Improvement of Genetic Testing for Cutaneous Melanoma in Countries With Low to Moderate Incidence

Key Points

Question

Is it necessary to update guidelines for genetic testing for melanoma susceptibility genes in countries with low to moderate incidence?

Findings

The mutation rates in a series of unselected patients tested between 2004 and 2015 were 67 of 1032 (6.5%) and 32 of 263 (12.1%) in a subgroup younger than 40 years at first melanoma based on the rule of 2, and 38 of 408 (9.3%) based on the rule of 3.

Meaning

In countries with a low to moderate incidence of melanoma, we propose using the rule of 3, except in families and patients with a first melanoma before the age of 40 years, for whom the rule of 2 should be maintained.

Abstract

Importance

Genetic testing for melanoma-prone mutation in France, a country with low to moderate incidence of melanoma, is proposed in cases with 2 invasive cutaneous melanomas and/or related cancers in the same patient, or in first- or second-degree relatives (rule of 2). In preclinical studies, these rules led to disclosure of mutation(s) in more than 10% of these families, the threshold widely accepted to justify genetic testing for cancers.

Objective

To reconsider these criteria in a general population testing of patients.

Design, Setting, and Participants

This was a retrospective study, performed from 2004 to 2015 at Angers and Lyons University Hospitals, of a cohort of 1032 patients who underwent genetic testing.

Main Outcomes and Measures

Frequency of mutation in high (CDKN2A, CDK4, and BAP1) and intermediate (MITF) susceptibility genes; statistical effect of histologic subtype, age, dysplastic nevi syndrome, and associated cancers on mutation rate; and evaluation of cases with anamnestic uncertainty.

Results

The mutation rate was 67 of 1032 patients (6.5%). Their mean (SD) age was 54.5 (14.2) years [range, 18-89 years], and 543 (52.6%) were men. It increased to 38 of 408 patients (9.3%) when applying a rule of 3 (those with ≥3 primary melanomas or genetically related cancers) (P = .68) and to 27 of 150 patients (18.0%) with a rule of 4 (4 primary melanomas or related cancer) (P < .001). The impact of age at first melanoma was observed only in those younger than 40 years, with a rate of 32 of 263 (12.1%) (P = .12) for the rule of 2 and 22 of 121 (18.2%) (P = .001) for the rule of 3. Use of the rule of 2 in patients younger than 40 years reduced the number of missed CDKN2A-mutated-families when applying the rule of 3 from 14 of 43 to 7 of 43. Anamnestic uncertainty, found in 88 families (8.5%), if excluded, would have led us to withdraw of only 21 cases (23.8%), and only 1 mutation would have been missed.

Conclusions and Relevance

We propose using the rule of 3 to recommend genetic testing in France and countries with low to moderate incidence of melanoma, except in families and patients with a first melanoma occurrence before age 40 years in whom the rule of 2 could be maintained.

This study evaluates the efficiency of using the “rule of 3” vs the widely accepted “rule of 2” for the detection of melanoma-prone mutations in France, a country with low to moderate melanoma incidence.

Introduction

Cutaneous melanoma (CM) incidence continues to increase in the white population, representing 2.7% of the total incidence of cancer in France, where the standardized incidence of melanoma is 10.8 per 100 000 inhabitants in males and 11.0 per 100 000 in females. An epidemiological study has shown that both patient characteristics (fair pigmentation phenotype, high number of melanocytic nevi, and presence of dysplastic nevi syndrome [DNS]) and environmental factors influence the risk of melanoma; a family history positive for melanoma represented the most important risk factor. Five principal susceptibility genes have been identified, comprising high-risk susceptibility genes: CDKN2A (cyclin-dependent kinase inhibitor 2A), CDK4 (cyclin-dependent kinase inhibitor 4), and BAP1 (breast cancer 1 [BRCA1] associated protein 1); an intermediate-risk susceptibility gene, MITF (microphthalmia-associated transcription factor), and an intermediate- to low-risk susceptibility gene, on allelic variants, MC1R (melanocortin 1 receptor).

