[Skip to Navigation]
Sign In
Figure 1.  Overall Survival for the Proposed and Current Nodal Classification Systems for Cutaneous Melanoma
Overall Survival for the Proposed and Current Nodal Classification Systems for Cutaneous Melanoma

Kaplan-Meier overall survival estimates are given for the proposed (A) and the American Joint Committee on Cancer, 8th edition (AJCC-8) nodal classification systems in the National Cancer Database discovery cohort (B). A forest plot of the univariate hazard ratios (HRs) for overall survival are given for the proposed (C) and the AJCC-8 nodal classification systems (D). Vertical dashed lines represent the reference HR of 1 for patients with negative lymph nodes; error bars, 95% CI.

Figure 2.  Validation of the Proposed Nodal Classification System for Cutaneous Melanoma
Validation of the Proposed Nodal Classification System for Cutaneous Melanoma

Kaplan-Meier estimates are given for patients with cutaneous melanoma in the National Cancer Database, excluding those undergoing completion lymph node dissection (CLND) after sentinel lymph node biopsy (SLNB) for micrometastatic disease and those with unknown SLNB or CLND status (A). The proposed N3a and N3b nodal categories were combined given that few patients had 4 or more occult positive lymph nodes who did not undergo CLND (proposed N3a [n = 14]). Kaplan-Meier estimates are also given for overall survival (B) and cause-specific mortality for the proposed nodal classification system in the Surveillance, Epidemiology, and End Results database (SEER-18) validation cohort (C).

Figure 3.  Proposed Novel Overall Staging System for Node-Positive (Stage III) Cutaneous Melanoma
Proposed Novel Overall Staging System for Node-Positive (Stage III) Cutaneous Melanoma

A, Proposed overall staging subgroups for patients with node-positive (stage III) melanoma based on recursive partitioning analysis using both American Joint Committee on Cancer, 8th edition (AJCC-8) T category and the proposed nodal classification groups. B, Kaplan-Meier curves for overall survival of the proposed stage III subgroups for cutaneous melanoma.

