ÌÇÐÄvlog

BackgroundÌý Moles, or melanocytic nevi, are both markers of an increased risk of cutaneous melanoma and direct precursor lesions. Recent strategies to reduce the burden of advanced disease have focused on early detection and ongoing surveillance of moles for malignant degeneration. Inherent in this approach is the notion that moles exhibit a certain risk of transformation into melanoma; however, this risk is unknown.

ObjectiveÌý To estimate the risk of moles transforming into cutaneous melanoma.

DesignÌý We first constructed a model of transformation based on the assumption that the minimal number of moles turning into cutaneous melanoma per year is roughly equivalent to the number of melanomas diagnosed each year with associated nevic components. The annual risk was then calculated as the number of mole-associated melanomas diagnosed in 1 year (stratified by 10-year age groups) divided by the number of moles in a the same 10-year age group. We also estimated the cumulative risk during the lifetime of an individual mole by using a modification of the standard life table method.

ResultsÌý The annual transformation rate of any single mole into melanoma ranges from 0.0005% or less (ie, ≤1 in 200Ìý000) for both men and women younger than 40 years to 0.003% (about 1 in 33Ìý000) for men older than 60 years. The rate is similar between men and women younger than 40 years but becomes substantially higher for men older than 40 years. For a 20-year-old individual, the lifetime risk of any selected mole transforming into melanoma by age 80 years is approximately 0.03% (1 in 3164) for men and 0.009% (1 in 10Ìý800) for women.

ConclusionsÌý The risk of any particular mole becoming melanoma is low, especially in younger individuals. However, since moles can disappear, ones that persist into old age have an increased risk of malignant degeneration. For young people with innumerable moles and no other associated risk factors, systematic excision of benign-appearing lesions would be of limited benefit.

ALTHOUGH THE incidence of cutaneous melanoma (CM) has increased dramatically during the past several decades, earlier detection of lesions has led to a greater percentage of curable lesions. In 2002, the American Cancer Society estimated 53Ìý600 cases of CM with only 7400 deaths attributable to the disease.¹ Current strategies to minimize the morbidity and mortality associated with CM focus on sun protective behaviors and scrutiny of individuals with identifiable precursors and markers of risk, such as moles, also known as melanocytic nevi (MN).

The presence of multiple moles has been shown to be associated with a greater risk of CM. Retrospective case-control studies^2-9 have found a direct relationship between number of moles and risk of CM (roughly 2- to 14-fold). Many of the current efforts to curb the rise in melanoma have centered on the detection of early changes in these melanocytic nevi. Benign and atypical moles have been shown to exist in histologic contiguity with CMs, suggesting that these melanocytic proliferations are also susceptible to malignant transformation.^10-13 The ABCD (asymmetry, border irregularity/changes, color variation/changes, diameter increase) campaign and mole mapping approaches rely on the fundamental supposition that moles can transform into CMs over time. To date, however, this rate of transformation has not been calculated. Given the tremendous number of nevi in the general population and the comparatively limited incidence of CMs, a prospective cohort study to assess the rate of nevic transformation in an unbiased population would be difficult in terms of time and resources. We thus set out to calculate the rate of transformation of nevi into CM by means of population-based estimates of CM incidence and numbers of moles.

To approximate the transformation rate, we developed the model shown in Figure 1. We estimated the annual transformation rate (τ) to be τ = (MN_T-y/MN_US), where MN_T-y represents the number of MN transforming to CM per year in the United States and MN_US, the total number of MN in the United States. We make 2 assumptions in this calculation. First, the number of moles is relatively stable during the course of 1 year. Second, the estimated number of CMs arising in MN per year is the same as the number of MN that transform into CMs per year. This may be an underestimate, since some MN that become CM are obliterated in the process and are no longer detected on histologic examination. Thus, MN_T-y = CM_MN-y = ϕ_MN/CM × CM_US-y, where CM_MN-y is the number of CMs arising in MN per year in the United States; ϕ_MN/CM, the fraction of all CMs that have an associated mole; and CM_US-y, the number of CMs per year in the United States.

