Clinical and Sociodemographic Characteristics Associated With Thick Melanomas: A Population-Based, Case-Case Study in France

Objective To identify clinical and sociodemographic factors associated with very thick melanoma (VTM) (Breslow thickness, ≥3 mm) in France.

Design Retrospective, population-based, case-case study using a survey of cancer registries and questionnaires to practitioners.

Setting Five regions covering 19.2% of the French territory and 8.2 million inhabitants.

Cases Cases included all incident melanomas with a Breslow thickness of 3 mm or greater (ie, VTM), diagnosed between January 1 and December 31, 2008, in residents of the study area (Alsace, Bourgogne, Champagne-Ardenne, Franche-Comté, and Lorraine, France), and a randomly selected sample of melanomas thinner than 3 mm.

Main Outcome Measures Circumstances of diagnosis, clinical and pathological characteristics of melanomas, and sociodemographic characteristics of patients (age, sex, residence, home and family life conditions, educational level, and smoking habits).

Results Among 898 melanomas, 149 (16.6%) were VTMs. Very thick melanomas were more often diagnosed in a general-practice setting than thinner melanomas. The rate of immediate clinical recognition by dermatologists was lower for VTMs than for thinner melanomas. In a multivariate logistic regression analysis, factors associated with VTM were the nodular and acrolentiginous types; the head and neck and lower limb locations; older age; male sex; and being single, separated, divorced, or widowed. When only factors related to patients were taken into account, older age, male sex, and living alone were independent risk factors for VTM. The most significant risk was observed for patients living alone.

Conclusions Intrinsic factors related to the tumor and sociodemographic characteristics of patients contribute to the occurrence of VTM. These factors should be better targeted in future secondary prevention programs.

During the last 3 decades, the global prognosis of melanoma, a cancer potentially recognizable on simple skin examination by patients and/or physicians, has highly improved in most developed countries, mainly as a consequence of earlier diagnosis.^1-3 However, very thick melanomas (VTMs) at diagnosis are still frequently identified in many countries and play a major role in the global mortality related to this cancer.⁴ Although risk factors of melanoma as a whole have been identified by many case-control studies,^5-10 factors associated with advanced tumors at diagnosis have not been sufficiently studied to date. Among the studies that addressed this question,^11-23 only a minority analyzed the role of various sociodemographic factors,^{11,14,16,18,23} and very few were population-based studies^18,19,21 (the study by Van Durme et al¹⁸ accomplished both). Beyond the well-established risk factors of melanoma, a better knowledge of risk factors for advanced-stage melanoma should help to better target people with a higher risk of dying from this cancer.²⁴ This prompted us to perform a population-based, case-case study of clinical and sociodemographic characteristics associated with VTM in France, which represents, to our knowledge, the first such study in Europe.

The study was performed in the northeast part of France, grouping together the 5 regions of Alsace, Bourgogne, Champagne-Ardenne, Franche-Comté, and Lorraine, which cover 19.2% of the French territory and include 8.2 million inhabitants, ie, 13.5% of the French population. This area has the highest density of cancer and/or melanoma registries in France, with 4 registries located in the regions of Alsace (Registre des Cancers du Haut-Rhin, Registre des Cancers du Bas-Rhin), Franche-Comté (Registre des Tumeurs du Doubs), and Champagne-Ardenne (Observatoire du Melanome en Champagne-Ardenne).^25,26

This study was approved by the institutional review board of Reims University Hospital, Reims, France. Incident cases of clinical stage I-II invasive melanoma (according to the American Joint Committee on Cancer classification used in 2008²⁷), diagnosed between January 1 and December 31, 2008, in residents of the study area, were identified through a systematic survey of private and hospital dermatologists, completed by data from the 4 cancer registries and/or from private and public pathology laboratories, as previously described.^28,29 Initials and date of birth were used to exclude doubles. Quiz Ref IDVery thick melanomas were defined a priori as melanomas with a Breslow thickness of 3 mm or greater. The 3-mm cut point was used in preference to 2 mm to obtain a group with a larger proportion of potentially life-threatening melanomas and to maximize contrast with thinner tumors and in preference to 4 mm to provide a sufficient number of cases for statistical analysis. All incident melanomas with a Breslow thickness of 3 mm or greater were eligible as VTM cases. A similar number of thinner melanomas were randomly selected among cases with a Breslow thickness less than 3 mm, without any matching criteria.

