Association of Patient Risk Factors and Frequency of Nevus-Associated Cutaneous Melanomas

Importance  The reported frequencies of associations between primary cutaneous melanomas and melanocytic nevi vary widely between 4% and 72%. However, earlier histopathologic studies were limited by their retrospective design and did not assess the influence of important patient-related risk factors.

Objectives  To identify the frequency of nevus-associated melanomas and correlate patient- and melanoma-related factors.

Design, Setting, and Participants  A prospective, single-center, observational study with systematic documentation of melanoma risk factors, clinical and dermoscopic criteria of excised lesions, and results of histopathologic examination was conducted at a university-based dermatology clinic. Participants included 832patients at high risk for developing melanoma. Evaluation was performed at regular intervals between April 1, 1997, and May 31, 2012, and data analysis was conducted between September 1, 2012, and December 31, 2013.

Main Outcomes and Measures  Assessment of the frequency of nevus-associated melanoma and the influence of patient- and melanoma-related factors on their manifestation.

Results  During the study, 190 melanomas (81 [42.6%] in situ and 109 [57.4%] invasive) were diagnosed in 113 of the 832 patients (13.6%); there were 42 women (37.2%) and 71 men (62.8%). The median (SD) Breslow thickness of invasive melanomas was 0.42 (0.43) mm. Histopathologic examination revealed remnants of melanocytic nevi in 103 melanomas (54.2%). Most nevus-associated melanomas were found on the trunk (67 [65.1%]); however, statistical significance for the localization was not present (P = .06). In univariate analyses, reported as odds ratios (95% CIs), nevus-associated melanomas were found significantly more frequently in patients of lower melanoma risk (risk group 1 [>50 common and/or ≤3 atypical nevi], 2.75 [1.14-6.64]; P = .02), with more than 100 nevi (1.63 [1.02-3.60]; P = .04), or with the diagnosis of in situ melanoma (14.01 [6.14-31.96]; P < .001). In contrast, nevus-associated melanomas were found significantly less frequently in patients with 1 or more previous melanomas (0.28 [0.21-0.83]; P = .005). All other factors (eg, age, skin type, hair color, and melanoma thickness) showed no significant influence on the manifestation of nevus-associated melanomas. These observations were confirmed in a separate analysis including all 109 invasive melanomas. Multivariate regression analysis identified 3 independent patient-related factors (high nevus count, low risk for melanoma, and female sex) and 1 melanoma-related factor (in situ melanoma) to be indicative of a significantly increased probability of nevus-associated melanomas.

Conclusions and Relevance  In this prospective study of a high-risk patient cohort, 54.2% of primary melanomas were associated with melanocytic nevi. Patients with many nevi and without previous melanomas or traits of familial atypical mole and multiple melanoma syndrome had a higher frequency of nevus-associated melanomas. These patients could thus benefit from sequential digital dermoscopy in addition to total-body photography.

During the past few decades, melanoma has emerged as an important public health challenge.¹ Steadily increasing incidence rates and disturbing trends in melanoma mortality have sparked a heightened commitment to early detection and prevention.² However, financial and personnel resources for melanoma screening are limited. The long-term follow-up of high-risk patients by sequential digital dermoscopy and total-body photography are examples of procedures that are time consuming and demand highly trained personnel.^3,4 To increase the effectiveness of long-term surveillance, a previous study⁵ suggested selection of patients by a detailed assessment of melanoma risk factors. However, after selection of suitable high-risk patients, the correct selection of lesions for further digital follow-up remains controversial. Is it reasonable to apply sequential digital dermoscopy on large numbers of preexisting melanocytic lesions in high-risk patients? What is the frequency of nevus-associated melanomas in high-risk patients? Which patient-related or melanoma-related factors influence the incidence of nevus-associated melanomas?

