Cumulative Incidence and Predictors of CNS Metastasis for Patients With American Joint Committee on Cancer 8th Edition Stage III Melanoma

Abstract

PURPOSE

Improved understanding of the incidence, risk factors, and timing of CNS metastasis is needed to inform surveillance strategies for patients with melanoma.

PATIENTS AND METHODS

Clinical data were extracted from the databases of 2 major melanoma centers in the United States and Australia for 1,918 patients with American Joint Committee on Cancer (AJCC) 8th edition stage III melanoma, diagnosed from 1998-2014, who had (negative) baseline CNS imaging within 4 months of diagnosis. The cumulative incidence of CNS metastasis was calculated in the presence of the competing risk of death, from stage III presentation and at benchmark time points 1, 2, and 5 years postdiagnosis.

RESULTS

At a median follow-up of 70.2 months, distant recurrence occurred in 711 patients (37.1%). The first site of distant metastasis was CNS only for 3.9% of patients, CNS and extracranial (EC) for 1.8%, and EC only for 31.4%. Overall, 16.7% of patients were diagnosed with CNS metastasis during follow-up. The cumulative incidence of CNS metastasis was 3.6% (95% CI, 2.9% to 4.6%) at 1 year, 9.6% (95% CI, 8.3% to 11.0%) at 2 years, and 15.8% (95% CI, 14.1% to 17.6%) at 5 years. The risk of CNS metastasis was significantly influenced by patient sex, age, AJCC stage, primary tumor site, and primary tumor mitotic rate in multivariable and conditional analyses. High primary tumor mitotic rate was significantly associated with increased risk of CNS metastasis at diagnosis and all subsequent time points examined.

CONCLUSION

Similar rates of CNS metastasis were observed in 2 large, geographically distinct cohorts of patients with stage III melanoma. The results highlight the importance of primary tumor mitotic rate. Furthermore, they provide a framework for developing evidence-based surveillance strategies and evaluating the impact of contemporary adjuvant therapies on the risk of CNS metastasis development.

INTRODUCTION

In most Western populations, the most common sources of CNS metastases from cancer, defined as metastases in the brain or leptomeninges, are lung cancer, breast cancer, and melanoma. Although melanoma is less common than lung or breast cancer, it has the highest risk of CNS metastasis among common solid tumors.¹ Historically, the median survival of patients with melanoma with CNS metastasis has been approximately 4 months (range, 2-7 months).^2-4 However, the introduction of more effective systemic therapies (ie, immunotherapies, targeted therapies) and surgical and radiation therapies (eg, stereotactic radiosurgery [SRS]) is improving outcomes for these patients.^5-7 Immunotherapies have demonstrated greatest efficacy in patients with small, asymptomatic metastases, whereas SRS is most effective for tumors < 3 cm in diameter.^8-10 Thus, there is a strong rationale to identify patients with CNS metastases before they become large and/or symptomatic. An improved understanding of the incidence, timing, and risk factors for CNS metastasis among at-risk patients with melanoma is critical to developing evidence-based guidelines for CNS surveillance.

Population-based studies using the SEER database show that a low proportion of patients with cutaneous melanoma overall will be diagnosed with CNS metastasis. Kromer et al¹¹ reported a 1.2% incidence of brain metastases (BM) among 79,785 patients diagnosed from 2010 to 2013, and Zhang et al¹² reported 1.3% among 116,119 patients diagnosed from 2010 to 2015 at initial presentation of melanoma. However, the risk of developing CNS metastasis among patients with regionally metastatic disease (stage III), the most common pattern of progression from clinically localized melanoma, is higher. Among 402 patients with American Joint Committee (AJCC) 7th edition high-risk stage III (IIIA[N2a]-IIIC) melanoma enrolled in the S0008 clinical trial of adjuvant biochemotherapy or high-dose interferon (HDI), 15% of patients (n = 59) relapsed in the CNS during follow-up, and the CNS was the initial site of relapse for 9% of patients (n = 34).¹³ Among patients with stages I-III melanoma who had recurrence, the incidence of CNS metastasis was reported as 19.8% (n = 607).¹⁴