Mutation frequency rates of CDKN2A can range from 5% to 72% depending on the selection criteria used. In Europe, the probability of developing a melanoma for CDKN2A carriers reaches 13% at age 50 years and 58% at age 80 years. This risk is modulated by other genetic risk factors, such as associated variants of MC1R, and by sun exposure, mainly depending on the latitude. An excess of nonmelanoma cancers has been observed in CDKN2A carriers, such as exocrine pancreas adenocarcinoma and central nervous system (CNS) tumors.CDK4 was identified as the second high-risk melanoma susceptibility gene, described in 18 melanoma-prone families to date. Lifetime risk for developing melanoma in CDK4 carriers is similar to that of CDKN2A mutation carriers. BAP1 germline mutations have been associated with multiple tumors (including cutaneous and ocular melanomas, mesothelioma, and renal clear cell [RCC] carcinoma) and could be considered as a mutation indicating high melanoma susceptibility. Recently, a recurrent p.Glu318Lys oncogenic germline mutation was identified in the MITF gene, an intermediate-to-medium penetrance melanoma susceptibility allele.

MC1R is a low- to moderate-risk gene. To date, genetic testing for its mutations is considered to be of uncertain clinical usefulness because the associated cancer risk is too small to formulate recommendations for screening or prevention.

Given the high probability for developing melanoma in the population with a CDKN2A/CDK4 mutation and the favorable impact of early detection on the prognosis, genetic counseling can be offered to families likely to be carrying a melanoma-prone mutation. According to the first American Society of Clinical Oncology (ASCO) guidelines published in 1996, criteria to justify BRCA1 testing in patients with breast cancer were chosen to reach a pretest probability of more than 10% positivity. This cutoff was afterward used widely to ascertain a minimal cost-efficiency in cancer genetic testing. Thereafter, different guidelines for genetic testing for melanoma were proposed, depending on overall incidence in different countries. In France, a country with a low to moderate incidence of melanoma, individuals with personal history of at least 2 primary CMs or with at least 1 other affected first- or second-degree relative are considered as appropriate candidates to test for CDKNA2, CDK4, BAP1, MITF, and MC1R gene mutations (rule of 2). In 2014, under the auspices of the French National Institute of Cancer and the French Society of Dermatology, guidelines in France approved the rule of 2 and proposed considering only invasive melanoma diagnosed before age 75 years. Taking into account the other cancers observed in syndromes associated with melanoma-susceptibility gene mutations, experts have also included ocular melanoma, exocrine pancreas adenocarcinoma, RCC carcinoma, mesothelioma, and CNS tumors in the spectrum of cancers justifying testing. In areas with higher melanoma incidence (eg, the United States), genetic testing should be proposed to individuals with at least 3 personal or familial CMS and/or pancreatic cancers (rule of 3). In areas with very high melanoma incidence (eg, Australia), the rule of 4 should be applied.

Data from our everyday clinical experience suggested that the percentage of detected mutations in unselected patients was lower than 10% when applying the rule of 2. This discrepancy might be due to the stringent selection process of the patients included in initial studies performed to establish the frequency of melanoma-prone mutations in the French population.

The aim of our study was to evaluate the efficiency of the detection of melanoma-prone mutations in a large series of 1032 patients recruited according to the rule of 2 at 2 melanoma referral centers in France between 2004 and 2015.

We calculated the mutation rate for high- and intermediate-susceptibility genes (CDKN2A, CDK4, BAP1, and MITF) in all index patients in whom we performed genetic testing applying the rules of 2, 3, and 4. Accordingly, our secondary aim was to describe conditions in which the testing rule could be lowered by 1 or 2 digits (ie, using the rule of 2 rather than the rule of 3) in the presence of various conditions: presence of multiple melanomas in 1 single patient or family member, presence of a lentigo maligna melanoma (LMM) or in situ melanoma, impact of age at first melanoma in index patients, presence of a DNS, and the presence of other associated cancers. Our tertiary aim was to evaluate the impact of anamnestic uncertainty on the efficiency of genetic testing. We identified a subgroup of patients for whom the anamnesis of family and/or personal history was uncertain and compared it with the whole sample using more precise data.