The presence of many common or atypical acquired nevi is an accepted factor indicating an increased risk for the development of melanoma.⁶ In contrast, the rate at which benign melanocytic lesions may also act as melanoma precursors remains a subject of debate.⁷ In several retrospective studies,^8-12 between 18% and 85% of patients with melanoma reported preexisting pigmented lesions at the site of melanoma. This information is of little help since both a benign nevus and a slowly progressing early melanoma could have been noticed by the patients. Friedman et al¹² compared clinical and histologic reports of patients with melanoma and found that only 70% of patients showing histologic evidence of a nevus reported a preceding pigmented lesion, and 35% of patients without histologic evidence erroneously recalled a preceding mole. A wide range of frequencies for nevus-associated melanoma was reported in histopathologic studies^9,13-20 assessing primary cutaneous melanomas, with 4% to 72% of the cases showing nevus cells in association with a melanoma. For several reasons, the histopathologic approach also might not be completely reliable: (1) in studies including patients with high nevus counts, a cutaneous melanoma may have invaded an adjacent nevus coincidentally, (2) nevocytic remnants might have been overgrown by more advanced melanomas, (3) studies prospectively assessing melanomas for preexisting moles on serial sections may reveal higher frequencies than retrospective studies based on pathology reports, (4) studies mainly including melanomas of the superficial spreading histotype may reveal higher frequencies because associated nevi were observed most often in superficial spreading melanoma,^18,21 and (5) differing morphologic criteria used for the definition of nevus-associated cutaneous melanoma may influence the frequency of the diagnosis. The aim of the present prospective study was to analyze the frequency of nevus-associated melanomas as well as critical patient- and melanoma-related factors in a precisely characterized cohort of high-risk patients monitored by the use of long-term sequential digital dermoscopy and total-body photography.

The study was approved by the local ethics committee at the University Medical Center in Göttingen, Germany, and performed in accordance with the Declaration of Helsinki principles.²² Participants provided written informed consent for all invasive surgical procedures and oral consent for participation in the program. There was no financial compensation. All patient-related data were routinely acquired as part of the clinical surveillance program of high-risk patients. Data were deidentified for use in the study. Patients’ characteristics, including age, sex, hair color, eye color, skin type (Fitzpatrick I-IV classification), presence of ephelides, history of previous melanoma, nevus count, and the number of atypical nevi, were prospectively entered into a database (Microsoft Access, version 10; Microsoft Corp). Patients were assigned to 3 risk groups depending on the presence of melanoma risk markers. Group 1 included patients with multiple (>50) common and/or 3 or fewer atypical nevi.²³ Group 2 included patients with atypical mole syndrome (defined as ≥50 common and >3 atypical nevi).²⁴ Group 3 participants had familial atypical mole and multiple melanoma (FAMMM) syndrome (defined by a family history of at least 2 melanomas in first- or second-degree relatives and the presence of atypical mole syndrome criteria).²⁵ Data on 832 patients and 14 896 monitored lesions with at least 1 follow-up examination were collected between April 1, 1997, and May 31, 2012.

During their first appointment, patients underwent a head-to-toe visual examination including the use of a nonpolarizing, handheld dermatoscope (Heine Delta 20; Heine Optotechnik). A 7-point checklist algorithm was used to differentiate benign melanocytic lesions from melanomas.²⁶ Lesions suspected to be malignant (score ≥3 points) were excised immediately. Lesions scoring less than 3 points but presenting clinical or dermoscopic criteria of atypia²⁷ were marked on digital overview images and were electronically stored (FotoFinder medicam 800HD; FotoFinder Systems GmbH).

During follow-up appointments, patients were asked about new or changing nevi, and digital overview images (total-body photography) were compared with the corresponding body surface. Nevi that had been previously documented were compared with the baseline image on a split screen.