For each melanoma, questionnaires were mailed to referring physicians, the majority of whom were private and hospital dermatologists, while the rest were general practitioners (GPs), surgeons, or other specialists. Several physicians (most often dermatologists and GPs) were contacted to obtain exhaustive information, when necessary. Physicians were asked either to retrospectively complete the questionnaire using data available in medical records or, as often as possible, to question patients and/or their families during a follow-up visit or by telephone. Physicians could require the assistance of a clinical research assistant.

The following data were collected for each case:

1. Basic sociodemographic data (ie, patients' age, sex, geographic region of residence, and administrative district of residence [namely, French département ]);

2. Clinical and histological characteristics of the tumor (ie, anatomical location; clinical appearance [pigmented or not]; Breslow thickness; ulceration; and histological subtypes comprising superficial spreading melanoma [SSM], nodular melanoma [NM], lentigo maligna melanoma [LMM], acral lentiginous melanoma [ALM], and other or unclassified subtypes);

3. Circumstances of diagnosis of melanoma (ie, whether the lesion was first identified in a general-practice setting or by directly consulting a dermatologist, another medical specialist, or a surgeon; and whether the lesion was recognized as a probable melanoma at first examination by a GP or a dermatologist, when applicable);

4. Detailed sociodemographic data (ie, patients' area of residence [urban vs rural, as defined by the French National Institute for Statistics and Economic Studies]; marital status [married or cohabiting, single, widowed, divorced, or separated]; number of children; whether the patient was living alone or with relatives at the time of diagnosis of melanoma; the total number of persons living with the patient; educational level [defined as low for patients with no high school diploma or equivalent and high in other cases]; whether the patient had worked outdoor during his or her professional life and the time spent working outdoors; and smoking habits [current smoker, former smoker, or nonsmoker]).

Quantitative variables were described as mean (standard deviation) or median (first and third quartiles) according to their distribution, and qualitative data as number and percentage. Comparisons between VTMs and thinner melanomas were performed using univariate analyses (t test, Wilcoxon test, χ² test, or Fisher exact test, as appropriate) and multivariate analyses (stepwise logistic regressions, with enter and removal limits set at 0.20 and factors significant at P = .20 included). P < .05 was considered significant. All statistical tests were performed using SAS statistical software (SAS Institute Inc).

Among 1075 incident melanomas identified through all data sources, questionnaires were obtained in 898 cases (83.5%), either directly from dermatologists (n = 603) or secondarily from dermatologists, surgeons, general practitioners, or other physicians after identification by cancer registries and pathology laboratories (n = 295). Breslow thickness was known in 887 of these 898 cases (98.8%), ranging from 0.03 to 20.0 mm (mean, 1.67 mm; median, 0.75 mm). A total of 149 melanomas (16.6%) had a Breslow thickness of 3 mm or greater (range, 3.0-20.0 mm; mean, 5.85 mm; median, 4.5 mm), making up the group of VTMs. Among the 743 cases thinner than 3 mm, 149 were randomly selected for analysis. Basic sociodemographic data and clinical and histological characteristics of the tumor were obtained for all cases. The mean Breslow thickness in the group of thinner melanomas was 0.88 mm (median, 0.68 mm; range, 0.15-2.80 mm). Complementary informative data regarding the circumstances of diagnosis and detailed sociodemographic data were obtained for 139 VTMs (93.3%) and 137 thinner cases (92.0%) by extraction from medical records in all cases and completed by direct patient interviews in 163 cases (59.7%).

