Deyo score
0	1594 (90.7)	2145 (93.1)	.03	1388 (89.6)	2351 (93.6)	<.001
1	135 (7.7)	139 (6.0)		135 (8.7)	139 (5.5)
2	22 (1.2)	15 (0.6)		19 (1.2)	18 (0.7)
≥3	6 (0.3)	6 (0.3)		7 (0.4)	5 (0.2)
Insurance
Private	1142 (65.0)	1204 (52.2)	<.001	951 (61.4)	1395 (55.5)	<.001
Medicare	466 (26.5)	958 (41.6)		478 (30.9)	946 (37.6)
Other governmental	57 (3.2)	62 (2.7)		43 (2.8)	76 (3.0)
None	54 (3.1)	37 (1.6)		43 (2.8)	48 (1.9)
Unknown	38 (2.2)	44 (1.9)		34 (2.2)	48 (1.9)
Primary site
Head and neck	472 (26.9)	1119 (48.6)	<.001	409 (26.4)	1182 (47.0)	<.001
Trunk	523 (29.8)	457 (19.8)		459 (29.6)	521 (20.7)
Extremities	752 (42.8)	709 (30.8)		666 (43.0)	795 (31.6)
Overlapping/NOS	10 (0.6)	20 (0.9)		15 (1.0)	15 (0.6)
Histologic characteristics
Superficial spreading	656 (37.3)	654 (28.4)	<.001	596 (38.5)	714 (28.4)	<.001
Lentigo maligna melanoma	154 (8.8)	464 (20.1)		121 (7.8)	497 (19.8)
Nodular	50 (2.8)	35 (1.5)		60 (3.9)	25 (1.0)
Malignant acral lentiginous	12 (0.7)	15 (0.6)		10 (0.6)	17 (0.7)
Malignant desmoplastic	6 (0.3)	24 (1.0)		15 (1.0)	15 (0.6)
Spindle cell	5 (0.3)	14 (0.6)		5 (0.3)	14 (0.6)
Other/NOS	874 (49.7)	1099 (47.7)		742 (47.9)	1231 (49.0)
Ulceration
Not present	1685 (95.9)	2242 (97.3)	.02	1478 (95.4)	2449 (97.4)	<.001
Present	72 (4.1)	63 (2.7)		71 (4.6)	64 (2.6)
Breslow depth, mm
<0.8	1287 (73.2)	1857 (80.6)	<.001	1096 (70.8)	2048 (81.5)	<.001
0.8-1.0	246 (14.0)	240 (10.4)		242 (15.6)	244 (9.7)
1.1-2.0	224 (12.8)	208 (9.0)		211 (13.6)	221 (8.8)
Clinical node stage reported
Yes	1352 (77.0)	1837 (79.7)	.04	1147 (74.0)	2042 (81.3)	<.001
No	405 (23.0)	468 (20.3)		402 (26.0)	471 (18.7)
Facility type
Nonacademic	1757 (100)	0	NA	1025 (66.2)	732 (29.1)	<.001
Academic	0	2305 (100)		524 (33.8)	1781 (70.9)
Facility volume
Low	1025 (58.3)	524 (22.7)	<.001	1549 (100)	0	NA
Top decile	732 (41.7)	1781 (77.3)		0	2513 (100)

Abbreviations: NA, not applicable; NOS, not otherwise specified.

^aP value refers to Wilcoxon rank sum test rather than χ².

^bThis group includes American Indian, Aleutian, or Eskimo as well as those for whom race was coded as Other or Unknown. The Asian/Pacific Islander group includes a large number of subgroups: Chinese, Japanese, Filipino, Hawaiian, Korean, Vietnamese, Laotian, Hmong, Kampuchean, Thai, Asian Indian or Pakistani, not otherwise specified (NOS), Asian Indian, Pakistani, Micronesian, NOS, Chamorran, Guamanian, NOS, Polynesian, NOS, Tahitian, Samoan, Tongan, Melanesian, NOS, Fiji Islander, New Guinean, Other Asian, including Asian, NOS and Oriental, NOS, and Pacific Islander, NOS.

Unadjusted (SE) survival for cases treated at LVFs was 94.3% (0.6%) at 3 years, 90.0% (0.9%) at 5 years, and 78.7% (1.8%) at 10 years; at TDVFs, survival was 95.7% (0.4%) at 3 years, 90.6% (0.7%) at 5 years, and 78.8% (1.6%) at 10 years. Survival at nonacademic centers was 95.0% (0.6%) at 3 years, 90.7% (0.8%) at 5 years, and 80.5% (1.7%) at 10 years; at academic centers, survival was 95.3% (0.5%) at 3 years, 90.2% (0.8%) at 5 years, and 77.4% (1.7%) at 10 years. Multivariable survival analyses of the 4062 patients were then conducted, with separate models created using either facility type or facility volume (Table 3). There were 417 total events (deaths) during the study period. The regression including facility type demonstrated that treatment at academic facilities vs nonacademic facilities was associated with significantly improved overall survival, with a nearly 30% reduction in the hazard of death (hazard ratio [HR], 0.730; 95% CI, 0.596-0.895). Similarly, treatment at TDVFs was associated with improved survival compared with LVFs (HR, 0.795; 95% CI, 0.648-0.977). Subgroup analysis of head and neck primary melanomas also demonstrated an association with improved survival in academic vs nonacademic centers (HR, 0.718; 95% CI, 0.539-0.955) and TDVFs vs LVFs (HR, 0.700; 95% CI, 0.526-0.933).