During the study, 1770 melanocytic lesions were excised to either rule out or confirm the diagnosis of melanoma. All lesions were submitted to a highly standardized process of histopathologic assessments requiring agreement of at least 2 independent dermatopathologists to arrive at a diagnosis. In equivocal cases, a third experienced dermatopathologist was consulted. A total of 190 cutaneous melanomas were diagnosed using serial hematoxylin-eosin–stained sections. Immunohistochemical stainings, including S100-, Melan-A–, HMB-45–, and Ki-67–directed antibodies, were performed on all melanocytic tumors suspected as being malignant. To differentiate benign nevus cells from melanoma cells, we applied histopathologic criteria adapted from Elder et al⁹ and Friedman et al.¹² A diagnosis of a melanoma in association with a nevus was made whenever an abrupt transition between malignant cells and benign melanocytic cells became apparent, with the latter showing maturation and a lack of cytologic atypia.^14,16 In addition, we studied the immunohistochemical staining patterns of melanoma and nevus cells within the nevus-associated tumors, which demonstrated a loss of cytoplasmic reactivity for HMB-45 and a lack of proliferation by negative staining for Ki-67 in nevus cells. The nevus remnants were not further classified. In accordance with a report by Stolz et al,¹³ a diagnosis of nevus-associated melanoma was made when (1) benign nevus cells at the dermoepidermal junction were present contiguously to the melanoma, (2) nevus cells were present in the dermis below the melanoma, or (3) a combination of the 2 aforementioned patterns occurred.

Immunohistochemical staining was performed by applying the streptavidin-biotin complex method with alkaline phosphatase as the labeling enzyme and fast red chromogen as the substrate (detection kit K5005; Dako). The following antibodies (all obtained from Dako) were used: S-100 (clone, S-100; dilution, 1:3000), HMB-45 (clone, HMB45; dilution, 1:200), Melan-A (clone, A 103; dilution, 1:300), and Ki-67 (clone, MIB-1; dilution, 1:100). All antibodies were stained by an auto-immunostainer (Immunostar 80, Shandon Varistain 24–4; Thermo Shandon Ltd).

Plausibility, accuracy, and completeness of stored data were verified. Nevus-associated melanomas and melanomas without nevus association were compared using univariate and multivariate analyses. We applied logistic regression models, using each patient’s identification number as the random variable. We accounted for potential intraindividual correlations in the case of multiple melanomas by using the generalized estimating equation method to estimate the model variables.

All recorded melanoma risk factors were entered into multivariate logistic regression analysis as independent variables. To attain the best-fitting models (in terms of lowest quasi-Akaike information criterion), we performed a forward selection strategy, thereby determining which markers remained significantly associated with the diagnosis of nevus-associated melanoma while simultaneously adjusting for other criteria included in the regression model. For univariate analyses of the melanoma detection patterns, the categorical variables were investigated using χ² tests.

The results of univariate and multivariate modeling were expressed as odds ratios (ORs) with 95% CIs, and P < .05 was considered to indicate a significant difference. The statistical analyses were conducted from September 1, 2012, to December 31, 2013, and performed using SAS, version 9.3 (SAS Institute Inc).

During the study, 190 melanomas were diagnosed in 113 of the 832 high-risk patients (13.6%). In 68 patients (60.2%), a single melanoma was diagnosed; in 45 patients (39.8%), more than 1 melanoma occurred during the course of the study. There were 2 melanomas detected in 26 patients, 3 in 11 patients, 4 in 5 patients, and 5, 6, and 7 in 1 patient each. Forty-two patients (37.2%) were female and 71 patients (62.8%) were male (Table 1). The median (SD) age at the time of the first melanoma excision was 42.2 (14.3) years, and women were significantly younger than men (37.6 vs 46.8 years; P < .001). Many patients showed features of skin type I (23 [20.4%]) or II (74 [65.5%]); most had blond (52 [46.0%]) or brown (48 [42.5%]) hair and blue-gray (59 [52.2%]) eye color. Clinical examination revealed ephelides and/or freckles in 29 patients (25.7%). Other important melanoma risk factors included a high nevus count (≥50 common nevi in 87 patients [77.0%]), high median (SD) count of atypical nevi (23 [12.6]), previous melanoma (79 patients [69.9%]), or family history of melanoma in first- or second-degree relatives (15 [13.3%]).