We stratified by sex and age groups (ie, <20 years, 20-29 years, 30-39 years, 40-49 years, 50-59 years, 60-69 years, ≥70 years), since the number of MN and risk of CM are different with respect to these 2 factors. Since the frequency of moles has not been carefully documented in the nonwhite population, we restricted our analysis to the white population.

To calculate Ï•_MN/CM, we performed a computer-assisted natural language search (SURGE Anatomic Pathology Software; Computer Trust Corporation, Boston, Mass) of all consecutive CMs diagnosed in a community-based general dermatopathology practice (Pathology Services Inc [PSI], Cambridge, Mass) between 1993 and 1997. During these years, 339Ìý886 specimens were evaluated at PSI, including 1615 invasive CMs. Criteria for the diagnosis of dysplastic melanocytic nevi (DMN) and for distinguishing small melanoma cells from nevus cells are outlined in Table 1 and Table 2, respectively. As a standard of practice, most CMs diagnosed at PSI are evaluated by at least 2 trained dermatopathologists. If there were disagreements, a senior dermatopathologist (Terence Harrist, MD, or Wallace Clark, Jr, MD) would be consulted to render a definitive diagnosis. A test set of 346 randomly selected melanoma cases that were reviewed by 2 pathologists at PSI showed only 5 disagreements in diagnosis (<2%). Discordant diagnoses and tumors in which precursor lesions were not clearly present or absent were excluded. We selected a community-based general dermatopathology service to minimize referral bias that may exist at tertiary care centers. We included only invasive CMs, with or without associated MN, in the current analysis, since accurate population-based age-specific rates of melanoma in situ are not available. Moreover, in 2 recent studies, the reliability of diagnosis of melanoma in situ has been challenged. Brochez et al¹⁵ reported that the DMN is the most common pigmented lesion to be misdiagnosed as melanoma, especially melanoma in situ. Furthermore, Megahed et al¹⁶ recently performed immunohistochemical analysis of 104 in situ melanomas and found that that nearly a third of the lesions were actually invasive melanomas.

To derive the CM_US-y by sex and age group, we used the sex- and age-specific incidences of CM in the white population available through the Surveillance, Epidemiology, and End Result study (1973-1997; available at: ) and the sex- and age-specific number of individuals in the white population on July 1, 1997, as projected by the US Census Bureau using the 1990 Census (available at: ; statistics available on request from the authors). Since our estimates are based on 10-year age groups while the Surveillance, Epidemiology, and End Result incidence is reported in 5-year age groups, we calculated the total number of CMs per year for each age group (CM_agegroup) by CM_agegroup = [Σ(CM_subagegroup × P_subagegroup)]/100Ìý000, where CM_subagegroupis the incidence of CM in each sub–age group and P_subagegroup is the number of white individuals in the sub–age group.

We used published population-based MN counts that stratified by age and sex to estimate MN_US. MacKie et al¹⁷ examined 432 (204 male, 228 female) healthy British nonhospitalized volunteers who did not have a personal or family history of CM. Four dermatologists counted the number of moles (defined as "brown pigmented lesions 3 mm or greater in diameter which is present throughout the year even in the absence of solar stimulation"). Nicholls¹⁸ examined 1518 (948 male, 570 female) individuals from Sydney, Australia, and counted pigmented lesions "2 or more mm in diameter and either palpable or seen to be deforming the surface architecture of the skin when viewed tangentially in a strong light"; seborrheic keratoses were excluded when encountered. Both studies stratified their mean MN counts into 10-year age groups that are compatible with our stratification. For some age groups that were further substratified (ie, <20-year and ≥70-year age groups), we calculated the mean number of MN (MN_agegroup) within these age groups by MN_agegroup = [Σ(MN_subagegroup × N_subagegroup)±Õ/ΣN_subagegroup, where MN_subagegroup and N_subagegroup are the mean MN count and individuals within the sub–age groups, respectively. We then calculated the MN_US, segregated by sex and age groups, by multiplying the sex- and age-specific MN_agegroup by the respective number of male and female white individuals in that age group within the US population. A transformation rate was calculated for both men and women in each age group.

