Management of the positive sentinel lymph node in the post-MSLT-II era

Abstract

Background and Objectives

The publication of MSLT-II shifted recommendations for management of sentinel lymph node biopsy positive (SLNB+) melanoma to favor active surveillance. We examined trends in immediate completion lymph node dissection (CLND) following publication of MSLT-II.

Methods

Using a prospective melanoma database at a high-volume center, we identified a cohort of consecutive SLNB+ patients from July 2016 to April 2019. Patient and disease characteristics were analyzed with multivariate logistic regression to examine factors associated with CLND.

Results

Two hundred and thirty-five patients were included for analysis. CLND rates were 67%, 33%, and 26% for the year before, year after, and second-year following MSLT-II. Factors associated with undergoing CLND included primary located in the head and neck (59% vs 33%, P = .003 and odds ratio [OR], 5.22, P = .002) and higher sentinel node tumor burden (43% vs 10% for tumor burden ≥0.1 mm, P < .001 and OR, 8.64, P = .002).

Conclusions

Rates of CLND in SLNB+ melanoma decreased dramatically, albeit not uniformly, following MSLT-II. Factors that increased the likelihood of immediate CLND were primary tumor located in the head and neck and high sentinel node tumor burden. These groups were underrepresented in MSLT-II, suggesting that clinicians are wary of implementing active surveillance recommendations for patients perceived as higher risk.

1 |. INTRODUCTION

Melanoma comprises less than 1% of all skin cancer diagnoses in the United States but is responsible for about 80% of skin cancer-associated mortality.^1,2 Prognosis varies considerably by stage, with lymph node involvement remaining one of the most important prognostic factors.³ Over the past decade, novel systemic therapies have revolutionized the treatment of advanced (Stages III-IV) melanoma, leading to improved disease-free survival in these groups.⁴ These novel systemic therapies were designed to suppress the proliferation of melanoma cancer cells, as with BRAF/MEK inhibitors, or stimulate a robust immunogenic tumor response, as with checkpoint inhibitors. The surgical management of melanoma has also evolved in the past several years. Namely, two independent randomized control trials have evaluated the role of immediate completion lymph node dissection (CLND) for sentinel lymph node biopsy positive (SLNB+) patients.^5,6 SLNB is used in clinically nodenegative melanoma (ie, no evidence of involved nodes on physical exam or any available imaging studies) for primary lesions greater than 0.8 mm Breslow thickness.⁷ The procedure was designed to be diagnostic as a staging procedure and has been shown to be a reliable prognostic procedure in melanoma.⁸ However, the therapeutic benefit of immediate CLND after positive SLNB came into question after publication of the MSLT-II trial and De-COG SLT. MSLT-II, an international multicenter randomized trial of patients with a positive SLNB, demonstrated improved 3-year disease-free survival with immediate CLND but no long-term melanoma-specific survival benefit in the 1755 patients studied.⁶ De-COG SLT, another multicenter randomized trial in Germany, found no survival benefit in 473 patients with positive SLNB undergoing CLND versus active surveillance at 72 months follow-up.⁵

Based on the results of these trials, the most recent guidelines from the National Comprehensive Cancer Network (NCCN) recommend active surveillance as the “preferred” strategy for SLNB+ melanoma. In the latest guidelines from the Society of Surgical Oncology, active surveillance is recommended in most high- and low-risk patients.⁹ Despite these broad changes in guidelines and professional consensus, some still cautioned against active surveillance in a high-risk subset of patients with tumor features not included in trial criteria (extracapsular nodal extension, involvement of more than one nodal basin, primary tumor microsatellitosis).¹⁰ In this group of patients, there may be an increased risk of understaging with SLNB alone, which has subsequent implications for adjuvant therapy.¹¹ In addition, MSLT-II included a relatively small cohort of patients with head and neck melanomas (13.7% of study population) and overall sentinel lymph node tumor burden was low (interquartile range [IQR] for the observation group was 0.23–1.38 mm).

In light of the rapidly changing practice for treatment of SLNB+ melanoma, we identified two aims: first, to analyze trends in the use of immediate CLND vs active surveillance in our institution since the publication MSLT-II results; and second, to identify patient and disease factors that increased the likelihood of immediate CLND. Given the limitations of MSLT-II, we hypothesized that practice patterns may be more variable among patients not well represented by the study. At this time, national data examining the trends before, during, and after trial publication are unavailable. As a high-volume center, we were able to identify consecutive cases to analyze trends in dissection rates over time.

