Risk Factors and Predictors of Survival Among Patients with Amelanotic Melanoma Compared to Melanotic Melanoma in the National Cancer Database

J Clin Aesthet Dermatol. 2021;14(12):36-43.

by Zachary H. Hopkins, MD; Ryan P. Carlisle, BS; Zachary E. Frost; Julia A. Curtis, MD; Laura K. Ferris, MD, PhD; and Aaron M. Secrest, MD, PhD

Drs. Hopkins, Curtis, and Secrest are with the Department of Dermatology, University of Utah in Salt Lake City, Utah. Mr. Carlisle is with the School of Medicine, University of Utah in Salt Lake City, Utah. Mr. Frost is with the Speech Pathology Undergraduate Program, University of Utah in Salt Lake City, Utah. Dr. Secrest is also with the Department of Population Health Sciences, University of Utah in Salt Lake City, Utah. Dr. Ferris is with the Department of Dermatology, University of Pittsburgh in Pittsburgh, Pennsylvania.

FUNDING: No funding was provided for this article.

DISCLOSURES: Dr. Ferris is an investigator for Castle Biosciences and an investigator and consultant for DermTech and Scibase. Dr. Secrest receives support from a Dermatology Foundation Public Health Career Development Award. The other authors report no conflicts of interest relevant to the content of this article. The American College of Surgeons and the Commission on Cancer have not verified and are not responsible for the analytic or statistical methodology employed, or the conclusions drawn from these data by the investigators.

ABSTRACT: Background. Amelanotic melanoma (AM) is a rare form of melanoma lacking pigment. Data on AM risk factors and factors predicting survival are limited.

Objectives. We sought to identify predictors of AM, survival differences in AM and melanotic melanoma, and AM-specific survival rates.

Methods. Using 2004 through 2015 National Cancer Database data, we compared 358,543 melanoma cases to 1,384 AM cases. Multivariable logistic regression identified AM risk factors, and AM survival was explored using Kaplan-Meier and multivariable Cox regression.

Results. Increased age; tumor location on the face, scalp, and neck; increased Breslow thickness; metastatic disease; ulceration; and higher mitotic rate were associated with AM. Five- and ten-year survival rates were higher for patients with MM (melanotic melanoma) than AM tumors (75.4% vs. 58.8% and 62.4% vs 45.1%; log-rank P<0.0001). No survival difference was seen after adjusting for staging factors. Among patients with AM, more recent diagnosis was associated with improved survival. Increased age, T4 tumor size, higher N-stage, metastasis, and ulceration predicted poorer survival. No survival advantage was seen for chemotherapy, immunotherapy, or radiation therapy, likely due to confounding.

Conclusion. AM is more common in older patients on sun-exposed skin and is diagnosed at later stages. Advanced staging at diagnosis explains the survival differences. In patients with AM, regional and metastatic disease were the primary contributors of poorer outcomes. In at-risk patients, the threshold to biopsy should be lower for suspicious nonpigmented lesions.

Key words: Amelanotic melanoma, melanoma, NCDB, melanoma survival, skin cancer, dermatology


Amelanotic melanoma (AM) is rare, composing 2 to 8 percent of all melanomas.1 Clinically, lesions are characterized by little to no pigment and can present as red-to-pink macules, papules, or plaques.2,3 However, the clinical appearance is not reported in major databases, and only the histologic definition of amelanotic melanoma is used.4 One study compared central histologic reviews and available clinical data and found 95 percent of clinically melanotic melanomas were coded as melanotic, whereas, 80 percent of clinically amelanotic melanomas were coded as amelanotic.4 Thus, the histologic definition is more narrow and may under-report clinically amelanotic lesions. Since large-scale data comparing histologic and clinical definitions to outcomes are not available, it is unclear which definition is superior. 

Histologically, AM likely represents a true subtype of melanoma rather than a poorly differentiated melanoma subgroup without melanin-producing ability.2 AM is more difficult to diagnose, frequently mistaken for other non-melanotic lesions, and only infrequently included on initial clinical differentials.3,5 Survival appears to be poorer in patients with histologically and clinically defined AM.1,4,6,7 Whether this is due to a more aggressive tumor phenotype or more advanced stage at presentation is not known. However, AMs have higher mitotic indices, grow faster, and portend to worse survival, independent of detection delay.4,7–10 One study found greater Breslow thicknesses and worse staging at presentation for AM compared to melanotic melanoma (MM) despite a similar time to diagnosis.11