Although the aforementioned studies demonstrated a consistent and significant risk of CNS metastasis for patients with stage III melanoma, each of the cohorts examined was relatively small, and the analyses used the AJCC 7th edition staging system and varied in their inclusion criteria and statistical methodologies.¹⁵ Furthermore, although current National Comprehensive Cancer Network (NCCN) guidelines for radiographic imaging of patients with stage III melanoma state, “consider imaging for baseline staging and to evaluate specific signs or symptoms,” and for patients with AJCC 8^th edition stage IIIC and higher melanoma, “consider including baseline brain MRI [magnetic resonance imaging] in asymptomatic patients,” those guidelines are based on low-level evidence and NCCN working group consensus.^{16(pME-4, ME-D1)} Thus, to facilitate the development of contemporary, evidence-based guidelines, we analyzed a large, multi-institution cohort of patients with AJCC 8th edition stage III melanoma to determine the incidence, timing, and risk factors associated with the development of CNS metastasis.¹⁷

PATIENTS AND METHODS

Patient Selection

Clinicopathologic data were extracted from the melanoma clinical research databases of The University of Texas MD Anderson Cancer Center (MD Anderson) and Melanoma Institute Australia (MIA). The overall cohort consisted of patients aged 16 years or older whose initial presentation was between 1998 and 2014 and who had AJCC 8th edition stage III melanoma arising from either an identifiable but previously untreated primary cutaneous tumor or an unknown primary site, with sufficient information to determine pathologic stage group (IIIA, IIIB, IIIC, or IIID). Patients with mucosal or uveal melanoma were excluded. To be included, patients also had to have negative baseline CNS imaging, including computed tomography (CT) and/or MRI of the brain, and/or positron emission tomography/CT of the whole body, within 4 months of diagnosis.

Statistical Analysis

Comparisons of categorical and continuous covariables were conducted using the χ² test and the Kruskal-Wallis test, respectively. Time to CNS metastasis, including parenchymal BM or leptomeningeal disease, was calculated from the date of initial presentation with melanoma to the date of last known vital status. The cumulative incidence of CNS metastasis was assessed with the Fine and Gray model in the presence of the competing risk of death from any cause, using a threshold of significance of .05. Multivariable models were constructed a priori with prespecified predictors to include age, patient sex, AJCC 8th edition stage group, and primary tumor factors. Separate models were constructed for primary site and mitotic rate to allow inclusion of the “melanoma of unknown primary” cohort without redundant adjustment. Extranodal extension was not included in multivariable models because of significant (> 10%) missing data. Institution was included in all multivariable models to adjust for potential intercenter heterogeneity.

RESULTS

Overall Cohort

Data for a total of 1,918 patients were extracted from the melanoma research databases of MD Anderson (50.6%) and MIA (49.4%; Table 1). Median overall follow-up for patients alive at last follow-up was 70.2 months (range, 0.6-237.0 months). The median age of the overall cohort was 56 years (range, 16-95 years), and the majority were male (64.8%). The distribution of AJCC 8th edition stage group across the overall cohort was 22.2% for IIIA, 28.8% for IIIB, 44.7% for IIIC, and 4.4% for IIID. A total of 711 patients (37.1%) experienced recurrence at distant sites, and 321 (16.7%) were diagnosed with CNS metastasis during follow-up. The first site of distant metastasis was CNS only for 3.9% (75/1,918), CNS and extracranial (EC) for 1.8% (34/1,918; ie, diagnosed within 7 days of each other), and EC only for 31.4% (602/1,918). Of the 109 patients whose first presentation of distant metastasis included CNS site(s), 46 (42.2%) were asymptomatic, 50 (45.9%) presented with CNS symptoms, 7 (6.4%) presented with non-CNS symptoms, and 6 (5.5%) did not have sufficient documentation for categorization (Appendix Table A1, online only).

TABLE 1.

Clinicopathologic Factors and Univariable Statistics at Diagnosis With Stage III Melanoma by Institution

Comparison of MD Anderson and MIA Cohorts

Compared with MD Anderson patients, those treated at MIA were older, more likely to present with stage IIIB or IIIC, and less likely to present with stage IIIA or IIID melanoma; they also had a higher proportion of patients with melanoma of unknown primary. Primary tumors from patients at MD Anderson were thicker and had higher mitotic rates (Table 1). MIA and MD Anderson cohorts also underwent different patterns of brain imaging surveillance. MD Anderson patients had more brain imaging events overall during follow-up (median, 6; range, 1-93) compared with MIA (median, 4; range, 1-39; P < .001); brain imaging was primarily MRI at MD Anderson (89%) and primarily CT (90%) at MIA (Appendix Table A1).

