Sentinel Lymph Node Biopsy for T1b Melanoma: Balancing Prognostic Value and Cost

Abstract

Background.

The purpose of this study is to report the additional prognostic information and cost associated with sentinel lymph node biopsy (SLNB) for patients with T1b melanoma.

Patients and Methods.

An institutional database was queried for patients with T1b melanoma (0.8–1.0 mm or < 0.8 mm with ulceration) with at least 5 years of follow-up. Results of SLNB, completion lymphadenectomy (CLND), recurrence, and melanoma-specific survival (MSS) were assessed. Institutional costs of melanoma care were converted to Medicare proportional dollars (MP $). A Markov model was created to estimate long-term costs.

Results.

Among the total 392 patients, 238 underwent SLNB. Median follow-up was 10.5 years. SLNB was positive in 19 patients (8.0%). Patients who underwent SLNB had higher 10-year nodal recurrence-free survival (98.6% versus 91.2%, p < 0.001) but not MSS (94.4% versus 93.2%, p = 0.55). Ulceration (HR 4.7, p = 0.022) and positive sentinel node (HR 11.5, p < 0.001) were associated with worse MSS. Estimates for 5-year costs reflect a fourfold increase in total costs of care associated with SLNB. However, a treatment plan that forgoes adjuvant therapy for resected stage IIIA melanoma but offers systemic therapy for a node-basin recurrence would nullify the additional cost of SLNB.

Conclusions.

SLNB is prognostic for T1b melanoma. Its impact on the overall cost of melanoma care is intimately tied to systemic therapy in the adjuvant and recurrent settings.

INTRODUCTION

The incidence of melanoma is increasing in USA, with nearly 100,000 new diagnoses annually.¹ More than two-thirds of new cases are thin melanomas (T1, ≤ 1 mm). In the modern era, the anticipated survival for T1 melanomas is greater than 95% at 10 years.² However, a minority of these patients have nonlocalized disease. Within the American Joint Committee on Cancer (AJCC) 8^th edition staging system, the presence of microscopic lymph node metastases in T1 melanoma is indicative of stage IIIA disease, with a 10-year survival of roughly 88%. Given this prognosis, most patients with thin melanoma and sentinel node disease are not offered adjuvant therapy at our center.³

Sentinel lymph node biopsy (SLNB) has been a mainstay of melanoma staging for more than 20 years.⁴ The prognostic value of the sentinel node as a predictor of distant recurrence and survival has been consistently demonstrated.⁵ The potential therapeutic value of SLNB was evaluated in the first Multicenter Selective Lymphadenectomy Trial (MSLT-1). Despite retrospective series that have suggested an impact of SLN biopsy on survival,^6-8 there was no evidence of improved melanoma-specific survival (MSS) in this trial. Nevertheless, recent guidelines from the American Society of Clinical Oncology and the Society of Surgical Oncology recommend SLNB for patients with intermediate-thickness melanoma (1.1–4.0 mm), as the result of the sentinel node biopsy in these patients may guide adjuvant therapy. Patients with T1b melanoma (0.8–1.0 mm thickness or < 0.8 mm with ulceration) may also be considered for SLNB,⁹ although its role in this low-risk group is controversial. The cost-effectiveness of SLNB also remains unclear.^10-12 Importantly, an estimation of the cost impact of SLNB must include the elevated surveillance regimen affiliated with uncovering a positive sentinel node.

The likelihood of finding metastatic disease in patients with thin melanoma is about 5%, with higher rates for tumors that are ulcerated or ≥ 0.75 mm in Breslow depth.^13,14 Although patients with thin melanoma who have a positive sentinel node are upstaged to IIIA, many retrospective series are too underpowered to exhibit the prognostic value of SLNB in this population.^15,16 While less invasive than a complete lymph node dissection (CLND), SLNB is associated with complications such as infection and lymphedema.¹⁷

Given that the removal of microscopic node metastases has not been shown to improve long-term survival,^18,19 the key question becomes whether the prognostic information gained from SLNB justifies its associated costs. The objective of this study is to determine whether the status of the sentinel lymph node is prognostic in patients with T1b melanoma and to estimate short- and long-term costs of the procedure.