On the basis of our findings for countries with a low to moderate incidence of CM, we recommend offering genetic testing if criteria are met for rule of 3 cases of CM, including all pathological subtypes or other genetically associated cancers (exocrine pancreas adenocarcinoma, ocular melanoma, RCC carcinoma, mesothelioma, and CNS tumors) in an index patient or in first- or second-degree relatives. However, the rule of 2 should be applied in cases of a first melanoma occurring before age 40 years.

Methods

Patients

All index cases of patients from Angers and Lyons University Hospitals who had genetic testing for melanoma susceptibility genes between April 2004 and September 2015 were retrospectively included. General data, including demographics, presence of a DNS (as defined by ≥50 nevi >5 mm in diameter and including ≥1 clinically atypical nevus), number of CMs, age at diagnosis of the first melanoma, and pathological features, were recorded. History of other melanoma-prone gene mutation cancers (exocrine pancreas adenocarcinoma, RCC carcinoma, CNS tumors, mesothelioma, or ocular melanoma) was recorded both in patient’s personal and first- or second-degree relative history. When there was a doubt about personal or family medical records, the case was not counted in the number of cases and the family was included in the uncertain anamnesis subgroup for further analysis; however, the patient and/or family was not excluded from the study. Genetic testing performed only in a research setting (research on other melanoma-associated cancers or conditions) was excluded. The study was approved by the ethical committee of Lyons (No. 0179; December 17, 2015). All patients gave their written informed consent and were not compensated for their participation in this study.

Mutation Testing

Mutation screening of CDKN2A exons 1α, 1β, 2, 3, CDK4 exon 2, MC1R, MITF, and BAP 1 was carried out at Gustave Roussy Cancer Institute. DNA sequencing by the Sanger method was used, and all mutations were confirmed by analysis of a second independent blood sample. For every family, the presence or absence of mutation in high-risk (CDKN2A, CDK4, and BAP1), and intermediate-risk (MITF) susceptibility genes was recorded. We excluded from our analysis the detection of MC1-R variants when not associated with another mutation. When 2 mutations were observed in 1 individual it was then considered as a unique statistical event. CDKN2A and BAP1 variants of unknown significance were considered as negative results (18 index cases were concerned).

Statistical Analysis

The statistical unit was the index patient or family, and all of them met the criteria of the rule of 2 (ie, ≥2 cases of CM or of genetically related cancer in the index patient or in his or her first- or second-degree relatives). This constituted the whole cohort. We then considered the subgroups in which the criteria for the rule of 3 or rule of 4 were fulfilled. The threshold of 10% was considered to validate a testing scenario. Secondary analysis included evaluation of the mutation rate in different subpopulations. In a first sample, stringently corresponding to French guidelines for genetic testing, we excluded families with only 2 melanoma cases if 1 was in situ melanoma, or if 1 of each was melanoma diagnosed after age 75 years. The statistical impact of in situ melanoma or LMM subtype was also assessed by exclusion of patients and/or families.

The impact of age at first melanoma diagnosis was evaluated with 3 different thresholds (at ages 40, 50, and 60 years). The effect of the presence of DNS on mutation rate was also evaluated by performing analyses in this subpopulation. Finally, we assessed mutation rate in 4 subgroups of index patients with personal and/or family history of pancreatic cancer, RCC carcinoma, CNS tumors, and mesothelioma, or ocular melanoma, and evaluated association of these cancers with the type of mutation. Anamnestic uncertainty in family or personal history was recorded for the entire sample. The subgroup in which this suboptimal information was found was compared with that of patients and families with complete medical records to evaluate the impact of this real-life common observation on the diagnostic efficiency.

Statistical analysis was conducted using the IBM IPSS statistical software (version 19). One-way statistical analysis was performed for each variable, and the variable distributions comparisons were made by χ² test, Fisher exact test for qualitative variables, and t test for quantitative variables. Statistical correlation between each type of mutations and clinical or pathological variables was performed by χ² tests. P < .05 was considered statistically significant.

Possible Biases

There are possible selection biases because, according to the Institut National du Cancer (INCa), the Gustave Roussy Cancer Institute performs about 70% of all melanoma-prone gene testing. Our series of 1032 cases represents 25% of the recruitment of this laboratory. We can estimate that our series represents 17.5% of the French population tested for melanoma-prone genes and could be considered as representative for the whole nation.

