From being a relatively infrequent tumour in the early 20th century, cutaneous melanoma (CM) has evolved to become a common malignancy with a significant burden to society and public health. Its incidence has been steadily increasing since the mid-1960s in most fair-skinned populations and is predicted to continue increasing for at least two more decades.[1, 2] Melanoma is the fifth leading cancer in males and the seventh in females in the U.S.A. It accounts for the majority of deaths from skin cancer, even though it represents < 5% of all cutaneous malignancies. Despite the dismal prognosis of patients with metastatic disease, CM is associated with higher 5-year survival rates compared with other malignancies, outranked only by prostate, thyroid and testicular cancer. While increasing in the 1970s and 1980s,[5, 6] mortality rates of melanoma have stabilized since the early 1990s in Australia, the U.S.A. and certain European countries, possibly reflecting the effects of early recognition and the diagnosis of thin melanomas with a more favourable prognosis.[7-9] On the therapeutic front, two novel agents, ipilimumab and vemurafenib, have recently been approved for the treatment of metastatic melanoma opening the era of targeted immune- and mutation-based therapy in advanced disease. Both these agents have provided promising results with meaningful effects on progression-free and overall survival. Nevertheless, the related cost is high and the overall prognosis, although better, remains poor. Early diagnosis and prevention remain the cornerstones for optimal melanoma containment. This review article discusses the recent trends in melanoma incidence and mortality, and outlines the current knowledge on established and newly investigated risk factors and disease associations.
Cutaneous melanoma (CM) is one of the most rapidly growing cancers worldwide, with a consistent increase in incidence among white populations over the past four decades. Despite the early detection of primarily thin melanomas and the improved survival rates observed in several countries, the rate of thick melanomas has remained constant or continues to increase, especially in the older age group. Current considerations in the epidemiology of melanoma focus on the observed survival benefit of females vs. males, the contributing role of indoor tanning in melanoma risk and the diverse effect of sun exposure in the development of different types of melanoma with respect to their clinical and mutational profile. Certain well-known risk factors, such as skin, hair and eye pigmentation and melanocytic naevi have been validated in large-scale association studies, while additional lifestyle factors and iatrogenic exposures, such as immunosuppressive agents and nonsteroidal anti-inflammatory drugs are being investigated. In addition, genome-wide association studies have revealed genetic loci that underlie the genetic susceptibility of melanoma, some of which are related to known risk factors. Recently, an interesting association of melanoma with Parkinson disease has been noted, with a higher than expected frequency of melanoma in patients with Parkinson disease and vice versa. This review article provides an update in the epidemiology of cutaneous melanoma and discusses recent developments in the field.
Melanoma incidence and mortality
Recent trends in melanoma incidence
The majority of epidemiological studies show an increase in melanoma incidence among white populations over the past few decades, presumed to be related to changing attitudes of recreational behaviour and sun exposure. Annually, this incidence increase varies among populations, ranging from 3% to 7%, equal to a doubling of rates every 10–20 years. The highest incidence rates worldwide have been reported in Australia and New Zealand with up to 60 cases per 100 000 inhabitants per year. In the U.S.A. the rates among white individuals were 25·4 per 100 000 men and 16·9 per 100 000 women per year from 2003 to 2007.[13, 14] From the 1990s onwards, studies from Australia, New Zealand, the U.S.A., and several Western European and Nordic countries indicate a stabilization or slowing down of incidence rates in both sexes, mainly in the younger age group. Dramatic increases, however, are consistently and uniformly observed in the age group > 60 years, particularly in male subjects. Current reports from the U.S.A. show an estimated average annual percentage change of 6·15% [95% confidence interval (CI) 4·31–8·02%] for the time period between 1992 and 2005 in this age group (> 60 years).