Since the mean number of nevi per person exhibited a diminution over time, in any individual, a certain number of moles probably disappear with age. To calculate the probability of an individual mole developing into a CM, we made the following assumptions. First, individual moles on males disappear at a constant rate of 2388 per 100Ìý000 moles per year starting at age 20 years, and individual moles on females disappear at a constant rate of 3320 per 100Ìý000 moles per year, also starting at age 20 years. Second, the rate at which individual moles develop into CM is constant across 10-year age intervals. The rates of moles disappearing were estimated from the data by using the aggregate average number of moles for different age groups from the 2 studies.^17,18 The probability of an individual mole developing into a CM by a particular age was then calculated, separately for males and females, by using a modification of the standard life table method. Starting with a hypothetical population of 100Ìý000 moles at age 20 years, for each subsequent age group, the number of existing moles is estimated to be 100Ìý000 minus the cumulative number estimated to have disappeared by that age minus the cumulative number estimated to have developed into CM by that age. To estimate the number of moles that develop into CM at each age, we then multiplied the estimated yearly rate at which moles turn into CM for that age group by the expected number of moles remaining at that age. The cumulative rate per 100Ìý000 is then the cumulative number of moles that have turned into CM by a particular age.

We identified 1615 invasive CMs in the pathology database (863 males, 752 females; Table 3) that exhibited an unequivocal precursor status. Overall, approximately 26% (28% in males and 25% in females) of all invasive CMs harbored a recognizable MN on histologic examination. As expected, the total number of CMs increased with age; however, the Ï•_MN/CM decreased with age. Since the number of CMs from individuals younger than 20 years was limited (11 total), estimates for this age group may be unstable. Nevertheless, there is a clear downward trend for Ï•_MN/CM with advancing age. There is also a general trend for men to have more CMs associated with MN than women. In the age group 70 years and older, we found that 20% of CMs from men and only 9% of CMs from women exhibited a precursor MN lesion. We did not distinguish histologic subtypes of MN, since the population-based mole count studies^17,18 did not clinically differentiate between MN subtypes.

Using the Ï•_MN/CM derived from our histologic examination, we estimated the transformation rate of MN into CM with our model (Table 4). We used the 1993 to 1997 Surveillance, Epidemiology, and End Result CM incidence rates (white population) since this dataset provided the most complete figures for all age groups. We also used the 1997 projections for the white population obtained from through the US Census Bureau. For mean total mole counts, we used the MacKie et al¹⁷ (Glasgow, Scotland) and Nicholls¹⁸ (Sydney, Australia) assessments. Overall, there is general agreement from both reports in that children and young adults have higher MN counts than older individuals and the peak mean mole count occurs around the third decade. In both surveys, there was a decline in the number of moles after the third decade. About 16% of the healthy population show no evidence of any MN.¹⁷ With advancing age, there is a clear inverse relationship between the MN count and CM incidence (Table 4).

The estimated annual rate of transformation of any single MN into CM is very low (Figure 2). The transformation rate increases with age and ranges from 0.0005% or less (ie, ≤1 in 200Ìý000) for both men and women younger than 40 years to 0.003% (about 1 in 33Ìý000) for men older than 60 years. The rate is similar between men and women younger than 40 years, but becomes substantially higher for men after age 40 years.