Table 3. Multivariable Analysis of Factors Associated With Survival^a.

Variable	Facility type model		Facility volume model
	HR (95% CI)	P value	HR (95% CI)	P value
Ageb	1.085 (1.072-1.097)	<.001	1.084 (1.072-1.097)	<.001
Sex
Female	1 [Reference]		1 [Reference]
Male	1.652 (1.320-2.066)	<.001	1.662 (1.328-2.080)	<.001
Charlson/Deyo score
0	1 [Reference]		1 [Reference]
1	1.640 (1.220-2.205)	.001	1.657 (1.232-2.229)	.001
2	3.576 (2.057-6.216)	<.001	3.778 (2.179-6.551)	<.001
≥3	1.362 (0.190-9.771)	.76	1.367 (0.190-9.815)	.76
Insurance
Private	1 [Reference]		1 [Reference]
Medicare	1.630 (1.200-2.213)	.002	1.637 (1.206-2.222)	.002
Other government	2.781 (1.388-5.574)	.004	2.755 (1.375-5.520)	.004
None	2.925 (1.342-6.373)	.007	2.870 (1.317-6.252)	.008
Unknown	0.902 (0.410-1.984)	.80	0.905 (0.412-1.990)	.81
Primary site
Head and neck	1 [Reference]		1 [Reference]
Trunk	1.051 (0.815-1.354)	.70	1.047 (0.811-1.352)	.72
Extremities	0.745 (0.580-0.956)	.02	0.749 (0.582-0.963)	.02
Overlapping/NOS	1.047 (0.385-2.852)	.93	1.036 (0.381-2.821)	.94
Ulceration
Not present	1 [Reference]		1 [Reference]
Present	2.231 (1.543-3.225)	<.001	2.228 (1.540-3.222)	<.001
Breslow depth, mm
<0.8	1 [Reference]		1 [Reference]
0.8-1.0	1.248 (0.929-1.677)	.14	1.247 (0.928-1.677)	.14
1.1-2.0	1.461 (1.108-1.928)	.007	1.455 (1.102-1.922)	.008
Facility volume
Low	NA		1 [Reference]
Top decile	NA		0.795 (0.648-0.977)	.03
Facility type
Nonacademic	1 [Reference]		NA
Academic	0.730 (0.596-0.895)	.002	NA

Abbreviations: HR, hazard ratio; NA, not applicable; NOS, not otherwise specified.

^aRace/ethnicity, histologic characteristics, and clinical node reporting status removed as variables from multivariate model after Akaike information criterion minimization.

^bHazard ratio for age reported for each additional year of age.

A sensitivity analysis was conducted for each of these models excluding all cases without confirmed clinically negative lymph node disease status (eTable in the Supplement). These regressions reproduced the finding of improved survival for patients treated at academic vs nonacademic facilities (HR, 0.713; 95% CI, 0.555-0.916; P = .008). In the sensitivity analysis, treatment at TDVFs did not have a statistically significant association with improved outcomes (HR, 0.828; 95% CI, 0.643-1.067; P = .14). In addition, to test the robustness of the association of facility volume with survival, we repeated the multivariable survival analysis comparing facilities above and below the median for annual facility case volume (1.71) and found a similar association of improved survival with higher facility case volume (HR, 0.640; 95% CI, 0.474-0.864; P = .004).

Discussion

In this study, evidence suggests that facility characteristics are significantly associated with long-term survival after MMS for early-stage (T1a-T2a category) invasive melanoma. To our knowledge, this is the first evidence of the role of facility-level factors in influencing patient outcomes after MMS for invasive melanoma, and the findings are consistent with data from other novel oncologic surgical procedures.^11,12 Although the absolute differences in survival between centers that we found for early-stage invasive melanoma are more modest than those found for more aggressive cancers, this magnitude of difference should be considered in the context of the higher incidence of early-stage melanoma as well as its favorable prognosis. We also recognize that unadjusted survival between centers was comparable; however, given the significant differences in baseline characteristics of the patients treated at these centers, in particular with respect to age, such unadjusted estimates have limited utility. These results have implications for furthering the understanding of associations between facility-level characteristics and cancer outcomes as a whole as well as for the continued use of MMS for invasive melanoma.