Table 1.  Univariate and Multivariable Analysis of Overall Survival in Cutaneous Melanoma
Univariate and Multivariable Analysis of Overall Survival in Cutaneous Melanoma
Table 2.  Overall Survival for the Proposed Nodal Staging System for Cutaneous Melanoma
Overall Survival for the Proposed Nodal Staging System for Cutaneous Melanoma
1.
Siegel  RL, Miller  KD, Jemal  A.  Cancer statistics, 2019.   CA Cancer J Clin. 2019;69(1):7-34. doi:10.3322/caac.21551 PubMedGoogle ScholarCrossref
2.
Whiteman  DC, Green  AC, Olsen  CM.  The growing burden of invasive melanoma: projections of incidence rates and numbers of new cases in six susceptible populations through 2031.   J Invest Dermatol. 2016;136(6):1161-1171. doi:10.1016/j.jid.2016.01.035 PubMedGoogle ScholarCrossref
3.
Schadendorf  D, van Akkooi  ACJ, Berking  C,  et al.  Melanoma.   Lancet. 2018;392(10151):971-984. doi:10.1016/S0140-6736(18)31559-9 PubMedGoogle ScholarCrossref
4.
Welch  HG, Mazer  BL, Adamson  AS.  The rapid rise in cutaneous melanoma diagnoses.   N Engl J Med. 2021;384(1):72-79. doi:10.1056/NEJMsb2019760 PubMedGoogle ScholarCrossref
5.
Lo  SN, Scolyer  RA, Thompson  JF.  Long-term survival of patients with thin (T1) cutaneous melanomas: a Breslow thickness cut point of 0.8 mm separates higher-risk and lower-risk tumors.   Ann Surg Oncol. 2018;25(4):894-902. doi:10.1245/s10434-017-6325-1 PubMedGoogle ScholarCrossref
6.
Eggermont  AM, Chiarion-Sileni  V, Grob  JJ,  et al.  Adjuvant ipilimumab versus placebo after complete resection of high-risk stage III melanoma (EORTC 18071): a randomised, double-blind, phase 3 trial.   Lancet Oncol. 2015;16(5):522-530. doi:10.1016/S1470-2045(15)70122-1 PubMedGoogle ScholarCrossref
7.
Weber  J, Mandala  M, Del Vecchio  M,  et al; CheckMate 238 Collaborators.  Adjuvant nivolumab versus ipilimumab in resected stage III or IV melanoma.   N Engl J Med. 2017;377(19):1824-1835. doi:10.1056/NEJMoa1709030 PubMedGoogle ScholarCrossref
8.
Long  GV, Hauschild  A, Santinami  M,  et al.  Adjuvant dabrafenib plus trametinib in stage III BRAF-mutated melanoma.   N Engl J Med. 2017;377(19):1813-1823. doi:10.1056/NEJMoa1708539 PubMedGoogle ScholarCrossref
9.
Eggermont  AMM, Blank  CU, Mandala  M,  et al.  Adjuvant pembrolizumab versus placebo in resected stage III melanoma.   N Engl J Med. 2018;378(19):1789-1801. doi:10.1056/NEJMoa1802357 PubMedGoogle ScholarCrossref
10.
Dummer  R, Brase  JC, Garrett  J,  et al.  Adjuvant dabrafenib plus trametinib versus placebo in patients with resected, BRAFV600-mutant, stage III melanoma (COMBI-AD): exploratory biomarker analyses from a randomised, phase 3 trial.   Lancet Oncol. 2020;21(3):358-372. doi:10.1016/S1470-2045(20)30062-0 PubMedGoogle ScholarCrossref
11.
Elmore  JG, Elder  DE, Barnhill  RL,  et al.  Concordance and reproducibility of melanoma staging according to the 7th vs 8th edition of the AJCC Cancer Staging Manual.   JAMA Netw Open. 2018;1(1):e180083. doi:10.1001/jamanetworkopen.2018.0083 PubMedGoogle Scholar
12.
Gershenwald  JE, Scolyer  RA, Hess  KR,  et al; for members of the American Joint Committee on Cancer Melanoma Expert Panel and the International Melanoma Database and Discovery Platform.  Melanoma staging: evidence-based changes in the American Joint Committee on Cancer Eighth Edition Cancer Staging Manual.   CA Cancer J Clin. 2017;67(6):472-492. doi:10.3322/caac.21409 PubMedGoogle ScholarCrossref
13.
Faries  MB, Thompson  JF, Cochran  AJ,  et al.  Completion dissection or observation for sentinel-node metastasis in melanoma.   N Engl J Med. 2017;376(23):2211-2222. doi:10.1056/NEJMoa1613210 PubMedGoogle ScholarCrossref
14.
Leiter  U, Stadler  R, Mauch  C,  et al; German Dermatologic Cooperative Oncology Group (DeCOG).  Complete lymph node dissection versus no dissection in patients with sentinel lymph node biopsy positive melanoma (DeCOG-SLT): a multicentre, randomised, phase 3 trial.   Lancet Oncol. 2016;17(6):757-767. doi:10.1016/S1470-2045(16)00141-8 PubMedGoogle ScholarCrossref
15.
Leiter  U, Stadler  R, Mauch  C,  et al; German Dermatologic Cooperative Oncology Group.  Final analysis of DeCOG-SLT Trial: no survival benefit for complete lymph node dissection in patients with melanoma with positive sentinel node.   J Clin Oncol. 2019;37(32):3000-3008. doi:10.1200/JCO.18.02306 PubMedGoogle ScholarCrossref
16.
Boffa  DJ, Rosen  JE, Mallin  K,  et al.  Using the National Cancer Database for outcomes research: a review.   JAMA Oncol. 2017;3(12):1722-1728. doi:10.1001/jamaoncol.2016.6905 PubMedGoogle ScholarCrossref
17.
Hankey  BF, Ries  LA, Edwards  BK.  The Surveillance, Epidemiology, and End Results program: a national resource.   Cancer Epidemiol Biomarkers Prev. 1999;8(12):1117-1121.PubMedGoogle Scholar
18.
van Buuren  S, Groothuis-Oudshoorn  K.  mice: Multivariate imputation by chained equations in R.   J Stat Softw. 2011;45(3):67. doi:10.18637/jss.v045.i03 Google Scholar
19.
Harrell  FE. Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis. Springer; 2001. Springer Series in Statistics.
20.
Schoenfeld  D.  Partial residuals for the proportional hazards regression model.   Biometrika. 1982;69(1):239-241. doi:10.1093/biomet/69.1.239Google ScholarCrossref
21.
Hothorn  T, Hornik  K, Zeileis  A.  Unbiased recursive partitioning: a conditional inference framework.   J Comput Graph Stat. 2006;15(3):651-674. doi:10.1198/106186006X133933Google ScholarCrossref
22.
Segal  MR.  Regression trees for censored data.   Biometrics. 1988;44(1):35-47. doi:10.2307/2531894Google ScholarCrossref
23.
R: a language and environment for statistical computing. R Foundation for Statistical Computing; 2013. Accessed May 28, 2021. https://www.R-project.org/
24.
Roberts  TJ, Colevas  AD, Hara  W, Holsinger  FC, Oakley-Girvan  I, Divi  V.  Number of positive nodes is superior to the lymph node ratio and American Joint Committee on Cancer N staging for the prognosis of surgically treated head and neck squamous cell carcinomas.   Cancer. 2016;122(9):1388-1397. doi:10.1002/cncr.29932 PubMedGoogle ScholarCrossref
25.
Nguyen  AT, Luu  M, Lu  DJ,  et al.  Quantitative metastatic lymph node burden and survival in Merkel cell carcinoma.   J Am Acad Dermatol. 2021;84(2):312-320. doi:10.1016/j.jaad.2019.12.072PubMedGoogle ScholarCrossref
26.
O’Sullivan  B, Huang  SH, Su  J,  et al.  Development and validation of a staging system for HPV-related oropharyngeal cancer by the International Collaboration on Oropharyngeal Cancer Network for Staging (ICON-S): a multicentre cohort study.   Lancet Oncol. 2016;17(4):440-451. doi:10.1016/S1470-2045(15)00560-4 PubMedGoogle ScholarCrossref
27.
Ho  AS, Kim  S, Tighiouart  M,  et al.  Metastatic lymph node burden and survival in oral cavity cancer.   J Clin Oncol. 2017;35(31):3601-3609. doi:10.1200/JCO.2016.71.1176 PubMedGoogle ScholarCrossref
28.
Ho  AS, Kim  S, Tighiouart  M,  et al.  Association of quantitative metastatic lymph node burden with survival in hypopharyngeal and laryngeal cancer.   JAMA Oncol. 2018;4(7):985-989. doi:10.1001/jamaoncol.2017.3852 PubMedGoogle ScholarCrossref
29.
Aro  K, Ho  AS, Luu  M,  et al.  Development of a novel salivary gland cancer lymph node staging system.   Cancer. 2018;124(15):3171-3180. doi:10.1002/cncr.31535 PubMedGoogle ScholarCrossref
30.
Patel  DN, Luu  M, Zumsteg  ZS, Daskivich  TJ.  Development and validation of an improved pathological nodal staging system for urothelial carcinoma of the bladder.   Eur Urol Oncol. 2019;2(6):656-663. doi:10.1016/j.euo.2018.12.012 PubMedGoogle ScholarCrossref
31.
Gershenwald  JE, Scolyer  RA.  Melanoma staging: American Joint Committee on Cancer (AJCC) 8th edition and beyond.   Ann Surg Oncol. 2018;25(8):2105-2110. doi:10.1245/s10434-018-6513-7 PubMedGoogle ScholarCrossref
32.
Bilimoria  KY, Balch  CM, Bentrem  DJ,  et al.  Complete lymph node dissection for sentinel node-positive melanoma: assessment of practice patterns in the United States.   Ann Surg Oncol. 2008;15(6):1566-1576. doi:10.1245/s10434-008-9885-2 PubMedGoogle ScholarCrossref
33.
Hewitt  DB, Merkow  RP, DeLancey  JO,  et al.  National practice patterns of completion lymph node dissection for sentinel node-positive melanoma.   J Surg Oncol. 2018;118(3):493-500. doi:10.1002/jso.25160 PubMedGoogle Scholar
34.
Broman  KK, Hughes  T, Dossett  L,  et al.  Active surveillance of patients who have sentinel node positive melanoma: an international, multi-institution evaluation of adoption and early outcomes after the Multicenter Selective Lymphadenectomy Trial II (MSLT-2).   Cancer. 2021;127(13):2251-2261. doi:10.1002/cncr.33483 PubMedGoogle ScholarCrossref
35.
Rachidi  S, Deng  Z, Sullivan  DY, Lipson  EJ.  Shorter survival and later stage at diagnosis among unmarried patients with cutaneous melanoma: a US national and tertiary care center study.   J Am Acad Dermatol. 2020;83(4):1012-1020. doi:10.1016/j.jaad.2020.05.088 PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    Original Investigation
    September 1, 2021