2 |. METHODS

2.1 |. Study design

This was a retrospective analysis of a prospectively identified cohort performed at a high-volume academic institution. The study was approved by the Institutional Review Board. Consecutive patients were identified as SLNB+ at time of surgery between July 2016 and April 2019 in an established, prospective institutional database. This data set captures the period immediately before and immediately after MSLT-II trial data first became publicly available in June 2017. Inclusion of patients in the year before this change in standard of care allowed us to understand trends in adoption of these new results. Patients with a positive SLNB went on to receive either active surveillance according to NCCN guidelines or immediate CLND. Patients who underwent CLND after ultrasound surveillance detected regional recurrence were included in this study under the active surveillance group.

Tumor-specific characteristics included for analysis were location (head and neck, trunk, upper extremity, and lower extremity), Breslow thickness, ulceration, largest nodal metastasis size, and number and ratio of SLNB+ nodes. Very high-risk tumor features (extracapsular extension, involvement of more than one nodal basin, primary tumor microsatellitosis) were infrequently encountered and as such were not included in the final analysis. Patient characteristics included for analysis were demographic information (age, gender, and race), body mass index (BMI), use of interpreter, distance between patient’s home address and the main hospital, primary insurance (Medicaid, Medicare, private, and self-pay), secondary insurance (yes/no), and comorbidities (diabetes, cardiac process, pulmonary process, and end-stage renal disease). In addition, we included timing of SLNB, receipt of adjuvant therapy, and whether the case was discussed at a multidisciplinary tumor board.

2.2 |. Statistical analysis

Statistical analyses were performed using JMP Pro 14.3.0 and STATA version 16.1. P <.05 were considered significant. Univariate and multivariate logistic regression was performed to determine odds of immediate completion lymph node dissection. χ² tests of independence and unpaired t tests compared characteristics of the active surveillance vs immediate completion lymph node dissection groups.

3 |. RESULTS

3.1 |. Study population

Two hundred and thirty-five SLNB+ patients were identified during the study period. Sufficient data were available on each patient and, as such, no patients were excluded from analysis. Rates of CLND decreased from 67% of SLNB+ patients in the year preceding MSLT-II to 33% in the year following publication of MSLT-II, and 26% in the second year following MSLT-II. Additionally, CLND rates were plotted by quarter after MSLT-II publication, which identified an inflection point 3 months after MSTL-II publication and is used as our institution’s timepoint for implementation of study guidelines (Figure 1). Overall, 90 of 235 patients (38%) with positive SLNB underwent immediate CLND and 145 of 235 patients (62%) underwent active surveillance (AS) (Table 1). The AS group was primarily composed of patients treated after publication of MSLT-II (88%). The majority of patients were Caucasian (90%) and male (58%), which is consistent with national melanoma demographics. The average age in this cohort was 57 (range, 4–93.2 years old). Mean BMI was 30.59 (range, 17.0–48.8). 37% of tumors were ulcerated. There was wide variation in Breslow depth (mean, 3.37 and IQR, 1.5–4.1 mm). Diameter of largest sentinel lymph node metastasis also varied widely (mean, 2.58 and IQR, 0.15–3.0). Patients were also stratified by Rotterdam criteria (<0.1mm, 0.1–1.0 mm, and >1.0 mm for the largest diameter of the largest metastasis in the sentinel node). 14% of patients had micrometastatic nodal disease; and tumor burden of 0. 1 to 1 mm made up 44% of the total compared with 43% for more than 1 mm. Most patients had one or two positive sentinel nodes (mean, 1.42 and range, 1–6), with only 7.7% of patients having greater than two positive sentinel nodes (Table 2).

FIGURE 1.

Quarterly trends in immediate completion lymph node dissection rates relative to the publication of MSLT-II. Immediate completion lymph node dissection rates by quarter (Q, or 3 month periods) in the two years following publication of MSLT-II, here labeled as peri MSLT-II and post MSLT-II. There was a steep and persistent decrease in rates of CLND 3 months after publication, marking an inflection point at Q2 in the peri MSLT-II period

TABLE 1.