PATIENTS AND METHODS

Patient Selection

Given that the therapeutic role of SLNB remains contentious, a classic cost-effectiveness approach is not justified. Rather, SLNB’s prognostic value was analyzed via a retrospective approach, and its cost implications were assessed through institutional data and long-term model projections. A prospectively maintained database was queried for patients with T1b melanoma of the trunk or extremities based on AJCC 8^th edition staging: Breslow depth 0.8–1.0 mm or < 0.8 mm with ulceration. Given that stage IIIA patients are associated with a 5-year expected survival of more than 90%,² assessment of the prognostic value of SLNB in T1b melanoma has poor validity without long-term follow-up. Therefore, the query was limited to patients diagnosed earlier than 2013. Patients meeting any of the following criteria were excluded: (1) second primary melanoma > 1 mm in Breslow depth, (2) stage IIIB–IV disease at the time of diagnosis, and (3) head and neck or ocular primary melanoma. We further excluded patients who did not receive their index melanoma treatment or consultation at our institution to minimize referral biases that select for recurrent or advanced cases.

The following characteristics were collected for each patient: age at diagnosis, date of diagnosis, and sex. The following primary tumor features were queried: anatomic site, Breslow depth, ulceration, mitotic rate, tumor infiltrating lymphocytes, regression, lymphovascular invasion, and perineural invasion. Treatment variables included sentinel node positivity, nonsentinel node positivity (when a completion node dissection was performed), and use of adjuvant systemic therapy. For each patient, the date and site of melanoma recurrence and the date and cause of mortality were assessed.

Survival Analysis

The primary outcome of interest was the relationship between sentinel node positivity and MSS. Secondary outcomes included node basin recurrence-free survival (NFS) and distant metastasis-free survival (DMFS). Risk factors associated with MSS were identified through univariate analysis using Kaplan–Meier curves and the log-rank test. Risk factors associated with MSS at the univariate level with p-value < 0.10 were incorporated into the multivariate regression through Cox proportional hazards modeling. Statistical analyses were performed using SPSS 25.0 (Armonk, NY).

Short-Term Costs

Costs of melanoma-related care, including surgery, clinic visits, imaging, and medications, are potentially impacted by SLNB. To assess the downstream cost impact of SLNB, institutional costs for each patient’s inpatient and outpatient care for 1 year following the date of diagnosis were determined via itemized costs from our hospital accounting database. Costs were restricted to those associated with melanoma diagnosis-related group codes. Both direct and indirect costs were included. To create more generalizable estimates, institutional costs were converted to Medicare reimbursement levels by matching each patient to the amount Medicare would reimburse for their care.²⁰ The ratios of the costs to Medicare reimbursement were based on US mean diagnosis-related group level reimbursements for inpatient admissions and the relative value scale-based reimbursements for outpatient visits. These conversion factors were then applied to all component costs for each patient. The cost estimates reported are not specific to a single institution but reflect base Medicare reimbursement. All costs were estimated based on 2012 US dollars using 2012 costs per unit for all services. Estimates were rounded to the nearest $10 and reported as Medicare proportional dollars (MP $). Given that the process of converting to Medicare reimbursement is unreliable with institutional records prior to 2005, only patients with melanoma diagnoses from 2005 to 2013 were included in the short-term cost analyses.

Long-Term Costs

Many patients who receive treatment for melanoma at our institution undergo long-term surveillance at local hospitals. For these patients, clinical outcomes are maintained through periodic correspondence. However, long-term costs of care are not adequately captured through the institutional accounting database. Because of this, a Markov cost-analysis model was created to estimate long-term costs of surveillance and treatment. Markov models are mathematical simulations in which patients transition between health states depending on potential outcomes are represented in the model. Passage of time is represented by Markov cycles (in this study, 1 year). Each Markov cycle is characterized by a disease state (i.e., disease-free survival), which is associated with a set of expected costs; For example, the costs relevant to disease-free survival include costs of surveillance imaging and clinic follow-up. Likelihood of changing between disease states within a given Markov cycle is represented by transition probabilities (i.e., node basin recurrence rate).