Large analyses evaluating predictors for AM-specific survival are limited. Single institution experiences and other databases (e.g., Surveillance, Epidemiology, and End Results [SEER] and the Genes, Environment, and Melanoma Study [GEM]) have suggested that AM is associated with advanced age, occurrence on sun-exposed areas (e.g., face, ears), higher stage at diagnosis, higher likelihood of ulceration, and poorer survival relative to MM.1,4,6,7,12 However, heterogeneity in these data exist. For example, the SEER study found AMs were more likely to ulcerate than MM; however, the GEM study found this effect was removed after adjustment for tumor thickness.13 Neither evaluated mitosis rate. Likewise, findings vary regarding sex predilection, the role of amelanosis in prognosis, and risk factors for poor outcomes in those with AM.13

In this study, we used data from the National Cancer Database (NCDB), the largest cancer database in the United States. We evaluated (1) factors associated with AM diagnosis, (2) the role of amelanosis as an independent predictor of survival, (3) factors predicting survival among those with AM, and (4) trends in treatment modality usage.

Methods 

The NCDB, a joint program of the American Cancer Society and the Commission on Cancer of the American College of Surgeons, captures 48 percent of all melanomas diagnosed in the United States.14 Our hypothesis and analysis plan were submitted and approved prior to being provided de-identified data for all cutaneous melanomas from 2004 to 2015. Institutional review board approval was not required.

Patient and variable selection. Only invasive tumors were selected. Amelanotic melanoma (AM) was selected using the International Classification of Diseases for Oncology morphology code 8730 (n=1,384). MM codes included 8720 to 8723, 8740 and 8741, 8743 to 8745, 8761, 8770 to 8774, and 8780 (n=357,159). Patients with missing histology information, lentigo maligna-type melanoma, or no primary tumor were excluded. Patients with missing follow-up times and vital status were included in logistic analyses evaluating association between MM and AM but were excluded for survival analyses. 

Demographic variables included age, sex, race/ethnicity, insurance, facility type and location, urban density, and Charlson-Deyo Comorbidity Index score. Disease-based variables included year of diagnosis, histology (MM/AM), Breslow thickness, regional lymph nodes status, American Joint Committee on Cancer (AJCC) staging (T/N/M), ulceration status, mitosis count per mm2, when treatment started (days from initial diagnosis), treatments received (chemotherapy, surgery, radiotherapy, immunotherapy), last contact or death (months from initial diagnosis), vital status, and date of death. Income and education were not used as these are extrapolated zip-code based data. Breslow thickness was categorically coded to match T-stage cutoffs because values over 8.89mm were truncated, similar to other studies.1 Age, year of diagnosis, and follow-up time remained continuous variables. 

Statistical analysis. Differences between MM and AM lesions were compared descriptively; the chi-squared test for categorical variables and quantile regression for medians of continuous variables were used.15

Multiple imputation. The presence of missing values for each variable is noted in Table 1. Missing data and resulting case-wise deletion can bias results if not missing completely at random, so multiple imputation utilizing chained functions was performed.16 Variables with missing data for imputation included ulceration, M-stage, mitoses per mm2, lymphovascular space invasion (LVSI), margin status, treatment(s) used, vital status, site of tumor, treatment facility type, insurance type, urban density, race, Breslow depth, N-stage, distance to clinic; and follow-up time. Vital status and follow-up time were only imputed for logistic regression models. Variables without missing data used in the imputation model included: patient age, sex, melanoma histology, Charlson Comorbidity Index score, and year of diagnosis. Twenty imputed datasets were created. Imputation model diagnostics were assessed graphically. The Stata “mi estimate” command was used for all imputed logistic and Cox regression models. 

Factors associated with AM. Factors associated with AM were evaluated using both univariable and multivariable logistic regression. Multivariable models were devised using a priori data, clinical knowledge, and expert opinion for important predictive factors. Variables related to disease development included age, sex, race, geographic region, urban density, and primary site.1,4,6–9,12 Variables related to disease phenotype or diagnostic differences included Breslow depth, N-stage, M-stage, ulceration, and presence of LVSI at diagnosis.1,4,6,7,17 Variables lacking a temporal relationship or clear role in pathophysiology included Charlson Comorbidity Index, facility type at diagnosis, and insurance type. An interaction term was tested between Breslow depth and ulceration. For model comparison, we utilized a lasso function to shrink a model with all predictors to one with statistically-based variable selection. 

Model diagnostics included linktest and graphical evaluation of standardized Pearson residuals, deviance residuals, and leverage. Sensitivity analysis included comparisons between the theory-based model and the lasso-derived model and comparisons between results from case-wise deletion and multiple imputation-derived datasets. 