Classification biases are also possible; for all included patients, we possessed the pathological proof of the diagnosis, and in families in whom uncertainty remained after revision of patient’s medical record, the CM and/or other cancers concerned were not taken into account in the total number for the family. However, these families were not excluded from the study but rather enrolled in the “uncertain anamnesis” subgroup analysis.

Results

Demographics and Medical History

This population-based study concerned 1032 patients, tested for genetic susceptibility to melanoma between 2004 and 2015 in Lyons and Angers University Hospitals. Their mean (SD) age was 54.5 (14.2) years [range, 18-89 years], and 543 (52.6%) were men. General characteristics of this sample are presented in Table 1. The median age at diagnosis was 51 years; there were at least 2 cases of melanoma or melanoma-related spectrum of cancers in each index case or in the patient’s first- or second-degree relatives.

Table 1. General Characteristics of the Sample of 1032 Patients^a.

Characteristic	No. (%)
Sex, male	543 (52.6)
Dysplastic nevi syndrome	84 (8.1)
Age at first melanoma	51 (14-88)
Multiple melanomas	581 (56.2)
Personal history
Pancreatic cancer	6 (0.6)
Renal cancer	15 (1.5)
CNS tumor	3 (0.3)
Ocular melanoma	5 (0.5)
Familial history
Cutaneous melanoma	459 (44.5)
Pancreatic cancer	133 (12.9)
Renal cancer	28 (2.7)
CNS tumor	20 (1.9)
Mesothelioma	1 (0.09)
Ocular melanoma	8 (0.8)
Histologic subtype
SSM	768 (74.4)
LMM	48 (4.7)
ALM	30 (2.9)
NM	23 (2.2)
Mucous membranes	4 (0.4)
Spitzoid	7 (0.7)
In situ	27 (2.6)
Unclassified	117 (11.3)
Unknown	8 (0.8)
Localization
Head and neck	84 (8.1)
Limb	346 (33.5)
Palm and soles	22 (2.1)
Trunk	562 (54.5)
Mucous	6 (0.6)
Unknown	12 (1.2)
Breslow index, mm
Median	1.00
Mean	0.55
Range	0.01-34.00
Anamnestic uncertainty	88 (8.5)

Abbreviations: ALM, acral lentiginous melanoma; CNS, central nervous system; LMM, lentigo maligna melanoma; NM, nodular melanoma; SSM, superficial spreading melanoma.

^aData are given as number (percentage) except where noted.

Genetic testing was principally offered because of multiple melanomas in a single patient (504 cases [48.8%]) or family history of melanoma (422 cases [40.9%]).

Frequency of Mutations in Whole Sample and in Subpopulations

The mutation rate in high- and intermediate-susceptibility genes was 67 (6.5%) (Table 2). These 67 events corresponded to 69 mutations because 2 patients harbored both MITF and CDKN2A mutations. Among these 69 mutations, 43 concerned CDKN2A (62.3%); 1, CDK4 (1.45%); 24, MITF (34.8%); and 1, BAP-1 (1.45%). The rate increased to 38 (9.3%) for the rule of 3 (eg, 0.09%; 95% CI, 0.06%-0.12%; P = .68). A threshold of 10% was reached when applying the rule of 4, with a significant mutation rate of 27 of 150 (18.0%) (eg, 0.18; 95% CI, 0.12%-0.24%; P < .001). Applying the rule of 3 in our cohort would have eliminated 29 patients and families with CDKN2A mutations, in 14 families and MITF mutations in 15 families. Exclusion of patients older than 75 years and in situ melanomas did not significantly change the mutation rate: 60 of 928 (6.5%).

Table 2. Frequency of Mutation.