Univariable Analysis of Predictors of CNS Metastasis

Univariable cumulative incidence of CNS metastasis for the study cohort (N = 1,918) was 3.6% at 1 year (95% CI, 2.9% to 4.6%), 9.6% (95% CI, 8.3% to 11.0%) at 2 years, and 15.8% (95% CI, 14.1% to 17.6%) at 5 years (Fig 1A), and was not different between the 2 institutions (P = .303). The cumulative incidence was higher for males (5 years, 18.1%; 95% CI, 15.9% to 20.5%) versus females (11.6%; 95% CI, 9.2% to 14.3%; P < .001; Fig 1B). Increasing stage group of disease (IIIA→IIIB→IIIC→IIID) was associated with higher cumulative incidence of CNS metastasis (P < .001; Fig 1C), with a 5-year risk of 6.5% (95% CI, 4.3% to 9.3%), 14.1% (95% CI, 11.1% to 17.4%), 20.4% (95% CI, 17.6% to 23.3%), and 29.4% (95% CI, 19.4% to 40.2%) for patients with stage IIIA, IIIB, IIIC, IIID disease, respectively. Primary tumor sites (P < .001; Fig 1D) had significantly different rates of CNS metastasis, with scalp conveying the highest risk (5 years, 29.0%; 95% CI, 20.6% to 37.9%). The cumulative CNS metastasis incidence was 16.7% (95% CI, 12.3% to 21.7%) for patients with unknown primary melanoma. Melanoma subtypes had significantly different risks of CNS metastasis, with acral subtype conveying the greatest risk (P = .041; Fig 1E). The presence of extranodal extension of lymph node metastasis was also associated with increased risk of CNS metastasis, compared with absence of this feature (P < .001; Fig 1F).

FIG 1.

Univariable cumulative incidence of CNS metastasis (A) for the overall cohort, and stratified by (B) patient sex (P < .001), (C) American Joint Committee on Cancer 8th edition stage group (P < .001), (D) primary tumor site (P < .001), (E) melanoma subtype (other group included desmoplastic, lentigo maligna, blue nevus-like, nevoid, polypoid, spindle cell, and spitzoid/nevoid subtypes; P = .041), and (F) extranodal extension (P < .001) (G) mitotic rate groups (< 0.001). SSM, superficial spreading melanoma.

Primary tumor mitotic rate was significantly associated with cumulative incidence of CNS metastasis when assessed either as a continuous (hazard ratio [HR],1.04; 95% CI, 1.03 to 1.54; P < .001) or categorical covariate. Increasing mitotic rate, grouped as 0-4 per mm² (low), 5-9 per mm² (moderate), and > 9 per mm² (high), predicted a higher cumulative incidence of CNS metastasis (P < .001; Fig 1G). The cumulative 5-year incidence was 9.1% (95% CI, 7.0% to 11.4%) for low, 18.8% (95% CI, 15.0% to 22.9%) for moderate, and 24.7% (95% CI, 20.5% to 29.2%) for high mitotic rate.

Multivariable Analysis of Predictors of CNS and EC Metastasis

Multivariable cumulative incidence of CNS metastasis was significantly increased for males (HR, 1.53; 95% CI, 1.18 to 1.99; P = .001), decreasing age at diagnosis (per 10-year increments: HR, 0.90; 95% CI, 0.83 to 0.97; P = .004), increasing AJCC 8th edition stage group (reference, IIIA; IIIB: HR, 2.07; 95% CI, 1.35 to 3.17; IIIC: HR, 2.46; 95% CI, 1.65 to 3.67; IIID: HR, 3.17; 95% CI, 1.75 to 5.74; P < .001), and mitotic rate groups (Table 2, model 1). Adjusted curves of the cumulative incidence function estimates for CNS metastasis illustrate the increased risk associated with male sex and increasing mitotic rate, stratified by AJCC stage group (Fig 2).

TABLE 2.