A Markov model was created to calculate the anticipated 5-year costs of care following a melanoma diagnosis. Transition probabilities were derived from the study population’s real-world outcomes. The following cost-related assumptions were made: wide local excision (WLE) without SLNB is performed in an outpatient clinic setting, WLE with SLNB is performed in a hospital facility, and patients with node basin recurrence undergo therapeutic lymph node dissection in a hospital facility. Patients with a positive sentinel node do not undergo completion lymph node dissection but receive escalated staging and surveillance (Fig. 1). Costs of clinic visits, surveillance imaging, and surgery were acquired from facility limiting charges to Medicare (www.cms.gov).

FIG. 1.

Markov model design and surveillance protocol. WLE wide local excision, SLNB sentinel lymph node biopsy, TLND therapeutic lymph node dissection

By using these parameters, we created a Markov model that simulates the study cohort’s outcomes over time through an exponential survival function and calculates the aggregate costs associated with these outcomes. Given that a Markov model is an iterative process, it is most suitable for estimating an outcome that accrues over time, such as cost; For example, patients within the cohort who recur in the node basin within the first year—who would then undergo therapeutic node dissection—would no longer accrue costs of node basin surveillance over the remaining years. Given that the patient population represents a treatment period preceding PD-1 antibody therapy, the baseline cost model did not incorporate PD-1 antibody in the treatment of melanoma. Additional cost projections were created to accommodate pembrolizumab in the adjuvant or recurrent setting. For these analyses, the regimen for pembrolizumab was based upon the KEYNOTE-054 trial.²¹ Since removal of microscopic node metastases was not anticipated to impact the rate of DMFS,^18,19 the model did not incorporate distant metastatic recurrence into costs associated with SLNB. Model analyses were performed using TreeAge Pro 2017 (TreeAge Software, Inc, Williamstown, MA).

RESULTS

Patient Characteristics

Between 1982 and 2013, 392 patients with T1b melanoma received their index treatment at our institution. Among these, 238 underwent SLNB, and 154 were treated with WLE only. The average age was 55.6 years. Patient demographics were comparable between those who did and those who did not undergo SLNB. Patients who underwent SLNB had thicker tumors (0.91 versus 0.82 mm, p < 0.001), were less likely to have an ulcerated tumor (10.1% versus 17.5%, p = 0.045), and had shorter follow-up (11.0 versus 13.4 years, p < 0.001) (Table 1). Among the 238 patients who underwent SLNB, 19 were found to have at least one positive sentinel node (8.0%). Fourteen of these patients underwent CLND, none of whom were found to have positive nonsentinel nodes.

TABLE 1.

Patient demographics and patterns of recurrence

	WLE	WLE + SLNB
	(n = 154)	(n = 238)	p-Value
Age, years (SD)	56.1 (15.9)	54.6 (15.7)	0.573
Female (%)	72 (46.8)	121 (50.8)	0.469
Follow-up, years (SD)	13.4 (6.3)	11.0 (4.6)	< 0.001
Breslow thickness, mm (SD)	0.82 (0.14)	0.91 (0.08)	< 0.001
Ulceration (%)	27 (17.5)	24 (10.1)	0.045
Mitosis identified (%)	104 (67%)	165 (69%)	0.708
SLNB + (%)		19 (8.0)
CLND (% of SLN+)		14 (73.7)
Nonsentinel node + (%)		0 (0)
Recurrence (%)
Local or in-transit only	5 (3.2)	3 (1.3)	0.272
Regional node only	14 (9.1)	2 (0.8)	< 0.001
Distant only	3 (1.9)	11 (4.6)	0.264
Regional node + distant	1 (0.6)	1 (0.4)	1.000
Distant total	12 (7.8)	18 (7.6)	1.000

WLE wide local excision, SLNB sentinel lymph node biopsy, SD standard deviation, CLND completion lymph node dissection