    Development and Validation of a Modified Pathologic Nodal Classification System for Cutaneous Melanoma

    Author Affiliations
    • 1Department of Radiation Oncology, Cedars-Sinai Medical Center, Los Angeles, California
    • 2Samuel Oschin Comprehensive Cancer Institute, Cedars-Sinai Medical Center, Los Angeles, California
    • 3Department of Biostatistics and Bioinformatics, Cedars-Sinai Medical Center, Los Angeles, California
    • 4Division of Hematology and Oncology, Department of Medicine, UCLA (University of California, Los Angeles) School of Medicine
    • 5The Angeles Clinic and Research Institute, Los Angeles, California
    • 6Department of Medical Oncology, Cedars-Sinai Medical Center, Los Angeles, California
    • 7Department of Dermatology, Cedars-Sinai Medical Center, Los Angeles, California
    • 8Division of Otolaryngology–Head and Neck Surgery, Department of Surgery, Cedars-Sinai Medical Center, Los Angeles, California
    JAMA Surg. 2021;156(11):e214298. doi:10.1001/jamasurg.2021.4298
    Key Points

    Question  Can nodal classification for cutaneous melanoma be improved through agnostic recursive partitioning analysis (RPA)?

    Findings  In this cohort study of 105 785 patients, RPA was used to generate a modified prognostic nodal classification schema based on comprehensive nodal burden. In contrast to the staging system of the American Joint Committee on Cancer, 8th edition, the proposed system demonstrated progressively increasing mortality risk with each nodal classification group without overlapping prognoses in both discovery and validation cohorts.

    Meaning  These findings suggest that a modified nodal classification system for cutaneous melanoma can improve mortality risk stratification.

    Abstract

    Importance  Given the evolving patterns of lymph node evaluation for cutaneous melanoma, it is unclear whether the current nodal classification system will continue to accurately reflect prognosis in the modern era. Existing nodal staging for cutaneous melanoma was developed primarily for patients undergoing completion lymph node dissection (CLND) for node-positive disease and does not produce groups with continuously increasing mortality.

    Objective  To develop and validate a modified nodal classification system for cutaneous melanoma.

    Design, Setting, and Participants  This retrospective cohort analysis included 105 785 patients with cutaneous melanoma undergoing surgery and nodal evaluation from January 1, 2004, to December 31, 2015, in the National Cancer Database. Extent of lymph node dissection was available for patients diagnosed in 2012 and onward. Multivariable models were generated with number of positive lymph nodes modeled using restricted cubic splines. A modified nodal classification system was derived using recursive partitioning analysis (RPA). The proposed lymph node classification system was validated in 85 499 patients from the Surveillance, Epidemiology, and End Results (SEER-18) database. Data were analyzed from April 9, 2020, to May 28, 2021.

    Main Outcomes and Measures  Overall survival.

    Results  Among the 105 785 patients included in the analysis (62 496 men [59.1%]; mean [SD] age, 59.9 [15.5] years), number of positive lymph nodes (hazard ratio [HR] per lymph node for 0 to 2 positive lymph nodes, 2.48 [95% CI, 2.37-2-61; P < .001]; HR per lymph node for ≥3 positive lymph nodes, 1.10 [95% CI 1.07-1.13; P < .001]), clinically detected metastases (HR, 1.35; 95% CI, 1.27-1.42; P < .001), and in-transit metastases (HR, 1.48; 95% CI, 1.34-1.65; P < .001) were independently associated with mortality. An RPA-derived system using these variables demonstrated continuously increasing mortality for each proposed lymph node classification group, with HRs of 1.83 (95% CI, 1.76-1.91) for N1a, 2.72 (95% CI, 2.58-2.86) for N1b, 3.79 (95% CI, 3.51-4.08) for N2a, 4.56 (95% CI, 4.22-4.92) for N2b, 6.15 (95% CI, 5.59-6.76) for N3a, and 8.25 (95% CI, 7.64-8.91) for N3b in the proposed system (P < .001). By contrast, the current American Joint Committee on Cancer (AJCC) nodal classification system produced a more haphazard mortality profile, with HRs of 1.83 (95% CI, 1.76-1.91) for N1a, 3.81 (95% CI, 3.53-4.12) for N1b, 2.59 (95% CI, 2.30-2.93) for N1c, 2.71 (95% CI, 2.56-2.87) for N2a, 4.51 (95% CI, 4.17-4.87) for N2b, 3.44 (95% CI, 2.60-4.55) for N2c, 6.06 (95% CI, 5.51-6.67) for N3a, 8.15 (95% CI, 7.54-8.81) for N3b, and 6.90 (95% CI, 5.60-8.49) for N3c. As a sensitivity analysis, the proposed system continued to accurately stratify patients when excluding those undergoing CLND for microscopic lymph node metastases. This system was validated for overall survival and cause-specific mortality in SEER-18. Last, a new overall staging system for node-positive patients was developed by RPA and demonstrated improved concordance vs the AJCC, 8th edition system (C statistic, 0.690 [95% CI, 0.689-0.691] vs 0.666 [95% CI, 0.666-0.668]).

    Conclusions and Relevance  The findings of this cohort study suggest that a modified nodal classification system can accurately stratify mortality risk in cutaneous melanoma in an era of increasing use of sentinel lymph node biopsy without CLND and should be considered for future staging systems.

    Introduction

    The global incidence of cutaneous melanoma has been increasing rapidly in recent decades.1-4 Although most cases of melanoma are early stage and treated via surgery alone, advanced-stage melanomas characterized by nodal or distant metastases have been historically associated with poor survival.5 Since 2015, novel agents such as immune checkpoint inhibitors and molecularly targeted therapies have been shown to improve survival in patients with regional nodal involvement in the adjuvant setting.6-10 However, given the heterogeneity of outcomes for patients with node-positive disease, accurate staging and risk stratification is needed to improve patient selection for adjuvant treatment decisions.

    In 2018, the 8th edition of the American Joint Committee on Cancer (AJCC-8) staging system for cutaneous melanoma incorporated major revisions to the T, N, and M categories and stage groupings.11,12 One of the changes was the formal inclusion of in-transit regional disease into nodal staging. However, this new system did not result in increased mortality with increased nodal classification for all groups. For example, patients with N1b and N1c disease can have a more advanced overall stage and worse prognosis than patients with N2a disease.12 Moreover, given the increased use of sentinel lymph node biopsy (SLNB) without completion lymph node dissection (CLND) for patients with microscopic nodal disease following the Multicenter Selective Lymphadenectomy Trial II (MSLT-II)13 and German Dermatologic Cooperative Oncology-Selective Lymphadenectomy Trial (DeCOG-SLT),14,15 it is unclear whether the current nodal staging will continue to accurately reflect prognosis for patients with positive lymph nodes in the modern era.