We then estimated the cumulative risk of any given mole transforming into CM. This was done by assuming that the number of moles and the incidence of melanoma vary across 10-year age groups but do not fluctuate significantly within the 10-year blocks. For instance, as shown in Table 5, the lifetime risk of a mole (on a 20-year-old individual) transforming into CM by age 80 years is approximately 0.03% (1 in 3164) for men and 0.009% (1 in 10Ìý800) for women. The estimated lifetime risk of malignancy is greatest (approximately 0.05%, or 1 in 2000) for a mole that is found on a male patient in his fifth decade of life.

We present estimates for the annual rate and cumulative risk of transformation of moles into CM. These rates change with respect to age because of 3 dynamic variables: fraction of melanomas associated with MN, incidence of CM, and mean number of MN. The annual risk of any one MN becoming malignant is approximately 0.0005% (or less than 1 in 200Ìý000) before age 40 years. This annual rate increases considerably during the ensuing decades to peak at approximately 0.003% (or 1 in 30Ìý000) for men older than 60 years. Although MN precursors are common in CM from younger patients, the low incidence of CM among the younger population and the higher total MN count contributes to a lower calculated rate of transformation for persons younger than 40 years. For older men, the increase in the calculated transformation rate results from the tremendous increase in CM incidence and large decrease in MN count with only a partial reduction in the fraction of CMs associated with precursor MN (Table 4).

For a 20-year-old individual, the estimated lifetime risks of any given mole becoming melanoma by age 80 years are about 1 in 3165 (100Ìý000/31.6) for males and 1 in 10Ìý800 (100Ìý000/9.26) for females (Table 5). Since some moles will disappear with age, not all moles will "survive" to realize malignant potential. This attrition leads to a slight increase in cumulative risk for moles detected on persons between ages 40 and 60 years.

Histologically, MN precursors can be identified in 28% and 25% of CMs from males and females, respectively. These percentages may be an underestimate of the actual precursor frequency, since MN precursors may be obliterated during tumor growth. Nevertheless, they are compatible with several other reports that observe frequencies ranging from 21% to 35%.^10,13,19 What is interesting, however, is the interaction between age and MN precursors. We found that for patients younger than 30 years, the percentage of CMs associated with an MN precursor is 50% or more, while for patients older than 60 years, this percentage drops to approximately 20%. The higher prevalence of MN among the younger population^13,17,18 is consistent with a greater percentage of CMs arising in MN. However, since most CMs in the population occur in the elderly without MN precursors, our results also suggest a possible etiologic mechanism whereby early-onset disease is related to an altered substrate, such as an MN or congenital nevus, whereas late-onset disease results from cumulative sun damage to non-MN melanocytes. Consistent with this hypothesis is the detection of oncogenic genetic lesions, such as N-ras mutations, in nevi not associated with CM.²⁰ However, more definitive analyses, such as genetic characterization of tumors, are required to substantiate this hypothesis.

We stratified our analysis by both age and sex because MN count and risk of melanoma vary significantly with respect to these demographic features. However, other factors that may influence the transformation rate were not included in our model. First, cumulative sun exposure may influence the number of MN. A recent study suggests that genetic influences may determine the emergence of MN, but sun exposure may influence the mean number.²¹ The 2 studies used in our analysis evaluated populations similar in origin (ie, Northern Europe and United Kingdom) but very different in sun exposure (ie, Glasgow, Scotland, and Sydney, Australia). Ambient sunlight may explain the higher mean MN count observed for males younger than 20 years in Australia¹⁸ compared with youths in the United Kingdom.¹⁷ Since the amount of sun exposure in the United States most likely falls within the range defined by Glasgow and Sydney and the mean total MN counts are similar (range, 5-33) between the 2 studies, we think that the estimates are appropriate for a healthy, unselected US white population. Second, anatomic variation in mole density may influence the likelihood of malignant degeneration. Although some studies report a higher MN density on intermittently sun-exposed sites (upper arms, upper back) compared with sun-protected sites (lower back, buttocks) and chronically irradiated sites (head and neck),^18,22-24 the relationship between MN density and the development of CM is unclear. For instance, the lower extremities of women have a relatively low density of MN^22,23 despite a higher risk of CM compared with other sites. Confounding factors that can account for the observed irregular correlation include the possibility of biologically distinct subsets of MN, differences in the pattern of sun exposure, and discrepancies in the amount of sunscreen applied to various body areas. Moreover, in our own histologic evaluation of specimens, anatomic designations are often ambiguous (eg, right shoulder). Thus, we did not stratify by anatomic site, to minimize introduction of these confounding factors.