Survival analysis. Survival differences between AM and MM were evaluated using the Kaplan-Meier with log-rank test. Cox proportional hazards regression was used to evaluate factors predicting survival differences. Survival differences between AM and MM were compared after adjusting for age and sex, followed by sequential adjustment for Breslow depth, N-stage, and M-stage, ulceration, and mitosis count. 

For patients with AM, univariable and multivariable Cox proportional hazards regression models, devised from a hierarchical framework (core variables including age, sex, race, tumor staging, mitotic count, and LVSI) based on prior data and expert opinion, were used to evaluate factors predicting survival differences. Model diagnostics and postestimation residuals were assessed. All analyses were performed using Stata version 14.2 (StataCorp LLC, College Station, Texas).

Results

We evaluated 358,543 cases of primary melanoma reported to the NCDB between 2004 and 2015. Of these, 1,384 (0.4%) were AM and 357,159 were MM (Table 1). Patients with AM were older; more likely to have tumors on the lips, external ear, face, scalp/neck, or upper limb/shoulder; and were more likely to have increased Breslow thicknesses, positive margins on initial biopsy or excision, ulcerated lesions, and be metastatic at presentation. Likewise, patients with AM were more likely to receive radiation therapy, chemotherapy, or immunotherapy.

Based on multivariable logistic regression (Table 2), factors predicting AM included age; and tumor location on the face, scalp/neck, or arms/shoulders; and a geographic location in the west. At presentation, AM tumors were more likely to be thicker, ulcerated, and have higher mitotic rates. Patients were also more likely to have metastatic disease.

A significant interaction was found between Breslow thickness and ulceration status. AM lesions were more likely to be ulcerated, but this effect was statistically significant for smaller lesions (Figure 1).

Median follow-up was 4.1 years (maximum, 13.1 years) for MM and 3.4 years (maximum, 12.8 years) for AM. Five- and 10-year survival rates for patients with AM were 59.1% (95% conofidence interval [CI], 56.0%–62.0%) and 47.0% (95 CI, 43.3%–50.7%), respectively. For MM, five- and 10-year survival rates were significantly higher at 76.0% (95 CI, 75.8%–76.1%) and 63.9% (95 CI, 63.7%–64.1%) (log-rank test P<0.0001) (Figure 2A). Unadjusted survival was reduced for patients with AM (hazard ratio, 1.84; 95% CI, 1.69–2.02)(Figure 2A). No statistical difference in survival between AM and MM was observed after controlling for age, sex, and staging variables (i.e., Breslow thickness, ulceration, N-stage, and M-stage) (adjusted hazard ratio, 1.01; 95% CI, 0.90–1.14) (Table 3).

Among patients with AM, survival improved over time (adjusted hazard ratio, 0.96; 95% CI, 0.92–0.997). Survival was decreased for regional and distant disease compared to localized disease (P<0.0001) (Figure 2B). Following multivariable Cox regression, factors associated with worse AM survival included increased age, 4.01- to 9.89+-mm lesions, increasing N-stage, metastatic disease, and ulceration (Table 4). 

Patients with AM were less likely to undergo or be recommended surgery. They were also slightly more likely to have biopsy followed by excision rather than Mohs, wide-local excision, or local excision as initial therapy. Patients with AM were more likely to receive radiation therapy, chemotherapy, and/or immunotherapy, and were more likely to receive neoadjuvant radiotherapy (Table 1). After multivariable adjustment, no therapy showed a statistically significant impact on survival, and no significant time trend in therapy usage was observed.

Discussion 

Our study of 1,384 United States cases of AM is the largest epidemiologic study of AM to date. However, the number of AM cases relative to that of MM was low (0.4%). While the NCDB cannot be used to infer population incidence, this rate is lower than general population estimates.1 However, it matches the rate seen in the population-based SEER database study.1 Another large database study from the GEM, showed a higher rate (8%), but this study was smaller and still falls at the lower end of estimates.4 One cause may be misclassification due to multiple histologic codes being attributed to a sample, leading to under-representation of AM. However, this has not occurred with patients in other reports to the best of our knowledge, and, in the GEM study with central pathologic review, most clinically pigmented melanomas were histologically identified as such (95%).2,4,6,18 Another reason may be the reliance on a histologic requirement for amelanosis. This may explain the similarity seen relative to the SEER study. Regardless, overlap is likely minimal, and while perhaps under-represented, we assert that inference between histologically defined groups is valid and applicable. 