    To refine the current nodal classification system used in cutaneous melanoma, we performed an agnostic clustering analysis based on the metrics of nodal burden that are currently used in AJCC-8 with data from a large national tumor registry. We then validated this modified system using overall and cause-specific mortality data from an independent registry. Last, we used unbiased partitioning analysis to develop a modified overall staging system for patients with positive lymph nodes.

    Methods
    Data Sources

    In this retrospective cohort study, patient data for nodal classification development and validation were abstracted from the National Cancer Database (NCDB) and from the Surveillance, Epidemiology, and End Results (SEER-18) database, respectively. The NCDB, which is sponsored by the American Cancer Society and the Commission on Cancer of the American College of Surgeons, includes more than 70% of newly diagnosed cancers in the US from more than 1500 cancer programs.16 The SEER-18 database, which includes approximately 35% of the US population across 18 cancer registries, was used for validation of the proposed system.17 Race/ethnicity were classified by coders for the NCDB and SEER-18. This study was deemed exempt from full review and patient consent requirements by the Cedars-Sinai Medical Center institutional review board owing to the use of deidentified publicly available data and followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

    Patients

    For the discovery NCDB data set, we included adult patients with cutaneous malignant melanoma (International Classification of Diseases for Oncology [Third Edition] [ICD-O-3] codes C44.0-C44.9 and histology codes 8720, 8721, 8742, 8743-8745, 8761, and 8780) who were diagnosed from January 1, 2004, to December 31, 2015, and underwent surgical resection and nodal evaluation (eFigure 1 in the Supplement). We excluded patients with incomplete follow-up, metastatic disease, incomplete pathologic staging, and noninvasive or unknown primary tumor and those who received neoadjuvant radiotherapy, chemotherapy, or immunotherapy given that these treatments could affect nodal burden. Patients were also excluded if they had an unknown number of positive or examined lymph nodes, indeterminate AJCC-8 nodal staging, unknown clinically occult vs clinically detected status, or unknown in-transit metastasis status. Given the prevalence of SLNB for cutaneous melanoma, we did not set a prespecified cutoff for minimum lymph node dissection.

    For the validation SEER-18 data set, we included adult patients with cutaneous malignant melanoma (same ICD-O-3 and histology codes as the NCDB cohort) diagnosed from January 1, 2000, to December 31, 2017 (eFigure 1 in the Supplement). We excluded patients with incomplete follow-up, patients who did not undergo surgery with lymph node evaluation, patients with an unknown number of positive or examined lymph nodes, and patients with unknown clinically occult vs clinically detected status.

    Statistical Analysis

    Data were analyzed from April 9, 2020, to May 28, 2021. Baseline characteristics were compared between patients with negative and positive lymph nodes using the Welch t test for continuous variables and Pearson χ2 test for categorical variables. The primary end point was overall survival, which was defined from the time of diagnosis to the time of last follow-up. Survival curves were estimated using the Kaplan-Meier method and were compared using a log-rank test. Cause-specific mortality curves were generated using cumulative incidence functions, with noncancer deaths as a competing risk, and compared with the k-sample test.

    Extent of lymph node dissection (SLNB alone, CLND alone, or SLNB and CLND) was available for patients diagnosed from 2012 onward and was captured by the RX_SUMM_SCOPE_REG_LN_2012 variable in the NCDB (2 indicates SLNB only; 3-5, CLND without SLNB; and 6-7, SLNB and CLND). Patients with at least 1 clinically occult positive lymph node who were missing the scope of lymph node surgery were excluded from patterns of care analysis and from a sensitivity analysis that excluded patients with microscopic nodal disease who underwent both SLNB and CLND.

    We determined the rate of missing data for race (0.87%), insurance type (1.47%), median zip code income (0.44%), median zip code educational attainment (0.39%), and population density (3.12%) and estimated these values using the multivariate imputations by chained equations algorithm.18 The model coefficients presented herein are the aggregated means among fitting the model on 10 complete imputed data sets with the variance equal to the imputation-corrected variance-covariance matrix. Univariate and multivariable survival analyses were generated using Cox proportional hazards regression models for overall survival and Fine-Gray models for cause-specific survival.

    The association between the number of positive lymph nodes and overall survival was modeled using a restricted cubic spline function, with the number of knots based on the lowest Akaike information criterion and their optimal location as defined by Harrell.19 The spline function of the log relative hazard ratio (HR) vs number of positive lymph nodes was plotted, and the change point in number of positive lymph nodes was estimated using piecewise linear regression modeling. The proportional hazards assumption was assessed by the scaled Schoenfeld residuals plot and goodness-of-fit test. Multicollinearity was assessed using the variable inflation factor.20

    Recursive partitioning analyses (RPAs) were used to develop a modified nodal classification system based on nodal variables independently associated with survival and to devlop a modified overall staging system for patients with positive lymph nodes based on T category (excluding patients with T category not otherwise specified) and the proposed nodal classification system.21,22 Binary partitions of these variables were identified to maximize mortality differences for each branch based on a permutation test of the log-rank statistic with Bonferroni adjustment for multiple comparisons. Performance of the AJCC-8 and proposed overall staging systems was evaluated in patients with positive lymph nodes using C statistics, which were calculated and corrected for possible optimism using internal out-of-sample bootstrap resampling with 200 replicates. Statistical analyses were performed using R, version 4.0.2 (R Foundation for Statistical Computing),23 with 2-sided tests and P < .05 indicating statistical significance.

    Results
    Patient Cohort

    Of 105 785 patients (62 496 men [59.1%] and 43 289 women [40.9%]; mean [SD] age, 59.9 [15.5] years) who met inclusion criteria, 17 837 (16.9%) had positive lymph nodes (eFigure 1 and eTable 1 in the Supplement). The median follow-up was 49.8 (95% CI, 49.5-50.1) months. Among patients with positive lymph nodes, the mean (SD) number of metastatic lymph nodes was 2.03 (2.85), and 3954 of 17 837 (22.2%) had clinically detected nodal involvement. In-transit disease was present in 1051 of 105 785 patients (1.0%).