Of 190 melanomas detected, 81 (42.6%) were in situ and 109 (57.4%) were invasive. In 103 melanomas (54.2%), remnants of melanocytic nevi were documented histopathologically. The median (SD) Breslow thickness of all invasive melanomas was 0.42 (0.43) mm. All but 7 melanomas (183 [96.3%]) showed a thickness of 1 mm or less. Superficial-spreading melanoma was the predominant melanoma histotype. Only 3 melanomas (1.6%) demonstrated either a lentigo maligna (n = 2) or lentigo maligna melanoma (n = 1) pattern, and none of these were nevus associated.

A total of 18 factors potentially associated with manifestation of nevus-associated melanomas were investigated. In univariate statistical analyses, nevus-associated melanomas were found significantly more frequently when associated with in situ melanomas comparedwith invasive melanomas (OR, 14.01 [95% CI, 6.14-31.96]; P < .001) and in patients with more than 100 nevi compared with 100 or fewer nevi (OR, 1.63 [95% CI, 1.02-3.60]; P = .04) (Table 2). Nevus-related melanomas were mostly localized on the trunk (67 [65.1%]) and least frequently on the head and neck area (4 [3.9%]). Only 6 (5.8%) nevus-associated melanomas were found on the upper limbs compared with 13 (14.9%) de novo melanomas; however, statistical significance for the localization was not present (P = .06).

Patients with nevus-associated melanomas reported 1 or more previous melanomas significantly less frequently (67 [65.0%]) than did patients with de novo melanomas (71 [81.6%]; OR, 0.28 [95% CI, 0.21-0.83]; P = .005). At the same time, 25 (24.3%) patients with nevus-associated melanomas were categorized as melanoma risk group 1 (low risk, with >50 common and/or ≤3 atypical nevi) vs only 12 (13.8%) patients with de novo melanomas (OR, 2.75 [95% CI, 1.14-6.64]; P = .02). All other factors (eg, tumor thickness, age, sex, skin type, hair color, and family history of melanoma) showed no significant influence on the incidence of nevus-associated melanomas.

We performed additional univariate analyses including only invasive melanomas (n = 109) to rule out a possible bias introduced by difficulties to distinguish in situ melanoma from nevus-associated in situ melanoma. Within this separate data set, nevus-associated invasive melanomas (n = 43) were mostly located on the trunk (25 [58.1%]), and all of the above-mentioned variables retained their statistical significance. Moreover, a family history of 1 or more melanomas in first- or second-degree relatives pointed to a significantly lower frequency of nevus-associated invasive melanomas (OR, 0.32 [95% CI, 0.07-0.94]; P = .03).

A forward selection strategy was performed to produce the best-fitting model and determine which markers remained significantly associated with the diagnosis of nevus-associated melanoma while simultaneously adjusting for other criteria included in the regression model. We identified 6 informative factors indicating a melanoma to be nevus associated (Table 3).

First, the absence of melanoma invasiveness significantly increased the probability of the manifestation of nevus-associated melanoma (OR, 20.25 [95% CI, 7.52-54.58]; P <.001). Next, a risk group stratification to the lowest study risk group (group 1, patients with multiple common [>50] and/or ≤3 atypical nevi: OR, 4.73 [95% CI, 1.36-16.50]; P = .008) was correlated with a significant increase of nevus-associated melanomas. In addition, a high total number of nevi (OR, 3.51 [95% CI, 1.27-9.70]; P = .002) and female sex (OR, 3.00 [95% CI, 1.20-7.51]; P = .002) were found to significantly influence the rate of nevus-associated melanomas.