Patient characteristics, presentation at tumor board, and timing of SLNB

	CLND (90)	Active surveillance (145)	P value
Patient characteristics
Sex, n (%)
Female	36 (40%)	63 (43%)	.684
Age
Mean (SD)	54.7 (16.7)	58.6 (16.2)	.074
Race, n (%)			.869
White	81 (90%)	131 (90%)
Black, Hispanic, or Asian	2 (3%)	4 (3%)
Unknown	6 (7%)	11 (7%)
BMI
Mean (SD)	31.5 (6.28)	30.0 (5.94)	.079
Comorbidity (any)a, n (%)
Present	24 (27%)	46 (32%)	.464
Distance traveled, miles
Mean (IQR)	96 (38–118)	112 (42–106)	.4697
Primary insurance n (%)			.865
Private	56 (62%)	86 (59%)
Medicare	26 (29%)	46 (32%)
Medicaid	5 (6%)	10 (7%)
Self-pay	3 (3%)	3 (2%)
Other
Tumor board, n (%)
Presented at tumor board	38 (42%)	43 (30%)	.066
Timing of SLNB, n (%)			<.001
Post-MSLT-II (after 6/8/2017)	24 (27%)	69 (48%)
Peri-MSLT-II (6/8/2017–6/8/2018)	29 (32%)	58 (40%)
Pre-MSLT-II (before 6/8/2017)	37 (41%)	18 (12%)
Adjuvant therapy, n (%)
Receipt of therapy 61	26 (29%)	35 (24%)	.541

Abbreviations: BMI, body mass index; CLND, completion lymph node dissection; IQR, interquartile range; SLNB, sentinel lymph node biopsy.

^aCardiac disease, pulmonary disease, diabetes mellitus, or end-stage kidney disease.

TABLE 2.

Tumor and sentinel node characteristics

	CLND (90)	Active surveillance (145)	P value
Tumor characteristics
Breslow depth
Mean (IQR)	3.46 (1.5–4.3)	3.32 (1.5–3.9)	.736
Primary site, n (%)			.002
Upper extremity	15 (17%)	26 (18%)
Lower extremity	12 (13%)	43 (30%)
Trunk	37 (41%)	58 (40%)
Head and neck	26 (29%)	18 (12%)
Ulceration, n (%)
Present	33 (37%)	55 (38%)	.890
Sentinel node characteristics
Number of positive sentinel lymph nodes, n (%)			.081
1 node	55 (61%)	107 (74%)
2 nodes	25 (28%)	30 (21%)
3 nodes	5 (6%)	6 (4%)
>3 nodes	5 (6%)	2 (1%)
Diameter of largest sentinel lymph node metastasis (mm), n (%)			<.001
<0.1 mm	3 (3%)	28 (19%)
0.1–1 mm	37 (41%)	65 (45%)
>1 mm	50 (56%)	52 (36%)
Mean (IQR)	3.49 (0.4–5.0)	2.03 (0.1–2.0)	.011

Abbreviations: CLND, completion lymph node dissection; IQR, interquartile range.

3.2 |. Patient and disease factors affecting CLND

On univariate analysis, there was no significant difference in rates of CLND based on patient sex, age, race, presence of any comorbidity or BMI (Table 1). Patients with head and neck melanoma had a CLND rate of 59% vs 33% for other sites (P = .003). Conversely, patients with lower extremity melanoma had lower rates of CLND compared with other sites (43% vs 22%, P = .004). Patients with primary melanoma of the trunk and upper extremity had similar rates of CLND compared with other sites (P = .89 and .86, respectively). There was no difference in use of CLND based on ulceration of the primary tumor or Breslow thickness. Treatment strategy varied considerably by sentinel node burden according to Rotterdam criteria. Patients with microscopic tumor burden (<0.1 mm) had a CLND rate of 9.7% compared with 42.7% in those with tumor burden more than or equal to 0.1 mm (P< .001). Patients who underwent SLNB before MSLT-II implementation (see Figure 1) had a CLND rate of 68% compared with 28% after implementation (P < .001). There was no difference in treatment strategy with respect to number of metastasis-positive sentinel lymph nodes (Table 2).

Multivariate logistic regression analysis identified several patient and disease factors that influenced likelihood of immediate CLND. The factor with the largest effect was sentinel lymph node tumor burden, with patients whose sentinel node metastasis was more than 1mm being most likely to undergo CLND (odds ratio [OR], 17.25, P < .001). Additional factors with large effect sizes were sentinel node biopsy timing and primary tumor location. Before MSLT-II implementation, patients were more likely to undergo CLND (OR, 7.73, P< .001). Patients with head and neck tumors were more likely to undergo CLND compared with tumors of the upper extremity (OR, 5.22, P = .002); no difference was found comparing upper extremity to the trunk or lower extremity. While not significant in univariate analysis, BMI over 30 was a predictor of CLND after multivariate analysis (Table 3).