    Among the 56 389 patients included in the patterns of care analysis from 2012 to 2015, 38 288 (67.9%) underwent SLNB alone, 11 254 (20.0%) underwent lymph node dissection without SLNB, and 6847 (12.1%) underwent SLNB and CLND. For the 7483 patients with at least 1 positive clinically occult lymph node, 3039 (40.6%) underwent CLND alone and 4444 (59.4%) underwent SLNB. Among the 4444 patients with clinically occult nodal involvement who underwent SLNB, 1854 (41.7%) did not undergo CLND. The rate of CLND omission for positive SLNB findings increased from 373 of 923 patients (40.4%) in 2012 to 556 of 1266 patients (43.9%) in 2015.

    Metastatic Lymph Node Features

    We examined the associations of individual nodal covariates included in the AJCC-8 system. On univariate analysis, clinically detected metastasis (HR, 4.27; 95% CI, 4.08-4.47; P < .001) and in-transit metastasis (HR, 2.50; 95% CI, 2.26-2.76; P < .001) were associated with increased risk of death. On multivariable analysis, these factors remained associated with 35% and 48% increased mortality, respectively, although the magnitude of the association was diminished when adjusting for other covariates (HR for clinically detected metastasis, 1.35 [95% CI, 1.27-1.42; P < .001]; HR for in-transit metastasis, 1.48 [95% CI, 1.34-1.65; P < .001]) (Table 1).

    Because the number of positive lymph nodes has a nonlinear association with mortality in many cancers,24-30 we used a restricted cubic spline function to assess this variable. We found that mortality risk continually increased with increasing number of positive lymph nodes without plateau (eFigure 2 in the Supplement). Mortality risk increased sharply with the initial 2 metastatic lymph nodes (HR per positive lymph node, 2.48; 95% CI, 2.37-2.61; P < .001) and continued to increase more gradually for 3 or more (HR per positive lymph node, 1.10; 95% CI, 1.07-1.13; P < .001).

    Proposed Nodal Staging System

    Using unbiased RPA with the number of positive lymph nodes, clinically occult vs detected metastasis, and in-transit metastasis, we identified 9 distinct clusters of patients (eFigure 3 in the Supplement). Two pairs of branches had convergent survival and were merged to create 7 separate groups. Specifically, patients with in-transit metastasis but no positive lymph nodes had nearly identical survival to patients with 2 to 3 clinically occult positive lymph nodes and were grouped. Similarly, patients with 1 clinically detected positive lymph node and patients with 1 clinically occult positive lymph node combined with in-transit metastasis were grouped together given similar survival rates. Thus, the proposed nodal staging system is as follows: N0 indicates 0 positive lymph nodes; N1a, 1 clinically occult positive lymph node; N1b, 2 to 3 clinically occult positive lymph nodes or 0 positive lymph nodes with in-transit metastasis; N2a, 1 clinically detected positive lymph node or 1 clinically occult positive lymph node with in-transit metastasis; N2b, 2 to 3 clinically detected positive lymph nodes with or without in-transit metastases; N3a, 4 or more clinically occult positive lymph nodes with or without in-transit metastases; and N3b, 4 or more clinically detected positive lymph nodes with or without in-transit metastases (Table 2).

    Kaplan-Meier estimates of the proposed nodal staging system and AJCC-8 classification system are shown in Figure 1A and B, respectively. The proposed nodal schema showed progressively increasing mortality, with 3-year overall survival rates of 89.9% (95% CI, 89.6%-90.1%) for N0, 79.2% (95% CI, 78.3%-80.1%) for N1a, 69.7% (95% CI, 68.2%-71.3%) for N1b, 58.9% (95% CI, 56.4%-61.6%) for N2a, 52.4% (95% CI, 49.5%-55.4%) for N2b, 43.2% (95% CI, 39.5%-47.2%) for N3a, and 36.1% (95% CI, 33.1%-39.4%) for N3b (Table 2). By contrast, the AJCC-8 system had multiple overlapping and crossing survival curves.

    In Cox proportional hazards regression, compared with the N0 category, the univariate HRs for RPA-derived N1a was 1.83 (95% CI, 1.76-1.91); for N1b, 2.72 (95% CI, 2.58-2.86); for N2a, 3.79 (95% CI, 3.51-4.08); for N2b, 4.56 (95% CI, 4.22-4.92); for N3a, 6.15 (95% CI, 5.59-6.76); and for N3b, 8.25 (95% CI, 7.64-8.91) (all P < .001) (Figure 1C). On the other hand, increasing nodal classification groups in the AJCC-8 system did not strictly correlate with increasing mortality (Figure 1D), with HRs of 1.83 (95% CI, 1.76-1.91) for N1a; 3.81 (95% CI, 3.53-4.12) for N1b; 2.59 (95% CI, 2.30-2.93) for N1c; 2.71 (95% CI, 2.56-2.87) for N2a; 4.51 (95% CI, 4.17-4.87) for N2b; 3.44 (95% CI, 2.60-4.55) for N2c; 6.06 (95% CI, 5.51-6.67) for N3a; 8.15 (95% CI, 7.54-8.81) for N3b; and 6.90 (95% CI, 5.60-8.49) for N3c (all P < .001). In multivariable Cox proportional hazards regression, the RPA-derived groups demonstrated steadily increasing mortality rates (eTable 2 in the Supplement) in contrast to nodal groups in the AJCC-8 system.

    Sensitivity Analysis

    Sentinel lymph node biopsy alone without CLND is now a standard of care for patients with micrometastases. Although CLND has been shown to have no effect on outcomes of patients with cutaneous melanoma and sentinel lymph node micrometastasis,13 CLND can influence the number of positive lymph nodes detected. Thus, patients undergoing SLNB alone could have a different prognosis than patients undergoing CLND, even if the same number of positive lymph nodes was detected owing to differences in diagnostic yield. Given that our proposed system depends heavily on the number of positive lymph nodes to classify patients, we performed a sensitivity analysis excluding patients with occult nodal disease undergoing CLND after SLNB (Figure 2A). All patients with micrometastases before 2012 were excluded because it was not possible to determine definitively whether they underwent CLND. Our proposed nodal classification system continued to accurately stratify mortality risk in this analysis. However, notably few patients with 4 or more occult positive lymph nodes underwent SLNB without CLND (proposed N3a category [n = 14]), so these patients were grouped with patients with 4 or more macroscopic lymph nodes (proposed N3b category).

    SEER Validation

    We validated our proposed RPA-derived nodal classification system in 85 499 patients with cutaneous melanoma from the SEER-18 database (51 087 men [59.8%] and 34 412 women [40.2%]; mean [SD] age, 59.2 [15] years). Similar to the discovery set, our nodal schema demonstrated continuously increasing overall mortality with increasing nodal classification group (Figure 2B). Cause-specific survival closely mirrored overall mortality rates by nodal classification group in the SEER-18 data set (Figure 2C).