The multivariate model was further improved by including information about the presence or absence of a FAMMM syndrome (P = .13) and hair color (P = .16), although the influence of each of these factors alone was not statistically significant. For the sample at hand, tumor thickness within invasive melanomas was not found to be associated with the presence of a nevus (P = .39).

Most of the 190 melanomas were detected at follow-up examinations (145 [76.3%]) (Table 4). De novo melanomas were detected more frequently at the first examination (30 [34.5%]) than were nevus-associated melanomas (15 [14.6%]; P = .001). The largest fraction of nevus-associated melanomas was excised because of combined clues from conventional dermoscopy as well as digital dermoscopy follow-up (45 [43.7%]); most de novo melanomas were found without information from follow-up examinations exclusively by conventional dermoscopy (43 [49.4%]). A comparable fraction of nevus-associated (14 [13.6%]) and de novo (16 [18.4%]) melanomas was detected exclusively by digital dermoscopy on follow-up (Table 4). Neither of the nevus-associated melanomas detected at follow-up examinations was reported as a new lesion by the patient or by comparison of total-body photographic images. When we focused on the 145 melanomas (76.3%) that were excised at follow-up examinations, subtle lesional changes were found in 59 of 88 (67.0%) nevus-associated melanomas and in 37 of 57 (64.9%) de novo melanomas by digital dermoscopy follow-up.

The incidence of cutaneous melanoma has risen steadily in most fair-skinned populations worldwide and is predicted to increase even more for at least another 2 decades.² Within melanoma detection programs, the dermoscopic follow-up of high-risk patients has proved to be effective but requires a high level of financial and personnel resources.²⁸ Hence, more knowledge about the frequency of nevus-associated melanomas in high-risk patients might help to optimize surveillance strategies. To this end, we (1) prospectively assessed histopathologic specimens on serial sections for the presence of nevus-associated melanoma, (2) prospectively documented patient characteristics and melanoma risk factors, (3) included only patients at increased risk for melanoma, and (4) monitored patients in a melanoma surveillance program by total-body photography and sequential digital dermoscopy.

Of 1770 excised lesions, 190 were histologically confirmed to be melanoma, and we found remnants of melanocytic nevi in 54.2% of melanoma cases. Several lines of explanation, which are not mutually exclusive, can be delineated for this high rate of nevus-associated melanomas. First, nevus-associated melanomas were reported^{14,15,19,29,30} to be more frequent in thinner tumors, possibly owing to the obliteration of nevocytic remnants in more advanced melanomas. Indeed, most melanomas in our study were in situ or very early invasive melanomas (only 6 melanomas were >1 mm in thickness), which may have favored the detection of nevus-associated melanomas. Second, nevus association has been reported^15,17,18 most frequently in cases of superficial spreading melanoma. In our study, all but 3 melanomas showed a superficial spreading melanoma histotype, which may have contributed further to the high frequency of nevus-associated melanomas. Third, the prospective study setting with a more thorough assessment of possible nevus remnants in melanomas may have increased the probability of detecting nevus-associated melanomas in comparison with previous studies^9,13,20 based on retrospective analyses of pathology reports in which such descriptive details might have been omitted. Finally, the presence of many common or atypical nevi in most of our patients may, by chance, increase the probability of melanoma developing in contiguity of a preexisting nevus and, thus, potentially leading to a false-positive diagnosis of nevus-associated melanoma. However, based on the dermoscopic images of each excised lesion, any such random effects can be ruled out because most nevus-associated melanomas developed dynamic changes within a preexisting nevus during dermoscopic follow-up. We did not observe dermoscopic cases of suspicious (melanoma) lesions growing into an unchanged neighboring nevus.

Our observation that nevus-associated melanomas preferentially develop on the trunk and least frequently on the head and neck is in line with earlier reports.^17,18 Moreover, our finding of a significantly higher association with in situ melanomas extends the results of Marks et al³¹ and Sagebiel,¹⁵ who found an inverse relationship between the presence of benign nevus cells and melanoma thickness. However, when focusing exclusively on invasive melanomas, we and other investigators¹⁸ found no significant influence of melanoma thickness. This difference might be explained by the low median thickness of invasive melanomas (0.42 mm) in our study.