In Europe, increases in melanoma incidence have been documented over the past few decades with wide north/south and east/west variations. The estimated age-standardized rates of new melanoma cases for 2012 were 11·4 per 100 000 for males and 11·0 per 100 000 for females, ranging from approximately six new cases for Central and Eastern Europe, to 10 cases for Southern Europe and up to 19 cases per 100 000 for Northern Europe. Melanoma incidence has been more profoundly raised in Northern Europe, with examples of 10-fold increases in Scandinavian countries between 1953 and 1997. A recent analysis of the GLOBOCAN 2008 dataset confirmed that the incidence rate of Central and Eastern Europe is less than half that of Western Europe. Variations among populations in Europe could be explained by differences in skin phenotype as well as in sun-exposure behaviours. Paradoxically, the rate of incidence differences appears to be greater than anticipated among neighbouring countries, e.g. Poland/Germany or Hungary/Romania, despite their expected share of common risk factors, i.e. latitude, ultraviolet (UV) index distribution, pigmentation characteristics. These differences, although not well understood, could be explained in part by differences in case reporting and registration. Only 20 out of 41 European countries have good-quality cancer registries, leading to the possibility of melanoma being under-reported in certain countries. The male/female ratio of melanoma varies among different countries. A male predominance has been recorded in countries with a high melanoma incidence, such as Australia and the U.S.A.[20, 21] In Europe, there is a discrepancy concerning sex predominance. The majority of Western and Northern European countries report higher incidence rates in females vs. males, whereas in most Central, Eastern and Southern European countries melanoma predominates in men. Differing patterns of melanoma incidence have been also noted in relation to age. Despite the reported stabilization of incidence in younger birth cohorts in several countries, recent data from the U.S.A. have shown an alarming increase of melanoma in young female subjects over the past three decades. An analysis of the Surveillance Epidemiology and End Results (SEER) data in the 20–49 age group showed an increase in age-adjusted rate from 8·1 in 1975 to 17·4 in 2008 for women, whereas in men the same rate increased from 8·3 to 12·5, respectively.[22, 23] A more recent analysis of 17 registries of the SEER programme revealed that for the age groups < 44 years, female subjects showed higher incidence rates, with the peak difference in the age group of 20–24 years. Men exhibited higher incidence rates after the age of 44 years. This difference was not observed for nonmelanoma skin cancers (NMSC), which are known to be strongly associated with cumulative exposure to UV radiation (UVR), suggesting that additional gender-related factors, such as endogenous hormones, may play a role in early-onset melanoma.
The anatomical location of melanoma varies according to age and shifts from truncal and extremity locations in younger ages to the head and neck location in advanced ages. The site development of melanoma has been associated with distinct amounts and patterns of sun exposure, based on the divergent model of melanoma development and the propensity for intrinsic melanocytic proliferation. Recent pooled analysis of case–control studies has provided support for this hypothesis showing that melanomas of the head and neck are associated with chronic sun exposure and are closely related to other indices of chronic actinic damage, such as solar keratoses, while truncal melanomas are associated with intermittent patterns of sun exposure and are more commonly seen in the context of multiple or atypical naevi.[26-28]
Levelling of mortality rates but persistence of thick tumours
Recent evidence suggests that the increases in incidence rates are not followed by a similar trend in mortality rates, which have remained relatively stable since the 1980s. The divergence between incidence and mortality likely reflects an improved survival due to the earlier detection of thin melanomas at a potentially curable stage. A trend towards thinner, less-invasive melanomas or tumours of earlier stage has been reported in Central Europe, U.S.A. and Australia during recent decades.[9, 29] Expanded skin screening and detection of biologically less-aggressive tumours with a low or clinically insignificant metastatic risk have been cited as possible causes of the stable mortality in view of the increasing incidence of the disease. Studies focusing on incidence patterns per tumour thickness have argued against the notion of a melanoma epidemic being simply due to the overdiagnosis of melanoma. In a U.S. population-based study, including approximately 300 million person-years of observation by the SEER programme over a 12-year period (1992–2004) and more than 70 000 new cases of malignant melanoma, significant increases for tumours of all histological subtypes and thicknesses were shown, including thick melanoma > 4 mm. Most importantly, this increase was independent of socioeconomic status, which was assessed as a surrogate marker of access to screening. Reports from other geographical areas have also reported an increased incidence of thick tumours, particularly in older men, with a parallel dramatic trend in mortality.[31-33] Mortality rates in the U.S.A. were noted to decline in the 1990s by 39% in women and 29% in men in the age group 20–44 years, but were increased by 70% and 157% in men aged 45–64 and ≥ 65 years, respectively. It has been postulated that melanoma in the elderly may have a different biological behaviour because of different biology or altered host defence mechanisms contributing to the observed higher incidence and mortality in this age group. Compared with other sociodemographic groups, older men tend to have thicker, more late-stage melanomas, usually on the head and neck area and more often of the nodular type. They are also less likely to be involved in self-skin examination, mass screening or physician-based examination.
Do women with melanoma have a better prognosis?