Multivariable Cumulative Incidence of CNS Metastasis From Stage III Diagnosis: Model 1 With Mitotic Rate Groups and Model 2 With Primary Tumor Site Groups

FIG 2.

Adjusted estimates of cumulative incidence of CNS metastasis by American Joint Committee on Cancer 8th edition stage group with baseline risk profile of (A) female and (B) male patients with mitotic rate of 0-4 per mm², (C) female and (D) male patients with mitotic rate of 5-9 per mm², and (E) female and (F) male patients with a mitotic rate of > 9 per mm².

A separate multivariable model was constructed adjusting for primary tumor site (Table 2, model 2). When compared with scalp, a significantly lower risk of CNS metastasis was observed for all other primary tumor sites, including head (nonscalp) and neck (HR, 0.58; 95% CI, 0.35 to 0.95; P = .031), upper extremity (HR, 0.39; 95% CI, 0.23 to 0.65; P < .001), trunk (HR, 0.63; 0.43 to 0.92; P = .016), lower extremity (HR, 0.50; 95% CI, 0.33 to 0.75; P < .001), and unknown primary (HR, 0.51; 95% CI, 0.32 to 0.81; P = .005). Finally, a model including subtype only demonstrated a significantly higher risk of CNS metastasis for acral compared with other melanoma subtypes (P = .015; data not shown).

Multivariable models were also constructed to assess the cumulative incidence of EC metastasis (Appendix Table A2, online only). Compared with the CNS model, the same factors were significant in the model, including mitotic rate groups. In the EC model including primary site, the only difference from the CNS model was that scalp versus head (nonscalp) and neck sites did not have significantly different risk of EC metastasis.

Conditional Cumulative Incidence of CNS Metastasis

Patients were selected for dynamic assessment from time points after initial stage III melanoma presentation on the condition of surviving 1, 2, and 5 years without CNS metastasis. To examine the probability of developing CNS metastasis in patients who had survived for those intervals without any CNS metastasis, conditional univariable cumulative incidence of CNS metastasis curves from each follow-up time point were constructed (Fig 3). The 5-year cumulative incidence of CNS metastasis in these patients starting from time points 1, 2, and 5 years after initial stage III presentation was 14.5% (95% CI, 12.8% to 16.4%), 10.8% (95% CI, 9.1% to 12.7%), and 7.4% (95% CI, 5.3% to 9.8%), respectively. Conditional univariable cumulative incidence of EC metastasis curves were also constructed (Appendix Figure A1, online only).

FIG 3.

Conditional, univariable cumulative incidence of CNS metastasis at time points (A) 1 year, (B) 2 years, and (C) 5 years after initial melanoma diagnosis.

In the conditional multivariable models including mitotic rate groups, at 1 and 2 years after the diagnosis of stage III melanoma without CNS recurrence, the subsequent cumulative incidence of CNS metastasis continued to be significantly higher for male patients with HRs of 1.64 (95% CI, 1.22 to 2.21; P = .001) and 1.86 (95% CI, 1.24 to 2.77; P = .002), respectively (Table 3). Age at diagnosis remained significant in conditional models at 1 year, but not at 2 or 5 years after diagnosis. With the exception of IIID (perhaps because of the small number of patients with IIID disease), AJCC 8th edition stage groups remained significant at 1 and 2, but not at 5, years. Compared with primary tumors with low (0-4 per mm²) mitotic rate, primary tumors with moderate (5-9 per mm²) mitotic rate conveyed a significantly higher risk of CNS metastasis 1 year after diagnosis (HR,1.61; 95% CI, 1.15 to 2.26; P = .006). At all assessed conditional survival time points (1, 2, and 5 years after diagnosis), high mitotic rate (> 9 per mm²) tumors conveyed significantly higher risk of CNS metastasis compared with low (0-4 per mm²) mitotic rate, with HRs of 2.09 (95% CI, 1.47 to 2.96; P < .001), 1.92 (95% CI, 1.23 to 3.01; P = .004), and 2.38 (95% CI, 1.16 to 4.89; P = .019), respectively.

TABLE 3.