It is likely that transformation rates among different types of moles, especially DMN, are not uniform. We developed our initial model with the use of unbiased total mole counts, since judgments on border or color irregularities used to clinically identify and count DMN are subject to interpretation. Our results, along with other published reports,^10,11,19,25 however, suggest that malignant transformation is not limited to DMN, since benign and congenital nevi can be observed in nearly half of the cases of melanomas with associated nevi. Of the 425 CMs with an MN fragment, we found that 43% of these MN precursors (or 11% of all invasive melanomas) had histologic findings consistent with DMN (see Table 1 for criteria). Although the prevalence of dysplastic nevi or atypical moles has been reported for the general population, the actual age-specific density of DMN in a given individual has not been accurately documented. Thus, it is not possible in our model to calculate transformation rates segregated by 10-year age groups. However, 2 studies have determined an overall mean DMN count of approximately 0.7 per person in healthy control populations as part of case-control series examining melanoma risk among individuals with DMN.^2,26 Using 0.7 as a rough mean number of DMN per person (male or female) and 0.11 as the fraction of invasive melanomas associated with DMN (Ï•_DMN/CM), we estimate an annual DMN transformation rate of 1 in 30Ìý089 moles for males and 1 in 39Ìý809 moles for females. We can also modify Ï•_DMN/CM on the basis of published figures. In series with at least 100 cases, the range of reported Ï•_DMN/CM appears to fall between 0.05²⁷ and 0.34.²⁸ One potential source of disparity between these 2 Ï•_DMN/CM figures is in the interpretation of the pathologic findings. For instance, Massi et al²⁷ interpreted the presence of an increased number of atypical melanocytes at the periphery of the melanoma as part of the tumor, while others, including us, consider these changes to represent residual DMN. Regardless of the interpretation, these values most likely reflect the possible range of Ï•_DMN/CM and, when applied to our model (without age stratification but as total population), yield a high DMN transformation estimate of 1 in 9735 and 1 in 12Ìý880 moles per year for males and females, respectively, and a low estimate of 1 in 66Ìý196 and 1 in 87Ìý580 moles per year for males and females, respectively.

In summary, clinicians engaged in the surveillance of moles for early detection of CMs are often confronted with the question of how likely it is that any given mole will turn into a melanoma. For a young patient, we can now say that this rate is approximately 1 in 3165 for a man and 1 in 10Ìý800 for a woman during the course of his or her lifetime. Our results have significant clinical implications. Some healthy individuals request indiscriminant removal of MN to minimize the likelihood of CM. Given our calculations, the overall risk reduction of such surgical eradication is probably limited, since very few lesions excised would ever have progressed to become CMs. Furthermore, in an effort to detect malignant degeneration within MN, photographic documentation ("mole mapping") of all moles on young, low-risk individuals will probably be associated with high cost and a low yield of detection. Targeted mapping of older individuals (at least older than 30 years) and patients with factors traditionally associated with an increased CM risk (eg, excessive sun exposure, family history, previous melanomas, or atypical MN) may be more effective.

Corresponding author and reprints: Hensin Tsao, MD, PhD, Department of Dermatology, Massachusetts General Hospital, Bartlett 622, 48 Blossom St, Boston, MA 02114.

Accepted for publication July 25, 2002.

This study was supported in part by grant CRTG 99-249-01-CCE from the American Cancer Society, Atlanta, Ga (Dr Tsao).