Gender differences in melanoma prognosis have been reported in several studies reporting a better survival rate in female subjects compared with men.[36, 37] This survival difference has been ascribed to an earlier diagnosis, i.e. thinner tumours occurring more commonly in females, and to differences in anatomical predilection, i.e. women present more frequently with tumours on the extremities that have a more favourable outcome. Recent evidence, however, shows that these factors can only partially explain this survival benefit. A population-based study of 10 538 patients with melanoma from the Netherlands investigated sex differences in melanoma survival after adjustment for several tumour-related variables. The study showed that male patients carried an almost double relative excess risk (RER) of dying from melanoma compared with females (RER of dying 1·87, 95% CI 1·65–2·10), after adjusting for age, period of diagnosis, anatomical location, Breslow thickness, nodal involvement and metastatic status. Interestingly, the strength of the risk estimate of the gender difference was comparable to other well-known prognostic risk factors, i.e. Breslow thickness, ulceration and histological type. Whether this survival gain applies for all melanoma stages is not clear, although a subgroup analysis suggested that the female superiority in survival persisted even in patients with advanced-stage disease. A recent study by Joosse et al. suggested that gender influenced the course of melanoma during all progression phases, with women showing a lower risk of lymph node invasion and visceral metastases. In addition, females with localized disease had a better survival than men, even after initial disease progression.
The causes of the survival advantage in women are not clear. Recent studies have shown that reproductive, hormonal or menstrual factors do not affect melanoma risk. Moreover, tamoxifen, an oestrogen-receptor inhibitor, does not improve survival in patients with metastatic melanoma and to date, no definite evidence exists to support the use of hormonal therapy in advanced stages of the disease. No prognostic differences have been observed between pregnant and nonpregnant women.[41, 42] In addition, a recent study showed a survival benefit in both pre- and postmenopausal women, supporting the fact that oestrogen does not influence survival. Further research is needed to investigate the molecular and immunological basis of gender differences in melanoma prognosis and survival.
Risk factors of melanoma
Cutaneous melanoma arises from a complex interaction between environmental and constitutional or phenotypic factors. This interaction not only influences the incidence of melanoma but also the clinical characteristics and oncogenic pathways through which the tumour develops.
Environmental risk factors
Natural ultraviolet radiation exposure
Lifetime exposure to UVR is a well-established risk factor for melanoma development. UVR has damaging effects on the skin via direct and indirect mechanisms, such as the formation of cyclobutane pyrimidine dimers, gene mutations, immunosuppression and oxidative stress.
The role of UVR exposure as a leading environmental cause of melanoma is further supported by a wealth of descriptive evidence, including a high prevalence of melanoma in populations that migrated from a low to a high ambient UVR environment,[44, 45] a higher incidence in fair-skinned compared with darker-skinned individuals and a latitude-dependent rise in melanoma rates among white populations with proximity to the equator.[46, 47] However, differences in rates between indoor and outdoor workers and variations in the anatomical distribution of the tumour suggest a complex association of melanoma with UVR that does not conform to a straightforward dose-relationship model. A history of intermittent exposure to excess UVR doses and of painful sunburns, as a marker of host sensitivity, were a consistent finding in the majority of case–control studies and were confirmed in recent systematic reviews.[48, 49] A pooled analysis of sun-exposure patterns and melanoma risk in 5700 cases and 7216 controls at different latitudes showed that different amounts and pattern exposures of UVR exert a differential effect on site-specific melanoma risk. Intermittent sun exposure and sunburns in childhood were strong predictors of melanoma on the trunk and limbs, i.e. less frequently exposed sites, with little variation across latitudes. Occupational exposure was more likely associated with melanoma on the head and neck, i.e. on continuously exposed sites, an effect that was, however, statistically significant only in low-latitude populations. These findings are not necessarily in conflict with results of previous meta-analytical studies showing an inverse relationship between melanoma risk and occupational exposure, but may, in fact indicate that this association is less evident in temperate areas and more pronounced in areas of higher ambient UVR levels.
Molecular studies have given further credibility to the notion that melanoma is a heterogeneous disease comprising distinct biological subtypes possibly arising through different causal pathways. For example, BRAF mutations, which are the principal causal mutation in melanoma (detected in roughly 50% of melanomas), are detected in lesions typically arising in anatomical locations of intermittent UVR exposure and in patients with high early-life ambient UV exposure.[51-53] In contrast, c-Kit mutations and cyclin D1 gene amplifications are commonly detected in acral and mucosal melanomas and in melanomas related to chronic sun exposure, in which a relative paucity of BRAF mutations is noted.[54, 55] These data suggest important differences between melanomas on a molecular level based on their site, type of sun exposure and constitutional and genetic factors that are not yet entirely clear.