Multivariable Cumulative Incidence of CNS Metastasis on the Condition of Surviving 1, 2, and 5 Years, Modeled With Mitotic Rate Groups, Including Melanoma of Unknown Primary as a Separate Group

Similarly, conditional multivariable models were constructed adjusting for primary tumor site at 1, 2, and 5 years after diagnosis (Appendix Table A3, online only). One year after diagnosis without CNS recurrence, the difference between scalp and other head and neck sites was no longer significant. However, patients with scalp primaries continued to have a significantly higher risk of CNS metastasis compared with all other sites. By the second year after diagnosis, scalp had a similar risk to trunk and other head (nonscalp) and neck sites. Five years after diagnosis, only melanoma of unknown primary patients had a significantly lower subsequent risk of CNS metastasis compared with scalp.

DISCUSSION

This study reports estimates of the cumulative incidence of CNS metastasis for a large cohort of patients (N = 1,918) diagnosed with AJCC 8th edition stage III melanoma from 1998 to 2014 from 2 of the largest melanoma centers in the world. We found that 16.7% of the overall cohort developed CNS metastasis during a median follow-up of 70.2 months. Furthermore, among the patients subsequently diagnosed with stage IV disease, CNS metastases were detected in 15.3% at the time of initial diagnosis of distant metastasis (n = 711), including 10.5% who presented with CNS metastasis only. Male sex, younger age, increasing AJCC 8th edition stage group, extranodal extension of lymph node metastases, primary tumor site, and mitotic rate were associated with increased risk of CNS metastasis.

Outside of clinical trials, standard-of-care treatment options for stage III melanoma across the United States and Australia during the time frame of this cohort were consistent, including primarily surgery (eg, therapeutic or completion lymph node dissection); selective use of adjuvant, postoperative radiotherapy to the regional node field; and/or adjuvant, systemic therapy with interferon. An acknowledged limitation of our study was the lack of available data indicating whether patients in the cohort were treated with adjuvant radiotherapy or HDI. However, previous prospective clinical trials in patients with stage III disease reported that although adjuvant radiotherapy and HDI improved local disease control, neither intervention significantly reduced the risk of distant metastasis or death.^18,19 Consistent with that lack of effect, we observed comparable rates of CNS metastasis with those previously reported by Samlowski et al¹³ in their analysis of patients with high-risk stage III disease treated with either biochemotherapy or HDI in the S0008 clinical trial (N = 402) with mandated CNS imaging, even though the cohorts differed in stage inclusion (eg, no AJCC 7th edition stage IIIA-N1a; only 5% stage IIIA-N2a). For all S0008 patients, the cumulative incidence of CNS metastasis at 2 and 3 years was approximately 10% and 12%, respectively; we observed 9.6% and 13.0%, respectively.¹³ Similar proportions of patients also recurred at CNS sites during follow-up in the Samlowski et al¹³ study (15%) and this study (16.7%). Recently, the Food and Drug Administration approved adjuvant treatment with ipilimumab (2015), nivolumab (2017), dabrafenib and trametinib (2018), and pembrolizumab (2019) for patients with stage III melanoma. Each treatment resulted in significant improvements in relapse-free survival and distant metastasis-free survival.^20-23 Although we hope and expect that these therapies will reduce the risk of CNS metastasis, to date no data about their impact on the incidence or timing of these tumors have been presented. The data presented here provide important information for interpreting such data when they become available, and future studies should address whether the predictors of CNS metastasis identified in this cohort remain significant among patients treated with these adjuvant therapies.

The univariable and multivariable analyses of clinical data from 2 large, geographically separate institutions yielded similar results with respect to the cumulative incidence of CNS metastasis, despite the difference in predominant CNS imaging modality used (MRI at MD Anderson, CT at MIA) and the greater frequency of imaging at MD Anderson. Notably, we attempted to minimize the potential for lead-time bias in the detection of CNS metastasis by limiting the cohort to patients with confirmed negative CNS imaging within 4 months of diagnosis. This requirement precluded the use of the full cohort of patients included in the melanoma AJCC 8th edition analysis, because baseline CNS imaging was not required nor were CNS imaging data collected for those patients. Although we did not observe a difference in the incidence of CNS metastases between the institutions, there was a difference in clinical presentations (Appendix Table A1). Among patients diagnosed with CNS metastases, 51.6% of the patients at MIA presented with symptomatic disease, whereas 32.8% were asymptomatic; at MD Anderson, 37.8% were symptomatic, and 55.6% were asymptomatic (P = .041).