Recurrence Patterns

Among the 238 patients who underwent SLNB, 22 developed recurrence at any site (9.2%): 3 recurred locally or in-transit, 2 recurred initially in the node basin alone (0.8%), 11 recurred initially with distant metastases (4.6%), and 1 recurred simultaneously in the node basin and at a distant site (0.4%). All node basin recurrences occurred in patients who were sentinel node negative. Of the 154 patients who did not have a SLN biopsy, 24 patients recurred (15.6%): 5 recurred locally or in-transit, 14 recurred initially in the node basin alone (9.1%), 3 recurred initially with distant metastases (1.9%), and 1 recurred at both sites simultaneously (0.6%) (Table 1). At last follow-up, the overall rate of distant metastasis was not different between groups (7.8% for WLE versus 7.6% for WLE with SLNB, p = 1.0).

Patients who underwent WLE with SLNB experienced greater 10-year NFS (98.6% versus 91.2%, p < 0.001) (Fig. 2A). However, SLNB was not associated with distant metastases, with a 10-year DMFS of 91.4% versus 93.0% (p = .957) (Fig. 2B). Among patients who underwent SLNB, those who were sentinel node positive trended toward worse 10-year DMFS (88.5% versus 91.6%, p = 0.079) (Fig. 2C).

FIG. 2.

Recurrence-free survival: (A) node basin recurrence-free survival comparing wide local excision with and without sentinel node biopsy, (B) distant metastasis-free survival comparing wide local excision with and without sentinel node biopsy, (C) distant metastasis-free survival comparing sentinel node positive and negative. WLE wide local excision, SLNB sentinel lymph node biopsy

Survival Outcomes

Melanoma-specific survival among patients who underwent WLE with SLNB was not significantly different from that of those managed with WLE alone (10-year MSS 94.4% versus 93.2%, p = 0.547) (Fig. 3A). Among patients who underwent SLNB, the finding of one or more positive sentinel nodes was associated with significantly worse 10-year MSS (77.0% versus 95.8%, p = 0.001, Fig. 3B). Melanoma-specific survival was not different between patients who were sentinel node positive versus those who did not undergo SLNB but later developed a node basin recurrence (10-year MSS 77.6% versus 64.6%, p = 0.550) (Fig. 3C). Univariate analyses indicated that age, Breslow depth, ulceration, and sentinel node positivity were associated with worse MSS. On multivariate analysis, only ulceration (HR 4.7, p = 0.035) and sentinel node positivity (HR11.5, p < 0.001) were associated with worse MSS.

FIG. 3.

Melanoma-specific survival: (A) comparing wide local excision with and without sentinel node biopsy, (B) comparing sentinel node positive and negative, (C) comparing sentinel node positive and patients who developed node basin recurrence. WLE wide local excision, SLNB sentinel lymph node biopsy

Cost Comparisons

Medicare-proportional cost data were available for 178 patients between 2005 and 2013, of whom 125 underwent SLNB. The median 1-year cost of melanoma-related care was MP $7300 for WLE with SLNB, and MP $2230 for WLE only (p < 0.001). Among patients who underwent SLNB, those with a positive sentinel node accrued significantly greater costs than node-negative patients (MP $10,800 versus MP $6720, p < 0.001).

A Markov decision-analysis model replicated the node basin recurrence and survival patterns within our patient population (Supplementary Materials). The baseline model is reflective of costs of surgery, clinical follow-up, and surveillance. Estimates for 5-year cost of treatment were $2480 for WLE alone and $10,220 for WLE with SLNB ($8650 for SLN-negative and $27,390 for SLN-positive). This translated into a cost of $96,750 per patient upstaged. Costs associated with treatment paradigms that incorporate adjuvant PD-1 antibody for sentinel node-positive disease or for node basin recurrence are presented in Table 2. These represent the estimated costs of a patient who presents with T1b melanoma, incorporating the probability of SLN positivity or node basin recurrence for each patient. Adoption of a strategy of routine adjuvant pembrolizumab for all sentinel node-positive patients elevated average projected 5-year costs to over $25,000 for each patient managed with WLE and SLNB. However, omitting pembrolizumab in the adjuvant setting and incorporating it routinely for node-basin recurrence results in near equivalence between WLE and WLE with SLNB. This was due to the higher rate of node-basin recurrence associated with omitting SLNB.