A comparison between VTMs and thinner melanomas for the circumstances of initial diagnosis is given in Table 1. Quiz Ref IDVery thick melanomas were more often identified in a general-practice setting than thinner melanomas (62.6% vs 47.9%; P = .01). Although the rate of correct diagnosis at first examination was much higher for dermatologists than for GPs in both groups (77.1% vs 47.7% for VTMs and 88.8% vs 43.2% for thinner melanomas; P < .001), dermatologists showed a lower clinical diagnostic performance for VTMs than for thinner tumors (77.1% vs 88.8%; P = .01).

Table 2 gives a univariate analysis comparing both groups for variables related to tumors and patients. Both types of variables exhibited significant differences between VTMs and thinner tumors. Quiz Ref IDCompared with thinner tumors, VTMs were more frequently located on the head and neck or the lower limb (60.4% vs 35.5%; P < .001). Conversely, VTMs were rarely located on the trunk (24.8% vs 43.0%; P = .001). They were more frequently clinically achromic (28.8% vs 12.3%; P = .001). They had different histological subtypes, with a high proportion of NM and ALM (54.4% vs 9.4%; P < .001). Quiz Ref IDCompared with patients with thinner melanomas, those with VTMs were older (age 64 years vs 55 years; P < .001) and were more likely to be male (58.4% vs 47.7%; P = .06), single, separated, divorced, or widowed (35.8% vs 23.5%; P = .03), living alone (31.9% vs 19.0%; P = .01) and less educated (73.4% vs 60.3%; P = .02). No difference was observed for the number of children, residence in a rural area, smoking habits, or working outdoors. In a multivariate analysis of characteristics of both patients and tumors, location on the head and neck or the lower limb, nodular or acrolentiginous histological subtypes, older age, male sex, and being single, separated, divorced, or widowed were identified as significant independent risk factors for VTM (Table 3). In a second multivariate model with only factors related to patients being taken into account, older age, the male sex, and living alone were associated with VTM (Table 4).

Breslow thickness is a powerful predictor of patient survival.⁴ It should therefore be a crucial end point in secondary prevention campaigns, rather than end points not directly related to mortality, such as number of people consulting a dermatologist or even the number of melanomas diagnosed. Unfortunately, most secondary prevention campaigns using common messages about risk factors for melanoma and clinical signs for early detection (in particular the ABCD rule [asymmetry, border irregularity, color variegation, and diameter >6 mm]) have failed to result in a reduction of incidence of thick melanomas.^4,30,31 Several reasons may explain these failures. First, risk factors for thick melanoma may differ from risk factors for melanoma as a whole. For instance, a high educational level has long been identified as a risk factor for melanoma,^6,32 whereas more recent studies (such as the present study) found that a low educational level was associated with thicker melanoma.^{11,14,16,23,24,33} At a geographic level, high socioeconomic status (including lower poverty, higher education, higher income, and lower unemployment) was associated with higher melanoma incidence rates in the United States, but this association did not persist for late-stage melanoma.³⁴ Second, people who are more prone to have melanoma diagnosed at an advanced stage may slip through the nets of common screening campaigns. Lastly, the clinical signs emphasized in most campaigns (mainly the ABCD rule) may be inadequate for an earlier detection of many thick melanomas.^35,36 Therefore, a better understanding of clinical, histological, and sociodemographic factors associated with thick melanoma is a crucial point for improving future campaigns aiming to reduce melanoma mortality.

Despite limitations due to its retrospective design and its relatively small patient sample, our study has several strong points. First, it is a population-based study in a large geographic French area including 8.2 million inhabitants, representing, to our knowledge, the first European population-based study to date addressing this question. Second, the collected data relied not only on cancer registries and pathology laboratories, but also on personal questionnaires to physicians, 60% of which were completed in the presence of patients, allowing more detailed and reliable sociodemographic information to be recorded. Lastly, in contrast to previous studies that focused on melanoma thicker than 1 mm,²² 1.5 mm,^11,16 or 2 mm,¹⁹ we analyzed factors associated with VTM (≥3 mm). Most of the VTMs were also ulcerated and therefore represented a larger proportion of life-threatening melanomas.