Cohort and Subgroups	Rule of 2	Rule of 3	Rule of 4
Whole cohort
CDKN2A, CDK4, MITF, and BAP1, No. (%)	67 (6.5)	38 (9.3)	27 (18.0)
Patients, No.	1032	408	150
Age <75 y and exclusion in situ
CDKN2A, CDK4, MITF, and BAP1, No. (%)	60 (6.5)	33 (8.9)	24 (17.7)
Patients, No.	928	367	135
Exclusion of other cancer
CDKN2A, CDK4, MITF, and BAP1, No. (%)	55 (6.6)	28 (9.5)	20 (18.2)
Patients, No.	831	293	110
Only cases of multiples primary melanomas
CDKN2A, CDK4, MITF, and BAP1, No. (%)	25 (5.8)	10 (8.4)	6 (14.6)
Patients, No.	432	118	41
Only familial cases of multiple melanomas
CDKN2A, CDK4, MITF, and BAP1, No. (%)	15 (5.0)	4 (5.3)	3 (13.6)
Patients, No.	296	75	22

Abbreviations: BAP1, breast cancer 1 (BRCA1) associated protein 1; CDKN2A, cyclin-dependent kinase inhibitor 2A; CDK4, cyclin-dependent kinase inhibitor 4; MITF, microphthalmia-associated transcription factor.

Exclusion of other melanoma-prone, gene mutation–associated cancers increased the efficiency of the gene testing, with a score close to 10% (28 of 293 [9.5%]) in the rule of 3 scenario (eg, 0.09%; 95% CI, 0.06%-0.13%; P = .50), and significant for the rule of 4 (20 of 110 [18.2%]; eg, 0.18% [95% CI, 0.11%-0.24%]; P = .002).

Subgroup Analysis

Exclusion of in situ melanomas (n = 27 [2.6%]) and LMM subtypes (n = 48 [4.6%]) did not increase the number of detected mutations with a threshold of 10% for the rule of 4 (eg, 0.18% [95% CI, 0.12%-0.25%]; P < .001 for LMM; eg, 0.18% [95% CI, 0.12%-0.24%]; P < .001 for in situ melanomas; eg, 0.19% [95% CI, 0.12%-0.25%]; P < .001 for both LMM and in situ melanomas) (Table 3).

Table 3. Frequency of Mutation After Exclusion of LMM and In Situ Melanomas.

Exclusion	Rule of 2	Rule of 3	Rule of 4
Exclusion of LMM and in situ melanomas
CDKN2A, CDK4, MITF, and BAP1, No. (%)	65 (6.8)	37 (9.8)	27 (19.0)
Patients, No.	957	378	142
Exclusion of LMM
CDKN2A, CDK4, MITF, and BAP1, No. (%)	66 (6.7)	38 (9.8)	27 (18.7)
Patients, No.	984	388	144
Exclusion of in situ melanomas
CDKN2A, CDK4, MITF, and BAP1, No. (%)	65 (6.5)	37 (9.3)	27 (18.2)
Patients, No.	1005	398	148

Abbreviations: BAP1, breast cancer 1 (BRCA1) associated protein 1; CDKN2A, cyclin-dependent kinase inhibitor 2A; CDK4, cyclin-dependent kinase inhibitor 4; LMM, lentigo maligna melanoma; MITF, microphthalmia-associated transcription factor.

One of the 27 patients (3.7%) whose first melanoma was in situ harbored a missense mutation in MITF. In 1 of the 48 patients (2.0%) with no family history of melanoma, with LMM subtype and whose second melanoma was an SSM, we found a CDKN2A mutation.

The statistical impact of the age of the index patient was studied in the 3 scenarios with thresholds at ages 40, 50, and 60 years (Table 4). The rule of 3 obtains a significant score higher than 10% at a cutoff age of 40 years with a positive rate of 22 of 121 (18.2%, eg, 0.18% [95% CI, 0.11%-0.25%]; P = .001). At this age, 32 of 263 (12.1%) presented at least 1 mutation for the rule of 2, but this result was not statistically significant (eg, 0,12% [95% CI, 0.08%-0.16%]; P = .12).

Table 4. Frequency of Mutation According to Age at First Melanoma.

Age, y	Rule of 2	Rule of 3	Rule of 4
≤40 at first melanoma
CDKN2A, CDK4, MITF, and BAP1, No. (%)	32 (12.1)	22 (18.2)	17 (33.3)
Patients, No.	263	121	51
≤50 at first melanoma
CDKN2A, CDK4, MITF, and BAP1, No. (%)	40 (8.1)	26 (12.5)	20 (25.6)
Patients, No.	490	208	78
≤60 at first melanoma
CDKN2A, CDK4, MITF, and BAP1, No. (%)	55 (7.6)	33 (10.9)	23 (22.3)
Patients, No.	723	302	103

Abbreviations: BAP1, breast cancer 1 (BRCA1) associated protein 1; CDK4, cyclin-dependent kinase inhibitor 4; CDKN2A, cyclin-dependent kinase inhibitor 2A; MITF, microphthalmia-associated transcription factor.