TABLE 2.

Estimated cumulative 5-year cost per patient of melanoma-related treatment

	WLE only	WLE + SLNB
Baseline model (no systemic therapy)	$2480	$10,220
PD-1 antibody* for SLN+ patients	$2480	$25,300
PD-1 antibody* for node basin recurrence	$12,210	$11,280
PD-1 antibody* for SLN+ or for node basin recurrence	$12,210	$26,360

^*Cost of PD-1 antibody based on pembrolizumab dosed Q3 weeks for 1 year

DISCUSSION

Using a large single-institution clinical database, this study demonstrated that SLNB has prognostic value for patients with T1b melanoma through a clear association between sentinel node status and MSS. These findings validate the prognostic role of SLNB in the staging of patients with T1b melanoma, among whom 8.0% may be expected to harbor one or more positive sentinel nodes. However, SLNB does not serve a therapeutic role in relation to MSS or DMFS and is associated with substantial costs that extend beyond the immediate postoperative period.

The incidence of sentinel node positivity among patients with melanoma ≤ 1 mm thick is approximately 5%.^22-25 However, because this rate is based on surgical cohorts, the true prevalence of nodal metastases in this population may be lower. Identifying patients who are more likely to harbor node metastases is important because SLNB carries a complication rate of up to 11%, including a lymphedema risk of 1–10%.^17,26 In a metaanalysis encompassing 60 studies, thickness ≥ 0.75 mm was associated with an 8.8% likelihood of sentinel node metastasis.¹³ A recent analysis of the National Cancer Database revealed that Clark level, lymphovascular invasion, ulceration, and mitotic rate are also predictive of occult node metastases in thin melanoma.²⁷ These findings were consistent with our sentinel node metastasis rate of 8.0% for T1b melanoma.

There are discordant reports regarding the association between sentinel node metastasis and survival in thin melanoma. Sentinel node positivity elevates T1b melanoma from stage IB to IIIA.² In two large retrospective cohorts, sentinel node status was associated with both disease-free and overall survival.^22,28 These findings were later corroborated by a multiinstitutional database.²³ However, several smaller series found no association between sentinel node status and MSS.^16,29-31 Given that patients with stage IIIA disease may experience a 10-year survival of 88%,² prognostic assessments in this population require adequate power and extended follow-up. To our knowledge, this is the first report of patients with T1b melanoma with median follow-up greater than 10 years. Our results strongly indicate that SLNB does possess prognostic value in this population.

The MSLT-2 trial demonstrated that CLND in the setting of a positive sentinel node conferred no survival benefit despite the presence of nonsentinel node metastases in 11.5% of patients.¹⁸ However, CLND did increase the likelihood of disease control in the regional node basin. Our results indicate that, for T1b melanoma, CLND has an exceedingly low likelihood of yielding additional metastases and therefore is unlikely to improve nodal basin recurrence. For these patients, the SLNB procedure itself may provide durable node basin control. MSLT-2 also implies that removal of microscopic nodal metastases has no impact on DMFS or MSS. Our results corroborate these findings.

As a staging tool, SLNB is relatively costly. While there are measures that may reduce the procedural cost of SLNB, such as performing the surgery under tumescent local anesthesia,³² such measures do not affect the long-term costs of surveillance and follow-up. Wilson et al.³³ found SLNB to be cost-effective when used to guide the use of adjuvant interferon, however, the principle outcome was not MSS but relapse-free survival, a metric for which utility values are ill defined. With a focus on thin melanomas, Agnese et al.¹¹ reported a cost ratio comparing SLNB with WLE alone that was similar to that of our 1-year costs. The authors suggested that, even if the removal of a positive sentinel node would prevent a melanoma-related mortality, such an intervention would have an exceedingly high cost per life saved. Within our study, early removal of node metastases was not associated with better MSS. Morton et al.¹⁰ used transition states derived from MSLT-1 to generate a Markov cost-effectiveness model for SLNB in intermediate-thickness melanoma. Sentinel node biopsy was cost-effective under the assumption that early CLND improve survival. However, MSLT-2 has since demonstrated no survival advantage for CLND in sentinel node-positive patients.