We identified intrinsic factors related to the tumor and sociodemographic factors related to patients as significant risk factors for VTM. The nodular or acrolentiginous histological subtypes were associated with VTM in multivariate analysis, independent of anatomical location and factors related to patients. In France, NM and ALM account for approximately 15% and 4% of all melanoma cases, respectively.^25,29 Both NM and ALM have been previously reported to be associated with a high Breslow thickness and a poor prognosis.^{16,24,25,28,37-39} Factors that may contribute to these subtypes being diagnosed at a more advanced stage are a more biologically aggressive behavior and a delayed diagnosis. Nodular melanomas lack an initial radial growth phase and have been well demonstrated to have an increased number of mitoses and a more aggressive growth rate.⁴⁰ Both NM and ALM frequently have clinical characteristics not typical of melanoma, which may lead to misdiagnosis at first consultation. In our study, 33% of NM and ALM cases were clinically amelanotic, compared with 15% among other histological subtypes (P < .001; data not shown). Nodular melanomas are frequently symmetric, invariant in color, and smaller than 6 mm in diameter and therefore lack the ABCD criteria.^19,35,41 Acral lentiginous melanomas frequently have atypical clinical features leading to the incorrect initial hypotheses of wart, hematoma, or traumatic lesion. The early recognition of NM and ALM is challenging, not only for GPs but also for dermatologists. Quiz Ref IDIn the present study, the rate of correct diagnosis at first examination by dermatologists was paradoxically lower for VTM (54.4% of which were NM or ALM) than in the group of thinner melanomas (77.1% vs 88.8%; P = .01). In view of the high proportion of NM and ALM among VTM cases, traditional messages mainly relying on the ABCD rule in secondary prevention campaigns have to be reconsidered, as previously mentioned by some authors.^36,42,43 Patients' and physicians' attention should be specifically drawn to any new or changing lesion, including amelanotic ones, and to any acral, atypical lesion without a clear diagnosis. Biopsy after a short period of observation should aid in the early differentiation of NM and ALM from inflammatory or traumatic lesions.

Melanomas located on the head and neck or the lower limb were more likely to be VTM than melanomas at other locations, including the trunk. These locations were a significant risk factor of VTM in multivariate analysis. This result is paradoxical, since both the head and neck region and the lower limbs are frequently uncovered anatomical areas, easily visible by patients themselves and their close relations. In contrast, the posterior aspect of the trunk, which is not easily accessible to self-examination, was not associated with VTM. Thick melanomas have been previously reported to be more frequent on the head and neck, as compared with the trunk, in hospital-based studies in various countries.^16,21,44 Indeed, head and neck melanomas are heterogeneous, including a high proportion of both typical, slow-growing LMMs and rapidly growing, often nodular and/or achromic melanomas. Obstacles to early recognition and/or excision of both types of tumors, particularly in elderly people, could contribute to the high frequency of VTM at this location. A high frequency of thick melanomas on the lower limbs has not been frequently reported to our knowledge. However, most ALMs are located on the foot,³⁷ and an Australian population-based study found that melanoma on the lower limb were also more likely to be NM.²¹ As a high proportion of melanomas were diagnosed in a general-practice setting in the present study, difficulties of GPs to regularly and systematically examine the lower limbs, as opposed to the trunk, could also play a role in a delayed diagnosis of melanoma at this location, particularly in elderly people.