A novel finding of our study is the significance of primary tumor mitotic rate as a predictor of CNS metastasis at the time of stage III melanoma diagnosis. High (> 9 per mm²) mitotic rate was also the only significant predictor after 5 years of surviving without CNS metastasis. Mitotic rate was included in the AJCC 7th edition melanoma staging system (< 1 per mm² v ≥ 1 per mm²), but only for stratification of pT1 (≤ 1 mm Breslow thickness) primary tumors. Subsequent multivariable melanoma-specific survival analyses (staging patients in the AJCC 7th edition) also demonstrated the prognostic significance of this covariate across stage groups.²⁴ Although this factor was not included in the AJCC 8th edition staging system, it was strongly recommended that it be recorded across its dynamic range (ie, as an integer value) for all primary melanomas to be used for development of integrated risk models and clinical tools; the association of mitotic rate and prognosis was also assessed in detail across stage I and II melanoma cohorts with similar findings.^17,25,26 For our stage III melanoma cohort, although we observed higher proportions of patients with mitotic figures present (v stage I and II cohorts), the association of mitotic rate values with the cumulative incidence of CNS was the same (eg, increasing risk for increasing mitotic rate). Our finding affirms the significance of primary tumor mitotic rate in stage III melanoma, supporting the need for detailed assessment of the dynamic range in larger cohorts and consideration of inclusion in future melanoma prognostic models.

Overall, most patients were diagnosed with CNS metastases after the diagnosis of EC metastases. Interestingly, however, the majority of patients found to have CNS involvement at the initial diagnosis of stage IV disease had CNS disease only. Together, the results support the need to identify which subset of patients would benefit from regular CNS imaging, which otherwise appears to be unnecessary for the majority of patients with stage III disease. Our finding that primary melanoma on the scalp conveys a substantially higher risk of CNS metastasis for patients with AJCC 8th edition stage III melanoma is consistent with a previous analysis of patients with AJCC 7th edition stage I and II melanoma from MIA.²⁷ Compared with extremity, truncal, and other head and neck primary melanomas, scalp tumors were also found to have an increased risk of EC metastasis, although not in comparison with other head and neck tumors. Additional molecular or immune factors may further refine risk models for CNS metastasis. Molecular features associated with CNS metastasis from melanoma in previous studies include tumor BRAF and NRAS mutation status, PTEN and PLEKHA5 protein expression, and microRNAs.^28-30 Because these data are not routinely collected in this patient population, a planned extension of this study will assess and incorporate such factors into multivariable models to test their added value for prediction of CNS metastasis.

In conclusion, our study demonstrates that increased risk of CNS recurrence for patients with stage III melanoma is strongly associated with male sex, younger age, increasing stage III subgroup, increasing primary tumor mitotic rate, and scalp primary site. The results from 2 large, independent melanoma treatment centers provide important data regarding the timing and risk of CNS metastasis that can be used to guide clinical surveillance strategies. Furthermore, they provide an important framework for future analyses to assess the impact of recently approved adjuvant therapies on the risk of CNS metastasis, as well as investigations of molecular and immune factors that may be targeted to further reduce risk.

APPENDIX

FIG A1.

Conditional, univariable cumulative incidence of extracranial metastasis at time points (A) 1 year, (B) 2 years, and (C) 5 years after initial melanoma diagnosis.

TABLE A1.

Patterns of Brain Imaging for the Cohort by Institution and Symptomatic Versus Asymptomatic First Stage IV Presentation Including CNS

TABLE A2.

Multivariable Cumulative Incidence of EC Metastasis From Stage III Diagnosis, Model 1 With Mitotic Rate Groups and Model 2 With Primary Tumor Site Groups

TABLE A3.