Given that SLNB does not have a therapeutic impact on MSS for thin melanoma, we believe that a traditional cost-effectiveness study is not possible. Nevertheless, our report of the long-term costs of SLNB along with its prognostic relevancy can provide a better perspective on the value of the procedure. Importantly, the cost implications of SLNB in this population are related to systemic treatment strategies. At our institution, adjuvant therapy is not routine for resected stage IIIA patients. For node basin recurrence, checkpoint blockade or targeted therapy is considered on a case-by-case basis. This approach was adopted because there has been no high-level evidence to suggest that systemic therapy in the adjuvant setting confers a significant survival advantage compared with delayed treatment in the recurrent setting. Furthermore, adjuvant anti-PD-1 therapy has not been shown to improve recurrence-free or overall survival for stage IIIA patients.²¹ Because SLNB is associated with lower node basin recurrence, this may provide long-term cost savings if such recurrences are routinely treated with systemic therapy. In our model, adopting a strategy of universal pembrolizumab for patients with node basin recurrence results in near-equivalent cost projections between WLE and WLE with SLNB. However, not all patients are candidates for this treatment, and only about half of patients who start PD-1 therapy complete the full course.²¹ Given that there are not adequate data on patient-reported utilities while on and off systemic therapy for melanoma, it is difficult to convert results from this study to cost per quality-adjusted life year. For these reasons, further research is necessary before the cost of SLNB can be compared against a societal willingness-to-pay threshold to establish true cost-effectiveness.

This study has several limitations. As a single-institution cohort from a tertiary referral cancer center, our dataset is susceptible to referral biases. For this reason, we excluded patients who were referred for recurrent disease. We modeled the surveillance protocol for sentinel node-positive patients based on guidelines from the National Comprehensive Cancer Network and the study protocol for the observation arm of MSLT-2.¹⁸ Therefore, the financial projections of the Markov cost model are generalizable outside of our institution. However, the cost findings in this study do not account for surveillance attrition, which can vary across different medical centers. Thus, cost-related projections for sentinel node-positive patients may overestimate real-world expenditures beyond 2–3 years. There were several patients for whom death from melanoma was not preceded by a detailed record of recurrence. This may account for the lack of a significant association between sentinel node metastases and DMFS. Finally, our 5-year cost projections do not account for complications related to SLNB, although these are typically mild.

CONCLUSIONS

Patients with T1b melanoma represent an overall low-risk population, but a subset does have sentinel node metastases. The sentinel node status holds significant prognostic value in this cohort—a finding that only manifests with extended follow-up. This prognostic value must be weighed against costs that extend beyond the perioperative period. Under most treatment algorithms, the use of SLNB in this population increases the long-term costs of melanoma-related care by 2–10-fold. However, SLNB may be cost-neutral within a treatment paradigm that omits upfront adjuvant therapy and reserves PD-1 antibody for the recurrent setting.

Supplementary Material

1729671_Supp_Fig

SUPPLEMENTARY FIG. 1 Comparison of node basin recurrence-free survival among patients who underwent wide local excision (WLE) only, according to the Markov decision model versus institutional cohort outcomes

SUPPLEMENTARY FIG. 2 Comparison of melanoma-specific survival among sentinel lymph node positive (SLN+) patients, according to the Markov decision model versus institutional cohort outcomes

SYNOPSIS.

Sentinel node biopsy for T1b melanoma (0.8–1.0 mm or ulcerated) remains controversial. The prognostic value of the sentinel node in this relatively low-risk population is demonstrated through a longitudinal clinical database. To counterbalance this value, estimates of the long-term cumulative costs of sentinel node biopsy are assessed through real-world patient-level institutional data and through a decision-analysis mathematical model. Results show that, while sentinel node biopsy holds useful prognostic information, this increases short-term melanoma management costs by roughly fourfold.