Like in prior studies, we found increasing patient age and sun-exposed skin were the two most important clinical risk factors for AM.1,4,6,7 Prior work hypothesized that this age-effect is mediated by lifetime ultraviolet exposure, differences in tumor biology, or later detection.1 The relatively minor contribution of age and the larger effect from anatomic distribution seen in this cohort may support chronic ultraviolet exposure as playing a more important role in this tumor’s pathophysiology as compared to MM rather than detection delay. 

Sex was not associated with AM after multivariable adjustment. This is unsurprising as prior reports came from small, largely unadjusted models, which led to small effect sizes.1,4 Hypotheses for previously noted sex differences included differential ultraviolet exposure.1 To investigate this, we tested an interaction term between sex and body location and found that women were more likely to have AM on the face (1.2% chance vs. 0.7% chance for men; P=0.03) with no significant sex differences seen for other body locations. This finding weakens this hypothesis, at least as it pertains to this melanoma subtype. 

AM presented with ulceration at thinner tumor sizes compared to MM (Figure 1). Early ulceration may be due to more aggressive histologic phenotype2,4,7,9,10 or ulceration triggering earlier clinical workup and detection of AM, leading to smaller tumors with ulceration being enriched in the database. Together, we believe our findings of earlier ulceration and higher mitotic rates in AM tumors offers support for AM being a more aggressive phenotype. Prior large databases did not evaluate mitotic rates when comparing these tumors, and we believe this offers important context. 

Five- and 10-year survival rates were lower for AM relative to MM (58.8% vs. 75.4% and 45.1% vs. 62.4%, respectively). These survival rates were similar to the SEER data.1 After adjusting for disease extent, no statistically or clinically meaningful difference in survival remained. Since the survival decrement appears dependent on staging rather than an independent unmeasured factor, we suggest that the more aggressive histologic AM phenotype leads to a more rapid achievement of higher tumor staging. This is supported by research reporting a worse prognosis independent of diagnostic delays.11 Given these data, one important clinical question may be whether amelanosis should affect staging workup (sentinel lymph node biopsy, imaging, etc.), especially in light of our finding that AM melanomas are more likely to be metastatic at presentation. Future prospective studies may be warranted to investigate this question and impact on outcomes. 

Direct assessment of survival predictors in AM has been limited by small case numbers; however, one prior study utilizing the national SEER database suggested that increased age and tumor spread (localized vs. regional vs. widespread) were significant predictors of poor outcomes.1 Our data corroborate this finding. We further found that current T/N/M staging factors largely predicted survival in AM in much the same way as MM except that the interaction between ulceration and Breslow depth was not statistically significant for any Breslow thickness. Thus, unlike in MM,19 ulceration in AM does not appear to modify survival at each T-stage, only itself as an independent risk factor. Notably, mitotic rate and LVSI were not statistically significant predictors of survival. 

Finally, we found that AM lesions were less likely to be surgically removed than MM lesions largely from medical recommendation and not patient refusal or technical barriers. This lends further support to more advanced, systemic disease at presentation. AM lesions were also more likely to be conservatively biopsied or treated with local ablative therapy, reinforcing the diagnostic uncertainty surrounding AMs. After diagnosis, patients with AM were more likely to undergo adjuvant and neoadjuvant radiation therapy as well as adjuvant chemotherapy and immunotherapy. Though we did not have sufficient evidence to support a survival advantage in patients receiving these therapies, this likely stems from residual confounding by indication rather than a lack of effect. While improved screening or diagnosis of AM with techniques such as dermoscopy, clinical imaging, or other techniques may improve outcomes, we suspect that advances in diagnosis may be more difficult to attain and have less impact than improvements in advanced therapy, especially for this subgroup of melanomas that more often present with advanced inoperable tumors.20 Indeed, given histologic differences,2,4,9,10 and apparent tendency towards more aggressive spread, therapies tailored directly towards this tumor group may be of import. Notably, we did see a small trend towards improved survival in later years. This may signal an optimistic trend, perhaps owning to more recent advances in immunotherapy especially. 

One important limitation is that the NCDB data are derived from patients receiving care at a Commission on Cancer–accredited hospital, limiting population-level metrics like incidence, prevalence, or mortality rate. Other limitations related to the NCDB setting include melanoma often being treated in the outpatient setting, outside of large referral centers, possibly leading to bias towards more high-risk, referred patients or tumors. Additionally, though multiple imputation was performed to minimize the risk of missing data, it is possible that these values were missing not at random, and lingering bias may persist. In AM, this may be due to more severe initial disease presentation where patients refused further workup, or where further staging was not indicated. Despite these limitations, the NCDB’s size and representation from many different clinical centers strengthens our findings and highlights important data concerning this rare tumor subtype. 