    Proposed Overall Staging System

    Given that both T and N classifications are critical for prognosis, we sought to generate a novel, combined overall staging system for patients with node-positive melanoma. To this end, we used unbiased RPA with both the AJCC-8 T classification and our proposed nodal classification groups as inputs (eFigure 4 in the Supplement). We combined clusters of patients with similar survival to create 4 groups of patients with positive lymph nodes, similar to the current AJCC-8 system (Figure 3A). Our proposed system showed markedly divergent survival probability by stage (3-year overall survival for IIIA, 90.5 (95% CI, 89.5%-91.5%); for IIIB, 77.1% (95% CI, 76.0%-78.4%); for IIIC, 56.9% (95% CI, 55.3%-58.5%); and for IIID, 40.7% (95% CI, 38.7%-42.8%) (P < .001 by log-rank test) (Figure 3B). The proposed overall staging system demonstrated improved concordance (optimism-corrected C statistic, 0.690; 95% CI, 0.689-0.691) compared with the AJCC-8 system (C statistic, 0.666; 95% CI, 0.666-0.668).

    Discussion

    Although the AJCC-8 nodal classification system has improved patient stratification from previous editions, we identified several limitations to the current system. Notably, we found that in the AJCC-8 system, increasing nodal classification did not correlate with increasing mortality for many groups. For example, patients with AJCC-8 N1b disease had worse survival than those with AJCC-8 N1c and N2a disease and similar survival to those with AJCC-8 N2c disease. The complexity of this nodal staging system culminated in an overall staging schema for patients with stage III disease where increasing nodal classification can lead to a lower overall stage. In particular, patients with N2a disease may have a lower overall stage compared with patients with N1b or N1c disease despite having the same T classification.12,31

    In our analysis, we found that the nodal metrics used in the AJCC-8 system were all independently associated with mortality in cutaneous melanoma. Specifically, an increasing number of metastatic lymph nodes was associated with increased mortality without plateau, albeit nonlinearly, and macroscopic nodal detection and in-transit metastases were associated with 35% and 48% relative increases in mortality, respectively, when adjusting for other covariates. Thus, the fundamental issue in the AJCC-8 system is not the variables that are used, but rather the relative weight assigned to them for nodal classification.

    Using agnostic RPA with these 3 metrics of total nodal burden, we developed a modified nodal classification system that resulted in continuously increasing mortality with increasing nodal classification. Notably, our system has some similarities to the current AJCC-8 system. For example, our RPA-derived system clustered patients identified with cut points for number of positive lymph nodes identical to those used in the AJCC-8 system (0, 1, 2-3, and ≥4). On the other hand, our system assigned less significance to in-transit metastases, which was associated with prognosis only for patients with limited nodal burden (0-1 positive lymph nodes), and somewhat more significance to quantitative lymph node involvement. We further used RPA to derive a novel overall staging system for patients with node-positive melanoma combining AJCC-8 T classification with our proposed N classification and found that it had improved concordance vs the current AJCC-8 staging system.

    We demonstrated the reproducibility of our RPA-derived staging system in a large cohort from the SEER-18 database using a more inclusive approach than the discovery cohort by allowing patients with distant metastasis, preoperative cancer treatments, and a wider time span. We found the prognostic nodal groups identified in our RPA analysis were associated with both overall and cause-specific mortality in the SEER-18 data set. Thus, we believe that our proposed nodal staging system is generalizable and capable of accurately stratifying patient prognoses across the mortality spectrum.

    After MSLT-II13 and DeCOG-SLT,14,15 patterns of nodal evaluation in cutaneous melanoma have shifted toward increased use of SLNB without CLND, even for positive sentinel lymph nodes. This shift may lead to an underestimate of the total number of positive lymph nodes that would be detected with CLND. In our study, among patients with known SLNB and CLND status, 41.7% of patients with a positive SLNB did not undergo CLND, consistent with previously reported studies.32-34 Although the reason that so many patients did not undergo CLND before the publication of MSLT-II and DeCOG-SLT is unclear, these patterns of care allowed us to demonstrate that our proposed nodal system retained accuracy when excluding patients undergoing CLND after positive SLNB findings and thus retain applicability in modern practice. The main limitation of applying the proposed nodal classification system in patients without CLND is that very few patients undergoing SLNB without CLND have 4 or more occult positive lymph nodes (N3a category in our proposed system). A simple potential modification that would address this issue would be to merge the proposed N3a and N3b categories into a composite N3 category. Thus, we believe our system is well suited for current practice patterns of nodal evaluation.

    Limitations

    Our work has several limitations. First, it is a retrospective study using cancer registry data. However, this limitation is mitigated in a study assessing intrinsic prognostic factors given that these factors are less subject to unmeasured confounding. Nevertheless, several variables that could potentially affect survival were not available for analysis, including marital status35 and molecular profiling, such as BRAF mutation status. Inclusion of these variables may affect the estimation of the magnitude of the independent association of each nodal covariable with mortality in multivariable analysis but would not affect our staging development given that RPA is a univariate process. In addition, the use of real-world data from a diverse patient population is in many ways the ideal setting for staging development and validation because the AJCC largely validates all its staging systems within the NCDB. Another consideration is the potential overlap between patients who are captured by the NCDB and SEER, which we tried to mitigate by using a wider range of diagnosis years for the SEER validation cohort and by using more inclusive selection criteria. Last, all patients in this study were treated before 2018, thus few patients received the current standard of care with immunotherapy or molecularly targeted therapies used today.

    Conclusions

    Although nodal variables used in the AJCC-8 cutaneous melanoma system are associated with mortality, this system produces classification groups with overlapping or inverted prognoses compared with neighboring groups. In the present cohort study, we used agnostic RPA to develop an improved classification system that more accurately stratifies patient prognoses across the mortality spectrum. We further validated our proposed schema for both overall and cancer-specific survival using an independent patient registry. We believe our system should be considered for future versions of cutaneous melanoma staging.

    Back to top
    Article Information

    Accepted for Publication: June 12, 2021.

    Published Online: September 1, 2021. doi:10.1001/jamasurg.2021.4298

    Corresponding Author: Zachary S. Zumsteg, MD, Department of Radiation Oncology, Cedars-Sinai Medical Center, 8700 Beverly Blvd, Los Angeles, CA 90048 (zachary.zumsteg@cshs.org).