Our study underlines the role of factors related to patients in the occurrence of VTM. Patients with VTM were on average nearly 10 years older than those with thinner melanomas. The risk of having a VTM was increased by a factor of 2.2 in men compared with women after adjusting for other tumor and patient-related variables. Numerous previous hospital-based^{13,14,16,21,22,44} or population-based^{17-20,25,28,38} studies in various countries reported the occurrence of thicker and more advanced melanomas in men^{13,14,17-22,25,44} and/or elderly patients^{13,14,16,19,21,25,28,44} than in women and younger patients. Melanomas in men and/or elderly patients frequently have tumor characteristics associated with an increased Breslow thickness, such as the location on the head and neck and nodular histological subtype. However, our study confirms that age and sex play a role in tumor thickness independent of these tumor-related factors. Although melanoma is not uncommon in young adults and has a higher incidence in women than in men in France, as in most developed countries, secondary campaigns aiming to reduce the incidence of thick tumors and melanoma-related mortality should more specifically target men and older people.

In the present study, multivariate analyses showed a significant association between VTM and the status of being single, widowed, or separated or living alone. Living alone was the most significant risk factor for VTM in a model where only patient-related factors were taken into account. To our knowledge, an association between a greater tumor thickness and patients living alone has only been reported twice before, in a German, hospital-based study¹⁶ comparing melanomas according to a 1.5 mm thickness cut point, and in an Italian, hospital-based study²³ in which Breslow thickness was categorized as 0 to 1 mm, 1 to 3 mm, and 3 mm or greater. In the study by Mandalà et al,²³ living alone was also demonstrated to be associated with a higher risk of melanoma-related death, independent of age and sex, in patients with a low socioeconomic status. In a study based on the tumor registry of Florida that only evaluated sociodemographic factors, Van Durme et al¹⁸ found that patients with melanoma who were unmarried were at higher risk of having regional lymph node or distant metastases at diagnosis. Our study confirms and extends these results. Because many people who live alone or are single, separated, or widowed are elderly subjects (including many older men), it is noteworthy that these family variables were associated with VTM independent of age and sex in our study, suggesting that not only elderly people but also younger, isolated subjects should be targeted in public health campaigns aiming to reduce the incidence of thick melanoma.

In conclusion, the recognition of factors related to advanced-stage disease is the cornerstone of efforts to improve melanoma outcome. Future prevention programs should particularly focus on atypical/non–ABCD criteria tumors, including ALM, NM, and achromic melanoma (particularly on the head and neck or the lower limb), and target men, elderly people, and patients living alone, regardless of their age.

Correspondence: Florent Grange, MD, PhD, Service de Dermatologie, Hôpital Robert Debré, Avenue du Général Koenig, 51092 Reims CEDEX, France (fgrange@chu-reims.fr).

Accepted for Publication: May 14, 2012.

Published Online: September 17, 2012. doi:10.1001/archdermatol.2012.2937

Author Contribution: Dr Grange had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Grange, Lipsker, Granel-Brocard, Velten, Truchetet, Buemi, Bernard, and Woronoff. Acquisition of data: Grange, Barbe, Aubin, Lipsker, Granel-Brocard, Velten, Dalac, Truchetet, Michel, Mitschler, Arnoult, Buemi, Dalle, Bernard, and Woronoff. Analysis and interpretation of data: Grange, Barbe, and Arnoult. Drafting of the manuscript: Grange, Barbe, and Mitschler. Critical revision of the manuscript for important intellectual content: Aubin, Lipsker, Granel-Brocard, Velten, Dalac, Truchetet, Michel, Arnoult, Buemi, Dalle, Bernard, and Woronoff. Statistical analysis: Grange and Barbe. Obtained funding: Grange. Administrative, technical, and material support: Aubin and Arnoult. Study supervision: Grange, Aubin, Dalac, and Truchetet.

Funding/Support: This study was supported by a grant from the Ligue Contre le Cancer.

Additional Contributions: We thank the physicians and dermatologists who provided data for this study and the following associations of dermatologists: Association d’Enseignement Post-Universitaire en Dermato-Vénérologie de Champagne-Ardenne, Association des Dermatologues de Bourgogne, Association d’Information Post-Universitaire des Dermato-Vénérologues de Strasbourg, Association de Formation des Dermatologues de Franche-Comté, Association Lorraine Post-Universitaire de Dermatologie.