Conclusion

Our findings highlight factors associated with AM and a significant staging-mediated discrepancy in survival between AM and MM. Our data also suggest that older age, higher N-stage, metastatic disease, and ulceration are important factors for survival among those with AM, while Breslow thickness confers less predictive power except in the thickest of tumors. The diagnosis of AM remains difficult, but techniques like dermoscopy and a lower threshold to biopsy suspicious lesions in at-risk patients is important. Lastly, AM tumors were more likely to be metastatic, thick, and inoperable and were more likely to receive adjuvant and neo-adjuvant therapies, reflective of advanced stage at presentation. This highlights the need not only for better diagnostics but also tailored staging workup strategies and advanced treatment therapies for this high-risk patient population. 

References

  1. Moreau JF, Weissfeld JL, Ferris LK. Characteristics and survival of patients with invasive amelanotic melanoma in the USA. Melanoma Res. 2013;23(5):408–413. 
  2. Cheung WL, Patel RR, Leonard A, Firoz B, Meehan SA. Amelanotic melanoma: a detailed morphologic analysis with clinicopathologic correlation of 75 cases. J Cutan Pathol. 2012;39(1):33–39. 
  3. McClain SE, Mayo KB, Shada AL, et al. Amelanotic melanomas presenting as red skin lesions: a diagnostic challenge with potentially lethal consequences: amelanotic melanomas presenting as red skin lesions. Int J Dermatol. 2012;51(4):420–426. 
  4. Thomas NE, Kricker A, Waxweiler WT, et al. Comparison of clinicopathologic features and survival of histopathologically amelanotic and pigmented melanomas: a population-based study. JAMA Dermatol. 2014;150(12):1306–1314.
  5. Strazzula L, Vedak P, Hoang MP, et al. The utility of re-excising mildly and moderately dysplastic nevi: a retrospective analysis. J Am Acad Dermatol. 2014;71(6):1071–1076. 
  6. Strazzulla LC, Li X, Zhu K, et al. Clinicopathologic, misdiagnosis, and survival differences between clinically amelanotic melanomas and pigmented melanomas. J Am Acad Dermatol. 2019;80(5):1292–1298. 
  7. Wee E, Wolfe R, Mclean C, et al. Clinically amelanotic or hypomelanotic melanoma: anatomic distribution, risk factors, and survival. J Am Acad Dermatol. 2018;79(4):645–651.e4.
  8. Chamberlain AJ, Fritschi L, Giles GG, et al. Nodular type and older age as the most significant associations of thick melanoma in Victoria, Australia. Arch Dermatol. 2002;138(5):609–614. 
  9. Liu W, Dowling JP, Murray WK, et al. Rate of growth in melanomas: characteristics and associations of rapidly growing melanomas. Arch Dermatol. 2006;142(12):1551–1558.
  10. Shen S, Wolfe R, McLean CA, et al. Characteristics and associations of high-mitotic-rate melanoma. JAMA Dermatol. 2014;150(10):1048–1055.
  11. Gualandri L, Betti R, Crosti C. Clinical features of 36 cases of amelanotic melanomas and considerations about the relationship between histologic subtypes and diagnostic delay. J Eur Acad Dermatol Venereol. 2009;23(3):283–287.
  12. Vernali S, Waxweiler WT, Dillon PM, et al. Association of incident amelanotic melanoma with phenotypic characteristics, MC1R status, and prior amelanotic melanoma. JAMA Dermatol. 2017;153(10):1026–1031.
  13. Gong H-Z, Zheng H-Y, Li J. Amelanotic melanoma. Melanoma Res. 2019;29(3): 221–230. 
  14. 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. 
  15. Conroy RM. What hypotheses do “nonparametric” two-group tests actually test? The Stata Journal. 2012;12(2):182–190. 
  16. Sterne JAC, White IR, Carlin JB, et al. Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ. 2009;338:b2393. 
  17. Egger ME, Gilbert JE, Burton AL, et al. Lymphovascular invasion as a prognostic factor in melanoma. Am Surg. 2011;77(8):992–997. 
  18. Huvos AG, Shah JP, Goldsmith HS. A clinicopathologic study of amelanotic melanoma. Surg Gynecol Obstet. 1972;135(6):917–920. 
  19. Gershenwald JE, Scolyer RA, Hess KR, et al. 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. 
  20. Swetter SM, Tsao H, Bichakjian CK, et al. Guidelines of care for the management of primary cutaneous melanoma. J Am Acad Dermatol. 2019;80(1):208–250. 
Facebook
Twitter
LinkedIn