    Author Contributions: Dr Zumsteg had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

    Concept and design: A. T. Nguyen, V. P. Nguyen, Ho, Zumsteg.

    Acquisition, analysis, or interpretation of data: A. T. Nguyen, Luu, Hamid, Faries, Gharavi, Lu, Mallen-St Clair, Ho, Zumsteg.

    Drafting of the manuscript: A. T. Nguyen, Luu, Gharavi, Zumsteg.

    Critical revision of the manuscript for important intellectual content: A. T. Nguyen, V. P. Nguyen, Hamid, Faries, Lu, Mallen-St Clair, Ho, Zumsteg.

    Statistical analysis: A. T. Nguyen, Luu, Ho.

    Administrative, technical, or material support: V. P. Nguyen, Lu, Ho, Zumsteg.

    Supervision: A. T. Nguyen, Faries, Mallen-St Clair, Zumsteg.

    Conflict of Interest Disclosures: Dr Hamid reported consulting or serving on the advisory board for Aduro, Akeso Health Sciences, Amgen Inc, BeiGene, BioAtla, Bristol Myers Squibb, Roche Genentech, GlaxoSmithKline PLC, Immunocore, Idera, Inc, Incyte Corp, Janssen Global Services, LLC, Merck & Co, Inc, NextCure, Inc, Novartis International AG, Pfizer Inc, Sanofi/Regeneron, Seattle Genetics, Tempus, and Zelluna Immunotherapy and serving on the speakers bureau for Bristol Myers Squibb, Novartis International AG, Pfizer Inc, and Sanofi/Regeneron. Dr Faries reported serving on the advisory board for Merck & Co, Inc, Nektar Therapeutics, Sanofi, Bristol Myers Squibb, Array Bioscience, Pulse Biosciences, Inc, and Novartis International AG. Dr Gharavi reported being a principal investigator/key opinion leader for Castle Biosciences. Dr Zumsteg reported that his spouse performs legal work for Johnson & Johnson and Allergan PLC through her law firm. No other disclosures were reported.

    Funding/Support: This project was supported in part by Cedars-Sinai Cancer.

    Role of the Funder/Sponsor: Cedars-Sinai Cancer had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