The phenotypic and genotypic traits of the patients in the present study had a significant effect on the frequency of nevus-associated melanomas. The number of common and atypical nevi is a known melanoma risk indicator.⁶ Since our analyses included the comparison of nevus-associated melanoma with de novo melanoma, the significant association of a high nevus count (>100 common nevi) with the presence of nevus-associated melanoma points to a relatively higher relevance of this risk indicator for nevus-associated melanoma. The results of a case-control study²⁹ confirmed this interpretation by reporting a higher risk for nevus-associated melanoma (OR, 14.4) than for nevus-nonassociated melanoma (OR, 4.7) in patients with many common nevi. These findings point to a role of acquired melanocytic nevi not only as a marker of melanoma risk but also as a marker for the development of nevus-associated melanomas. If confirmed in larger studies, this observation may have crucial implications on the design of future strategies for early melanoma detection. The findings could potentially lead to recommending a more thorough surveillance of preexisting melanocytic nevi in patients with high nevus counts.

Some authors^17,18 have demonstrated that the percentage of nevus-associated melanomas decreases with increasing age. In our study, the mean age at the time of first melanoma diagnosis was younger in cases of nevus-associated melanomas compared with de novo melanomas; however, no statistical significance could be demonstrated.

Patients with FAMMM syndrome carry an exceedingly high risk for melanoma development.^32-34 In our study, patients with FAMMM syndrome and patients with multiple melanomas developed nevus-associated melanoma less frequently compared with patients of the lower risk group (multiple common nevi and no previous melanoma). It is conceivable that the development of de novo melanomas is favored by genetic factors underlying FAMMM syndrome or the manifestation of multiple melanomas. Genetic analyses of patients with FAMMM syndrome often revealed germline mutations in the CDKN2A tumor-suppressor gene.^32,35 Family members who also presented an atypical mole syndrome phenotype carried the CDKN2A gene mutation 3 times more frequently than their relatives without atypical mole syndrome.³⁶ All significant variables for developing nevus-associated melanomas were confirmed in a separate subgroup analysis including only invasive melanomas (n = 109), thus ruling out a potential bias caused by a high proportion of in situ melanomas within the overall data set.

Possible limitations of our study are related to its setting, in which we selected patients at increased melanoma risk for long-term follow-up. Such a selection bias could make it difficult to present more generalized statements.

In this prospective study, 54.2% of the cutaneous melanomas identified were associated with melanocytic nevi. We were able to demonstrate that patients who develop nevus-associated melanomas show a distinct risk profile that differs from that of patients with de novo melanomas. Patients with many acquired nevi on the trunk appear to be at high risk for developing nevus-associated melanomas and may, therefore, particularly benefit from sequential digital dermoscopy in addition to total-body photography.

Corresponding Author: Holger A. Haenssle, MD, Department of Dermatology, Venereology, and Allergology, University Medical Center, Ruprecht-Karls-University Heidelberg, Im Neuenheimerfeld 440, D-69120 Heidelberg, Germany (holger.haenssle@med.uni-heidelberg.de).

Accepted for Publication: August 20, 2015.

Published Online: November 4, 2015. doi:10.1001/jamadermatol.2015.3775.

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

Study concept and design: Haenssle, Buhl.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Haenssle, Bertsch.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Haenssle, Mograby, Ngassa, Rosenberger.

Administrative, technical, or material support: Haenssle, Emmert, Schön, Bertsch.

Study supervision: Haenssle, Buhl, Schön.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported in part by a grant from the Cancer Society of Lower Saxony (Dr Haenssle).

Role of the Funder/Sponsor: The Cancer Society of Lower Saxony 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.

Additional Contributions: Parts of this work represent data from the doctoral thesis of Drs Mograby and Ngassa.