Multivariable Cumulative Incidence of CNS Metastasis at Conditional Survival Time Points 1, 2, and 5 Years After Stage III Melanoma Diagnosis, Modeled With Primary Tumor Site

AUTHOR CONTRIBUTIONS

Conception and design: Lauren E. Haydu, Jennifer L. McQuade, Jennifer Wargo, Merrick I. Ross, Jonathan R. Stretch, Patrick Hwu, Alexander J. Lazar, Jeffrey E. Gershenwald, John F. Thompson, Michael A. Davies

Financial support: Jeffrey E. Gershenwald, Michael A. Davies

Administrative support: Jennifer Wargo, Patrick Hwu, Alexander J. Lazar

Provision of study materials or patients: Jennifer L. McQuade, Rodabe N. Amaria, Jennifer Wargo, Jeffrey E. Lee, Robyn Matteo S. Carlino, Alexander M. Menzies, Georgina V. Long, Alexander J. Lazar, Jeffrey E. Gershenwald, John F. Thompson

Collection and assembly of data: Lauren E. Haydu, Jennifer L. McQuade, Jennifer Wargo, Janice N. Cormier, Anthony Lucci, Robyn P.M. Saw, Kerwin F. Shannon, Jonathan R. Stretch, Michael K.K. Wong, Matteo S. Carlino, Georgina V. Long, Alexander J. Lazar, Michael T. Tetzlaff, Richard A. Scolyer, Jeffrey E. Gershenwald, John F. Thompson, Michael A. Davies

Data analysis and interpretation: Lauren E. Haydu, Serigne N. Lo, Jennifer L. McQuade, Rodabe N. Amaria, Anthony Lucci, Jeffrey E. Lee, Sherise D. Ferguson, Robyn P.M. Saw, Andrew J. Spillane, Kerwin F. Shannon, Sapna P. Patel, Adi Diab, Michael K.K. Wong, Isabella C. Glitza Oliva, Hussein Tawbi, Matteo S. Carlino, Alexander M. Menzies, Georgina V. Long, Alexander J. Lazar, Michael T. Tetzlaff, Richard A. Scolyer, Jeffrey E. Gershenwald, John F. Thompson, Michael A. Davies

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Cumulative Incidence and Predictors of CNS Metastasis for Patients With American Joint Committee on Cancer 8th Edition Stage III Melanoma

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/journal/jco/site/ifc.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).

Jennifer L. McQuade

Honoraria: Merck

Consulting or Advisory Role: Bristol Myer Squibb

Travel, Accommodations, Expenses: Merck

Rodabe N. Amaria

Research Funding: Merck (Inst), Bristol-Myers Squibb (Inst), Novartis (Inst), Array BioPharma (Inst), Genentech (Inst), Iovance Biotherapeutics (Inst)

Jennifer Wargo

Leadership: Microbiome DX

Honoraria: GlaxoSmithKline, Genentech, Novartis, DAVAOncology, Bristol-Myers Squibb, Illumina, Gilead Sciences, Merck, AstraZeneca/MedImmune, SITC

Consulting or Advisory Role: GlaxoSmithKline, Genentech, Novartis, Bristol-Myers Squibb, Illumina, Merck, AstraZeneca

Speakers' Bureau: Novartis, Genentech, Precision Health Economics, PERS, Illumina, DAVAOncology, PHE

Research Funding: Genentech (Inst), Novartis (Inst), GlaxoSmithKline (Inst), Bristol-Myers Squibb (Inst)

Patents, Royalties, Other Intellectual Property: Patent submitted regarding the role of the microbiome in response to immune checkpoint blockade

Travel, Accommodations, Expenses: GlaxoSmithKline, Genentech, Novartis, DAVAOncology, Bristol-Myers Squibb, Illumina

Merrick I. Ross

Honoraria: Merck, Amgen, Novartis

Consulting or Advisory Role: Merck, Amgen

Speakers' Bureau: Amgen

Research Funding: Amgen (Inst)

Travel, Accommodations, Expenses: Merck, Amgen, Novartis

Anthony Lucci

Speakers' Bureau: Genomic Health

Robyn P.N. Saw

Honoraria: Merck Sharp & Dohme, Novartis

Consulting or Advisory Role: Merck Sharp & Dohme, Novartis

Kerwin F. Shannon

Honoraria: Merck Serono

Consulting or Advisory Role: Merck Serono

Patrick Hwu

Stock and Other Ownership Interests: Immatics, Dragonfly Therapeutics

Consulting or Advisory Role: Dragonfly Therapeutics, GlaxoSmithKline, Immatics, Sanofi

Research Funding: Genentech (Inst)