TABLE 3.

Multivariate logistic regression for odds of receiving completion lymph node dissection

	Odds ratio	95% Confidence interval	P value
Before MSLT-II implementationa	7.73	3.61–16.5	<.001
BMI > 30	2.01	1.04–3.87	.038
Comorbidity (any)	0.67	0.321–1.39	.278
Age	0.99	0.967–1.01	.222
Location
Upper extremity	Ref	Ref	Ref
Lower extremity	0.80	0.275–2.30	.675
Trunk	1.13	0.454–2.81	.793
Head and neck	5.22	1.80–15.1	.002
Sentinel lymph node tumor burden (Rotterdam criteria)
<0.1 mm	Ref	Ref	Ref
0.1–1 mm	9.15	2.08–40.2	.003
>1 mm	17.25	3.94–75.6	<.001

Abbreviation: CLND, completion lymph node dissection.

^aThe inflection point for implementation is based on an observation of trends in CLND by quarter after MSLT-II publication, Figure 1.

Of the 90 patients who had CLND, 17 (19%) were found to have positive nonsentinel lymph nodes. No patients with micrometastatic disease (Rotterdam < 0.1 mm) had a positive CLND. Multivariate logistic regression analysis identified two factors associated with CLND positivity: high SLN tumor burden (> 1 mm), which was also associated with CLND; and 3+ positive lymph nodes at SLNB, which was not associated with likelihood of CLND. Additional factors that were associated with CLND (timing of SLNB, BMI, and location in head and neck) were not significantly associated with CLND positivity (Table 4).

TABLE 4.

Multivariate logistic regression for odds of positive completion lymph node dissection

	Odds ratio	95% Confidence interval	P value
Before MSLT-II implementationa	0.72	0.182–2.85	.639
BMI >30	2.56	0.678–9.75	.167
Location, head and neck	0.33	0.060–1.80	.197
≥3 positive lymph nodes at SLNB	12.04	1.67–83.5	.013
Sentinel lymph node tumor burden > 1 mmb	25.71	2.54–231	.005

Abbreviations: BMI, body mass index; CLND, completion lymph node dissection; SLNB, sentinel lymph node biopsy.

^aThe inflection point for implementation is based on an observation of trends in CLND by quarter after MSLT-II publication, Figure 1.

^bNo CLND was positive for patients with sentinel node tumor burden < 0.1 mm.

4 |. DISCUSSION

Since the publication of MSLT-II, no study to our knowledge has analyzed the trend in rates of immediate CLND. Within our highvolume center, we identified a total of 235 consecutive SLNB+ patients treated immediately before, at the time of and in the short term following publication of MSLT-II. In line with emerging evidence and guidelines, overall rates of completion lymph node dissection decreased dramatically—from 67% in the year preceding MSLT-II, to 33% in the year following publication of MSLT-II, and finally to 26% in the second year following MSLT-II. However, adoption of guidelines favoring active surveillance was not uniform.

4.1 |. Sentinel lymph node tumor burden

With regard to disease-related factors, sentinel node burden was the most important factor determining likelihood of immediate CLND. Patients with higher sentinel node tumor burden by Rotterdam criteria were significantly more likely to undergo immediate completion lymph node dissection (OR, 8.04 for 0.1–1 mm and OR, 13.20 for > 1 mm diameter of largest sentinel node deposit) on multivariate logistic regression.³ Confidence intervals were wide due to a relatively low number of patients with micrometastatic disease (14% of patients) in the comparison group. No difference was seen upon analysis of Breslow depth, ulceration, number of positive sentinel nodes, or presentation at tumor board. While there was wide variation in Breslow depth and many tumors were ulcerated, there was little variation in number of positive sentinel nodes which limited comparison in this category.