    References
    1.
    Siegel  RL, Miller  KD, Jemal  A.  Cancer statistics, 2019.   CA Cancer J Clin. 2019;69(1):7-34. doi:10.3322/caac.21551 PubMedGoogle ScholarCrossref
    2.
    Whiteman  DC, Green  AC, Olsen  CM.  The growing burden of invasive melanoma: projections of incidence rates and numbers of new cases in six susceptible populations through 2031.   J Invest Dermatol. 2016;136(6):1161-1171. doi:10.1016/j.jid.2016.01.035 PubMedGoogle ScholarCrossref
    3.
    Schadendorf  D, van Akkooi  ACJ, Berking  C,  et al.  Melanoma.   Lancet. 2018;392(10151):971-984. doi:10.1016/S0140-6736(18)31559-9 PubMedGoogle ScholarCrossref
    4.
    Welch  HG, Mazer  BL, Adamson  AS.  The rapid rise in cutaneous melanoma diagnoses.   N Engl J Med. 2021;384(1):72-79. doi:10.1056/NEJMsb2019760 PubMedGoogle ScholarCrossref
    5.
    Lo  SN, Scolyer  RA, Thompson  JF.  Long-term survival of patients with thin (T1) cutaneous melanomas: a Breslow thickness cut point of 0.8 mm separates higher-risk and lower-risk tumors.   Ann Surg Oncol. 2018;25(4):894-902. doi:10.1245/s10434-017-6325-1 PubMedGoogle ScholarCrossref
    6.
    Eggermont  AM, Chiarion-Sileni  V, Grob  JJ,  et al.  Adjuvant ipilimumab versus placebo after complete resection of high-risk stage III melanoma (EORTC 18071): a randomised, double-blind, phase 3 trial.   Lancet Oncol. 2015;16(5):522-530. doi:10.1016/S1470-2045(15)70122-1 PubMedGoogle ScholarCrossref
    7.
    Weber  J, Mandala  M, Del Vecchio  M,  et al; CheckMate 238 Collaborators.  Adjuvant nivolumab versus ipilimumab in resected stage III or IV melanoma.   N Engl J Med. 2017;377(19):1824-1835. doi:10.1056/NEJMoa1709030 PubMedGoogle ScholarCrossref
    8.
    Long  GV, Hauschild  A, Santinami  M,  et al.  Adjuvant dabrafenib plus trametinib in stage III BRAF-mutated melanoma.   N Engl J Med. 2017;377(19):1813-1823. doi:10.1056/NEJMoa1708539 PubMedGoogle ScholarCrossref
    9.
    Eggermont  AMM, Blank  CU, Mandala  M,  et al.  Adjuvant pembrolizumab versus placebo in resected stage III melanoma.   N Engl J Med. 2018;378(19):1789-1801. doi:10.1056/NEJMoa1802357 PubMedGoogle ScholarCrossref
    10.
    Dummer  R, Brase  JC, Garrett  J,  et al.  Adjuvant dabrafenib plus trametinib versus placebo in patients with resected, BRAFV600-mutant, stage III melanoma (COMBI-AD): exploratory biomarker analyses from a randomised, phase 3 trial.   Lancet Oncol. 2020;21(3):358-372. doi:10.1016/S1470-2045(20)30062-0 PubMedGoogle ScholarCrossref
    11.
    Elmore  JG, Elder  DE, Barnhill  RL,  et al.  Concordance and reproducibility of melanoma staging according to the 7th vs 8th edition of the AJCC Cancer Staging Manual.   JAMA Netw Open. 2018;1(1):e180083. doi:10.1001/jamanetworkopen.2018.0083 PubMedGoogle Scholar
    12.
    Gershenwald  JE, Scolyer  RA, Hess  KR,  et al; for members of the American Joint Committee on Cancer Melanoma Expert Panel and the International Melanoma Database and Discovery Platform.  Melanoma staging: evidence-based changes in the American Joint Committee on Cancer Eighth Edition Cancer Staging Manual.   CA Cancer J Clin. 2017;67(6):472-492. doi:10.3322/caac.21409 PubMedGoogle ScholarCrossref
    13.
    Faries  MB, Thompson  JF, Cochran  AJ,  et al.  Completion dissection or observation for sentinel-node metastasis in melanoma.   N Engl J Med. 2017;376(23):2211-2222. doi:10.1056/NEJMoa1613210 PubMedGoogle ScholarCrossref
    14.
    Leiter  U, Stadler  R, Mauch  C,  et al; German Dermatologic Cooperative Oncology Group (DeCOG).  Complete lymph node dissection versus no dissection in patients with sentinel lymph node biopsy positive melanoma (DeCOG-SLT): a multicentre, randomised, phase 3 trial.   Lancet Oncol. 2016;17(6):757-767. doi:10.1016/S1470-2045(16)00141-8 PubMedGoogle ScholarCrossref
    15.
    Leiter  U, Stadler  R, Mauch  C,  et al; German Dermatologic Cooperative Oncology Group.  Final analysis of DeCOG-SLT Trial: no survival benefit for complete lymph node dissection in patients with melanoma with positive sentinel node.   J Clin Oncol. 2019;37(32):3000-3008. doi:10.1200/JCO.18.02306 PubMedGoogle ScholarCrossref
    16.
    Boffa  DJ, Rosen  JE, Mallin  K,  et al.  Using the National Cancer Database for outcomes research: a review.   JAMA Oncol. 2017;3(12):1722-1728. doi:10.1001/jamaoncol.2016.6905 PubMedGoogle ScholarCrossref
    17.
    Hankey  BF, Ries  LA, Edwards  BK.  The Surveillance, Epidemiology, and End Results program: a national resource.   Cancer Epidemiol Biomarkers Prev. 1999;8(12):1117-1121.PubMedGoogle Scholar
    18.
    van Buuren  S, Groothuis-Oudshoorn  K.  mice: Multivariate imputation by chained equations in R.   J Stat Softw. 2011;45(3):67. doi:10.18637/jss.v045.i03 Google Scholar
    19.
    Harrell  FE. Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis. Springer; 2001. Springer Series in Statistics.
    20.
    Schoenfeld  D.  Partial residuals for the proportional hazards regression model.   Biometrika. 1982;69(1):239-241. doi:10.1093/biomet/69.1.239Google ScholarCrossref
    21.
    Hothorn  T, Hornik  K, Zeileis  A.  Unbiased recursive partitioning: a conditional inference framework.   J Comput Graph Stat. 2006;15(3):651-674. doi:10.1198/106186006X133933Google ScholarCrossref
    22.
    Segal  MR.  Regression trees for censored data.   Biometrics. 1988;44(1):35-47. doi:10.2307/2531894Google ScholarCrossref
    23.
    R: a language and environment for statistical computing. R Foundation for Statistical Computing; 2013. Accessed May 28, 2021. https://www.R-project.org/
    24.
    Roberts  TJ, Colevas  AD, Hara  W, Holsinger  FC, Oakley-Girvan  I, Divi  V.  Number of positive nodes is superior to the lymph node ratio and American Joint Committee on Cancer N staging for the prognosis of surgically treated head and neck squamous cell carcinomas.   Cancer. 2016;122(9):1388-1397. doi:10.1002/cncr.29932 PubMedGoogle ScholarCrossref
    25.
    Nguyen  AT, Luu  M, Lu  DJ,  et al.  Quantitative metastatic lymph node burden and survival in Merkel cell carcinoma.   J Am Acad Dermatol. 2021;84(2):312-320. doi:10.1016/j.jaad.2019.12.072PubMedGoogle ScholarCrossref
    26.
    O’Sullivan  B, Huang  SH, Su  J,  et al.  Development and validation of a staging system for HPV-related oropharyngeal cancer by the International Collaboration on Oropharyngeal Cancer Network for Staging (ICON-S): a multicentre cohort study.   Lancet Oncol. 2016;17(4):440-451. doi:10.1016/S1470-2045(15)00560-4 PubMedGoogle ScholarCrossref
    27.
    Ho  AS, Kim  S, Tighiouart  M,  et al.  Metastatic lymph node burden and survival in oral cavity cancer.   J Clin Oncol. 2017;35(31):3601-3609. doi:10.1200/JCO.2016.71.1176 PubMedGoogle ScholarCrossref
    28.
    Ho  AS, Kim  S, Tighiouart  M,  et al.  Association of quantitative metastatic lymph node burden with survival in hypopharyngeal and laryngeal cancer.   JAMA Oncol. 2018;4(7):985-989. doi:10.1001/jamaoncol.2017.3852 PubMedGoogle ScholarCrossref
    29.
    Aro  K, Ho  AS, Luu  M,  et al.  Development of a novel salivary gland cancer lymph node staging system.   Cancer. 2018;124(15):3171-3180. doi:10.1002/cncr.31535 PubMedGoogle ScholarCrossref
    30.
    Patel  DN, Luu  M, Zumsteg  ZS, Daskivich  TJ.  Development and validation of an improved pathological nodal staging system for urothelial carcinoma of the bladder.   Eur Urol Oncol. 2019;2(6):656-663. doi:10.1016/j.euo.2018.12.012 PubMedGoogle ScholarCrossref
    31.
    Gershenwald  JE, Scolyer  RA.  Melanoma staging: American Joint Committee on Cancer (AJCC) 8th edition and beyond.   Ann Surg Oncol. 2018;25(8):2105-2110. doi:10.1245/s10434-018-6513-7 PubMedGoogle ScholarCrossref
    32.
    Bilimoria  KY, Balch  CM, Bentrem  DJ,  et al.  Complete lymph node dissection for sentinel node-positive melanoma: assessment of practice patterns in the United States.   Ann Surg Oncol. 2008;15(6):1566-1576. doi:10.1245/s10434-008-9885-2 PubMedGoogle ScholarCrossref
    33.
    Hewitt  DB, Merkow  RP, DeLancey  JO,  et al.  National practice patterns of completion lymph node dissection for sentinel node-positive melanoma.   J Surg Oncol. 2018;118(3):493-500. doi:10.1002/jso.25160 PubMedGoogle Scholar
    34.
    Broman  KK, Hughes  T, Dossett  L,  et al.  Active surveillance of patients who have sentinel node positive melanoma: an international, multi-institution evaluation of adoption and early outcomes after the Multicenter Selective Lymphadenectomy Trial II (MSLT-2).   Cancer. 2021;127(13):2251-2261. doi:10.1002/cncr.33483 PubMedGoogle ScholarCrossref
    35.
    Rachidi  S, Deng  Z, Sullivan  DY, Lipson  EJ.  Shorter survival and later stage at diagnosis among unmarried patients with cutaneous melanoma: a US national and tertiary care center study.   J Am Acad Dermatol. 2020;83(4):1012-1020. doi:10.1016/j.jaad.2020.05.088 PubMedGoogle ScholarCrossref
    ×