When the rule of 2 was applied by exception in case of occurrence of first melanoma before age 40 years instead of the rule of 3, only 7 of the 14 families with the CDKN2A mutation were left undiagnosed.

Eighty-four of 1032 (8.1%) presented with a DNS, 40 of the 84 (47.6%) had a familial history of CM, and 62 of the 84 (73.8%) had a history of multiple primary CM. A threshold of 10% of positive mutation rate (11 of 84 [13.1%]) was reached for the rule of 2 (eg, 0.13% [95% CI, 0.06%-0.20%]; P = .17). Interestingly enough, of the 44 who had multiple primary CM without a family history of CM or other relatives with cancer, 4 (9.1%) had a positive mutation with a mean age at first melanoma of 52 years.

The statistical impact of the presence of at least another associated cancer was evaluated. Variation of mutation rate in each subgroup according type of cancer was not statistically significant. No significant association was found between the type of mutation and specific type of cancer, probably because of the small number of patients.

Uncertain anamnesis was recorded in 88 families (8.5%) owing to doubt about CM family history in 29 (32.9%), about family history of another cancer in 13 (14.7%), about pathological features of 1 of the index patient’s CM in 19 (21.5%), or relative’s CM in 13 (14.7%), and about the pathological subtype of associated cancer in 4 (4.5%). Ten cases (11.3%) were considered to be uncertain for other reasons. If excluded, 21 cases (2.0%) would not be submitted to the test, but the 67 remaining cases met enough additional criteria. Among the 21 cases then putatively excluded, only 1 was positive with a MITF mutation.

Discussion

To our knowledge, we present herein the first general population testing study including all patients who received genetic counseling owing to concern about a family risk of melanoma according to the rule of 2, which is applied in countries with low to moderate melanoma incidence.

All 5 previous French studies (population minimum, 48 families; maximum, 483; mean, 203) were performed with a research goal; thus, genetic testing has been realized on hyperselected smaller cohorts with restrictive inclusion criteria. In these reports, the CDKN2A mutation rate varied from 8% to 44%. In 2 of these, a mutation rate lower than the threshold of 10% was found; 1 concerned 100 patients with only multiple primary melanomas (with a 9% positive mutation rate) and the other a sample of 89 individuals with either multiple family cases, multiple primary melanomas, patients diagnosed when younger than 25 years, or patients with another cancer and presumably non–photo-induced melanomas (an 8% positive mutation rate). The most recent study assessed the CDKN2A mutation rate in 483 families with at least 2 cases of melanomas in first-degree relatives and found a mutation rate in 17% of the sample.

A Swiss study found a CDKN2A mutation rate of 10% in their 41 pedigrees with multiple melanomas or family history of melanomas or pancreatic cancer. This rate varies from 6% (2 of 15 patients) in patients with multiple melanomas without family history, to 11% (2 of 18) in patient with a family history and at least 2 cases of melanomas. The most recent hospital-based study, performed in Italy, including only multiple primary melanomas cases (≥2), found a mutation rate of 19.9%, but a founder effect could have modified this rate. A recent international study showed a 2.2% prevalence of CDKN2A mutations in hospital and population-based series with at least 1 melanoma. Only 1.1% of patients with melanoma with neither a family history nor multiple primary lesions had a CDKN2A mutation, but the rate increased to 1.5% for those with multiple primary lesions and to 5.2% for those with 1 affected relative. In the same study, but looking at only European countries, 6.5% of the sample with at least 2 primary lesions and 29.3% with at least 3 primary lesions harbored CDKN2A mutations. The results of this Swiss population-based study are similar to ours.