Sapna P. Patel

Consulting or Advisory Role: Castle Biosciences, Incyte, Cardinal Health

Speakers' Bureau: Merck

Research Funding: Bristol-Myers Squibb (Inst), Celgene (Inst), Novartis (Inst), GlaxoSmithKline (Inst), Deciphera (Inst), Reata Pharmaceuticals (Inst), Provectus (Inst)

Travel, Accommodations, Expenses: Castle Biosciences, Incyte, Cardinal Health, Socrates Analytics, PCME, Rockpointe, Massachusetts General Hospital

Adi Diab

Honoraria: Array BioPharma

Consulting or Advisory Role: Nektar, CureVac, Celgene, Idera Pharmaceuticals

Research Funding: Nektar (Inst), Idera (Inst), Celgene (Inst), Pfizer (Inst), Merck, Apexigen

Travel, Accommodations, Expenses: Nektar

Michael K.K. Wong

Consulting or Advisory Role: Pfizer, Merck, Pfizer/EMD Serono, Regeneron, Bristol-Myers Squibb

Isabella C. Glitza Oliva

Consulting or Advisory Role: Array BioPharma, Bristol-Myers Squibb, Novartis

Speakers' Bureau: Array BioPharma

Research Funding: Bristol-Myers Squibb (Inst), Merck (Inst)

Travel, Accommodations, Expenses: Lilly

Hussein Tawbi

Consulting or Advisory Role: Novartis, Bristol-Myers Squibb, Genentech, Merck, Array BioPharma

Research Funding: Bristol-Myers Squibb (Inst), Novartis (Inst), Merck (Inst), GlaxoSmithKline (Inst), Genentech, Celgene

Matteo S. Carlino

Honoraria: Bristol-Myers Squibb, MSD, Novartis

Consulting or Advisory Role: Bristol-Myers Squibb, MSD, Amgen, Novartis, Pierre Fabre, Roche, Ideaya Biosciences

Alexander M. Menzies

Consulting or Advisory Role: MSD Oncology, Novartis, Pierre Fabre, Bristol-Myers Squibb, Roche

Georgina V. Long

Honoraria: Bristol-Myers Squibb, Merck, Pierre Fabre

Consulting or Advisory Role: Bristol-Myers Squibb, Merck, Novartis, Array BioPharma, Pierre Fabre, Aduro Biotech, OncoSec, Roche, Amgen

Alexander J. Lazar

Leadership: ArcherDX, Iterion Parmaceuticals, Nucleai

Stock and Other Ownership Interests: Archer Biosciences, Beta Cat Pharmaceuticals

Honoraria: Novartis, Bristol-Myers Squibb, Janssen Oncology, Genentech

Consulting or Advisory Role: Novartis, AbbVie, Bayer, BMS, Deciphera

Research Funding: MedImmune, AstraZeneca, Roche, Novartis

Patents, Royalties, Other Intellectual Property: Elsevier

Travel, Accommodations, Expenses: Bristol-Myers Squibb

Michael T. Tetzlaff

Consulting or Advisory Role: Myriad Genetics, Novartis, NanoString Technologies

Richard A. Scolyer

Consulting or Advisory Role: Bristol-Myers Squibb (Austria), Bristol-Myers Squibb (Switzerland), GlaxoSmithKline Australia, Dermpedia, Merck Sharp & Dohme, NeraCare, Novartis Pharmaceuticals Australia

Jeffrey E. Gershenwald

Consulting or Advisory Role: Merck, Novartis, Syndax, Bristol-Myers Squibb

Patents, Royalties, Other Intellectual Property: Mercator Therapeutics

John F. Thompson

Honoraria: GlaxoSmithKline, Bristol-Myers Squibb, MSD Australia, Provectus

Consulting or Advisory Role: GlaxoSmithKline, Bristol-Myers Squibb, MSD Australia, Provectus

Travel, Accommodations, Expenses: Provectus, GlaxoSmithKline

Michael A. Davies

Consulting or Advisory Role: GlaxoSmithKline, GenentechNovartis, Sanofi, Vaccinex, Bristol-Myers Squibb, Syndax, NanoString Technologies, Array BioPharma

Research Funding: GlaxoSmithKline (Inst), Genentech (Inst), AstraZeneca (Inst), Merck (Inst), Oncothyreon (Inst), Myriad Genetics (Inst), Sanofi (Inst)

No other potential conflicts of interest were reported.