Studies of the association between facility characteristics and cancer outcomes have suggested a number of underlying reasons for the observed survival differences between centers. Previous analyses of several cancer types have found that high-volume and academic hospitals are more likely to provide care according to guideline recommendations, which leads to improved long-term survival and decreased health care cost,^38,39,40,41 with some studies suggesting that this is due to these centers being more likely to discuss patients at multidisciplinary tumor boards.^{42,43,44,45,46,47} Although MMS is not yet included in national guidelines for invasive melanoma, it is possible that high-volume academic centers are more likely to follow the most current evidence and recommendations for the procedure in lieu of such guidelines. Although we controlled for histologic characteristics and primary tumor site in the survival analysis, we found that a larger proportion of patients treated at academic institutions and TDVFs had lentigo maligna melanoma lesions and/or tumors located on the head and neck—the disease subtypes for which MMS for melanoma has been best established.^6,48,49,50 Academic institutions and TDVFs were also more likely to report nodal stage, suggesting that they may also perform more extensive staging workups. Facility practice differences may also bear out in differential use of immunohistochemical stains to aid in the interpretation of sections, which has been shown to be necessary for achieving local control with MMS for melanoma.⁵¹ While variations in the use of immunohistochemical stains for interpreting MMS sections for melanoma according to facility volume and academic affiliation have not been established, a recent study⁵ found wide geographic variations in its use, ranging in the most recent study years (2013-2016) from 15.2% of cases in the lowest-use region to 51.4% of cases in the highest-use region.

Another hypothesis for the differences in patient outcomes between facilities is that the increased experience at higher-volume centers may lead to improved decision-making and better surgical technique. There has been an increasing recognition that high-volume surgeons, as well as overall high-volume facilities irrespective of individual surgeon volume, have substantially improved oncologic surgical outcomes.^52,53,54 This volume-outcome relationship may be especially prominent in the case of MMS for melanoma, where there is relatively less experience with the procedure than with standard surgical excision, as well as a lack of consensus on best practices. Such lack of consensus on best practices is particularly germane to the issue of the minimum acceptable margin for MMS excisions of invasive melanomas. Although the current standard of care is excision with 1-cm margins, local control has been demonstrated for early-stage melanoma with MMS excisions using less than a 1-cm margin.^55,56 A consensus on the minimum acceptable margin for these excisions has not yet been established. It is plausible that, in the absence of such a consensus margin, different practices with regard to minimum margins between centers may contribute to the observed difference in outcomes.

Although national guidelines have not yet specifically endorsed the use of MMS for invasive melanoma, its use for this indication has rapidly increased.^{5,7,49,51,56,57,58} As the efficacy of the practice as a whole is evaluated, it is especially necessary to ensure that it is performed with the best available techniques and evidence. While recent studies of national data have shown that MMS has similar or slightly improved survival relative to wide local excision for selected cases of invasive melanoma, patients treated at TDVFs compose a disproportionate share of the study cohorts.^7,59 In addition, much of the work showing the efficacy of MMS for invasive melanoma using MART-1 staining has been coordinated by institutions that have extensive experience with the procedure.^56,58,60

The findings of the present study as well as the increasing use of MMS for melanoma support the development of consensus guidelines for the procedure’s use.^2,5 Such guidelines may serve to reduce variations in practices and patient outcomes between facilities as has occurred after the establishment of consensus recommendations for other surgical procedures.^61,62,63,64 It has also been proposed that further improvements in the quality of care and outcomes achieved by surgeons performing MMS for melanoma across centers may be achieved by the establishment of training standards for the procedure as has been successfully executed for head and neck robotic surgery, another novel oncologic procedure.^{65,66,67,68,69,70}

Limitations

Our study is limited by several factors. We were unable to control our analysis for the use of immunohistochemical stains to evaluate MMS sections owing to the absence of this variable from the data set. It is possible that a component of the differences in outcomes between facilities can be attributed to differential use of this technique. In addition, the outcomes in our data set are reported as all-cause survival and do not include information on local recurrence or disease-specific survival. Thus, we were unable to evaluate the association between facility characteristics and these outcomes. In addition, a significant limitation of the NCDB is that it is a hospital-based registry, which limits the generalizability of our results to community settings wherein there may be surgeons treating invasive melanoma at high volume and achieving similarly excellent outcomes. The study’s generalizability is further limited by the fact that a number of academic hospitals do not report data to the registry.

Conclusions

We found that facility case volume and academic affiliation are significantly associated with long-term survival after MMS for T1a-T2a invasive melanoma. Further study of the underlying reasons for these differences in survival, as well as the development of consensus standards for the technique, may help to reduce such variations in patient outcomes across treatment centers.

Supplement.

eTable. Sensitivity Multivariable Analysis of Factors Associated With Survival Excluding Cases With Unreported Clinical Node Stage