Our study of 1032 French index cases confirms that current guidelines for genetic testing need to be updated. The mutation rate in the whole cohort does not reach the threshold of 10% established by the ASCO guidelines. Only 67 of 1032 tested patients (6.5%) had a mutation in a high-risk or intermediate-risk melanoma susceptibility gene, including 43 of 1032 patients with CDKN2A mutations (4.2%). When applying the rule of 3, the mutation rate increased to 38 of 408 (9.3%), then to 27 of 150 (18.0%) when following the rule of 4. On the basis of these data, obtained in general population testing conditions and not in the setting of early precise genetic research studies designed to discover melanoma-prone mutations, we recommend moving toward the use of the rule of 3 in France. However, when reducing the number of tests to be performed, from 1032 to 408 (a decrease of 60.5%), the number of undetected cases that results when strictly applying the rule of 3 scenario, in this case, 29 of 1032 (a decrease of 43.3%), including 14 CDKN2A mutations, the highest-risk melanoma susceptibility gene, is a cause for concern. This is why we also tested subgroups to reduce the number of undetected families.

Concerning in situ melanoma, the French expert group recommended exclusion because of the risk of pathological misclassification. However, a French study reported that incidence of in situ melanoma has increased over the past 3 decades and is associated with an 8.08-fold increased risk of secondary invasive melanoma (95% CI, 7.66-8.57), supporting inclusion of these cases in the indications for testing. Inclusion of LMM cases is debatable because LMM is thought to be more related to environment than to genetic background. Yet the study performed by the GenoMEL consortium has shown a mutation rate of 10 in 49 patients (20.4%) in the LMM subgroup of affected members of families with a history of melanoma (1.5% of all CDKN2A mutation). Our data, when excluding in situ melanomas and/or LMM, did not change the mutation rate. However, because our study was not intended to specifically address the question of including in situ melanomas or LMM in the criteria for genetic testing, we cannot give definitive conclusions on this point.

Previous studies have found a median age of 40 years (range 36-43 years) for first melanoma in families with a CDKN2A mutation. Soufir et al evaluated the mutation rate in a subgroup of families (n = 28) with at least 2 affected members, including 1 younger than 50 years. It found a mutation rate of 8 of 28 (28%) compared with 22 of 48 (46%) for the whole sample. Soura et al suggested that multiple primary melanomas could develop later in life as a result of cumulative sun damage, whereas CM before age 40 years could be more related to genetic influences. In our subgroup analysis, a first melanoma occurrence before age 40 years increased the mutation rate. Moreover, when applying the rule of 2 in families and patients in whom at least 1 case was diagnosed before age 40 years, we also reduced the number of missed families with a CDKN2A mutation from 14 to 7. This finding could lead to recommend pursuing the indication of the rule of 2 in these cases. Moreover, knowing the high penetrance of CDKN2A mutation, we can assume that these families could be detected later, after occurrence of a third event, which is likely to occur.

The presence of DNS in index case did not significantly change the mutation rate. Moreover, the presence of a case of a genetically associated melanoma cancer in a relative did not seem to be an additional criterion for modifying the rule of 3 or the rule of 4. The small effect size in our study lowers the value of this recommendation and deserves additional research.

Limitations

In general population genetic testing, anamnestic uncertainty may act as a classification bias because a patient’s and a family’s medical history is based on interrogation and pathological proof of existence of such cancers in patients’ families is not always obtained. We have tried to evaluate this point in our analysis of the anamnestic uncertainty subgroup, which represented 8.5% of our cohort. Exclusion of patients for whom the indication for testing was based only on the presence of the index patient’s melanoma and 1 uncertain family case (2%) would have led us to ignore just 1 case of mutation (of MITF). In the literature, families often overestimate the number of melanoma cases and may confuse melanomas with nonmelanoma skin cancer. An Australian study found a rate of uncertain anamnesis of 20% in a cohort of 482 families with a history of melanoma.

Conclusions

To our knowledge, this is the first French population-based, general population–based study that reflects clinical practice about genetic testing for melanoma. Absence of large founder effect and the size of studied sample lead us to believe that these results are representative. On the basis of our data, we suggest changing the recommendations for France and for countries with similar melanoma incidence and apply the rule of 3 instead of the rule of 2 melanomas or genetically related cancers in the same patient or in first- and second-degree relatives when proposing genetic testing, except when in cases in which the first melanoma in the patient or family occurred before the age of 40 years, where the rule of 2 can be maintained.