4.2 |. Head and neck primary

Patients in our study with head and neck melanoma were also significantly more likely to undergo immediate CLND (OR, 4.29). Reasons for this finding are likely multifactorial and deserve further investigation, but five hypotheses are discussed here. First, as mentioned earlier, MSLT-II enrolled a much smaller number of head and neck subjects compared with other anatomic regions. Among providers who care for patients with head and neck melanoma, this has raised concerns about the generalizability of the study findings to this patient subset. Studies evaluating specifically patients with melanoma of the head and neck region have, though retrospective in nature, have nevertheless shown little to no difference in disease- specific survival after CLND.^12,13 Second, the head and neck region includes many crucial structures in close proximity,¹⁴ thus making a nodal recurrence potentially more morbid and technically more challenging.¹⁵ Third, lymphedema after modified radical neck dissection lymph node dissection for head and neck cancer is less common, making the procedure less morbid overall.^13,14 Although many studies do in fact report lymphedema in head and neck cancer patients, no robust studies looking at lymphedema-related quality of life in head and neck melanoma patients have been performed.¹⁶ Fourth, melanoma of the head and neck portends a worse prognosis compared with other sites.¹⁷ Finally, we note that at our institution the head and neck melanoma operations are performed by otolaryngologists and oral and maxillofacial surgeons rather than general surgery-trained surgical oncologists. It is possible that variation in training background could translate to variation in practice. A recent study in the Netherlands found that for rates of sentinel node biopsy, head and neck melanoma was the only outlier.¹⁸ In this study, patients with head and neck primaries actually received substantially fewer sentinel node biopsies compared with those with primaries in the trunk or extremities. This difference has not been previously explored and could be the focus of future investigation.

4.3 |. Body mass index

Regarding patient characteristics and selective adoption of various treatment strategies, we unexpectedly found that BMI > 30 was associated with a twofold higher likelihood of immediate CLND. The increased rate of CLND in patients with obesity has not been reported in prior literature. One possible explanation for this is perceived difficulty of ultrasound surveillance in patients with increased body habitus, though data are lacking. Although no prior studies have specifically examined the effect of obesity on outcomes after lymph node dissection for melanoma, multiple prior studies have demonstrated that obesity is a risk factor for lymphedema after axillary node dissection for breast cancer.¹⁹ While other studies within the breast cancer literature suggest that obese individuals are at higher risk for local recurrence, this has not been demonstrated in melanoma.²⁰ While patients with obesity are likely at increased risk for lymphedema, this remains a topic of continued investigation and it would be useful to examine the relationship between obesity and immediate CLND on a multi-institutional level. Other patient characteristics, including gender, race, insurance coverage, distance traveled, and presence of medical comorbidity, were not found to be significant contributing factors.

4.4 |. CLND positivity

High sentinel node tumor burden by Rotterdam criteria (> 1 mm) was associated both with increased odds of undergoing CLND and increased odds of a positive nonsentinel node on CLND on multivariate logistic regression. In addition, the number of positive sentinel lymph nodes, while not found to be significantly associated with CLND, was associated with CLND positivity. These findings should be interpreted with caution due to low sample size (17 positive CLNDs). To date, data regarding the predicative value of sentinel lymph node tumor burden have been mixed, with some studies showing an association.^21–23 These variations are likely due in part to variation in defining “high” or “low” tumor burden.²⁴ Our observation that overall tumor burden in the sentinel nodes is a significant predictor of positive CLND is hypothesis-generating. In addition to better informing the decision to pursue CLND or active surveillance, future research is also needed to determine whether the current one-size-fits-all active surveillance timeline should be modified based on sentinel node burden.

4.5 |. Strengths, limitations, and future directions

As a retrospective single-institution study, our analysis is limited. The strengths of our study include a large, consecutive series of patients with no dropouts or loss to follow up, precise patient data (complete medical and demographic records), and granular tumor characteristics (full pathology reports). While we await national data on the change in practice patterns for immediate CLND in SLNB+ melanoma, our highvolume center’s experience provides a window into early implementation of MSLT-II and DeCOG-SLT trial results. Future directions include studies to clarify which patients with high nodal burden and head and neck primary may be safely managed with active surveillance.

4.5.1 |. Implications and relevance

Rates of immediate CLND after positive SLNB for melanoma decreased dramatically, albeit not uniformly, at a high-volume center following the publication of MSLT-II. Factors that were associated with higher rates of CLND were increased sentinel node tumor burden according to Rotterdam criteria and primary tumor located in the head and neck. Patients with high nodal burden and head and neck primaries were underrepresented in MSLT-II, suggesting that clinicians are wary of applying active surveillance recommendations to perceived higher-risk groups. In addition, we unexpectedly found that BMI over 30 was associated with a higher rate of immediate CLND.