Artificial ultraviolet radiation exposure: tanning beds
Over the past 20 years, indoor tanning has become popular among white populations, and particularly in young women. To date, there is strong evidence that sunbed users have an increased risk for melanoma development, even after adjustment for outdoor sun exposure. A dose dependency underlying this risk association has been also reported, expressed by the length, duration or number of treatment sessions. A recent meta-analysis by Boniol et al. showed an overall summary relative risk (RR) of 1·2 (95% CI 1·08–1·34) of melanoma development in ‘ever use’ of sunbeds and a 1·8% increase of risk for each additional session of sunbed use per year. In a subgroup analysis of subjects who first used sunbeds at an age below 35 years, the summary RR rose to 1·87 (95% CI 1·41–2·48) indicating a higher melanoma risk with an early onset of tanning bed exposure. A large case–control study from the Nurses’ Health Study reviewing cancer incidence data over a 20-year span (1989–2009) among 73 494 female nurses who used tanning beds prior to the age of 35 years showed a significantly increased risk of melanoma, basal cell carcinoma and squamous cell carcinoma (SCC), with multivariable-adjusted hazard ratios of 1·1, 1·5 and 1·5, respectively. It has been estimated that 5·4% of all new melanoma cases, diagnosed every year in 15 countries of the European Community and three more member countries of the European Free Trade Association, could be attributed to sunbed use. Interestingly, female subjects represented most of this burden, with 6·9% of all melanoma cases in women being related to the use of indoor tanning devices. This is further supported by the significant increase of trunk-located melanoma in females over the past 15 years, mainly in countries where indoor tanning is popular.[23, 59]
Phenotypic risk factors
The major constitutional risk factors for melanoma include fair pigmentation, poor tanning ability, multiple naevi, clinically atypical or dysplastic naevi and freckling. The majority of observational studies report a significant increase of RR for melanoma development in patients with multiple or clinically atypical naevi, while a recent pooled analysis using the original data from 15 case–control studies showed that the level of risk of an abnormal naevus phenotype, defined by the presence of multiple banal, clinically atypical, or large naevi, was consistent across studies and at different latitudes. In their meta-analysis, Gandini et al. showed a gradient of risk, proportional to the number of common or dysplastic naevi. The RR for patients with one dysplastic naevi was 1·45 (95% CI 1·31–1·60), 3·03 (2·23–4·06) for patients with three dysplastic naevi and 6·36 (3·80–10·33) for those with five dysplastic naevi. Patients with a high number of common naevi (> 100) carried a sevenfold increased risk for melanoma, compared with those with low numbers (0–15 common naevi) (RR 6·89; 4·63–10·25). A retrospective analysis of 10 case–control studies of melanoma in women found that the density of naevi correlates with the site of melanoma development with higher naevi counts on the arm strongly associated with melanoma on the trunk or limbs but not of the head and neck.
Apart from melanocytic naevi, pigmentary characteristics have been consistently associated with an increased likelihood of melanoma development. In the meta-analysis by Gandini et al., the RR for melanoma of ‘fair eye colour’, consisting of blue, green and hazel, was 1·62 (95% CI 1·44–1·81) compared with the ‘dark eye colour’. With respect to hair colour, the RR for melanoma of ‘light hair colour’, consisting of blond, red and light brown colour, was 1·87 (1·63–1·95) compared with the ‘medium dark, brown hair colour’, with red hair colour having the highest RR (3·64; 2·56–5·37). A recent meta-analysis by Olsen et al. investigated the contribution of eye and hair colour, skin phototype and the presence of freckling, on melanoma risk. The highest population-attributable fractions (PAF) were observed for skin phototypes I/II (0·27), presence of freckling (0·23), and blond hair colour (0·23). The PAF for blue/blue-grey eye colour was higher than for green/grey/hazel eye colour (0·18 vs. 0·13), while the PAF for red hair colour was 0·10.
Genetic risk factors
Heritable factors play an important role in melanoma predisposition. A family history of melanoma is associated with a significant twofold increased risk of melanoma and a PAF that varies between geographical regions (0·007 in Europe and 0·064 in Australia). A significant proportion of familial disease (30–40% of cases) is explained by inherited mutations of two high-penetrance genes with a critical role in cell-cycle control, namely CDKN2A and CDK4. The penetrance of mutations in these genes and the likelihood of developing melanoma in mutation carriers is variable, depending on co-inherited modifiers and environmental modulators, such as ambient light exposure. The use of the newest sequence technologies has uncovered additional high- and low-risk melanoma genes, many of which underlie genetically some of the phenotypic traits associated with melanoma. A mutation in BRCA-associated protein-1 (BAP1) was identified in families with multiple kindreds with cases of uveal and cutaneous melanomas. Among the low-penetrance genes, the most consistent association with melanoma has been that of the melanocortin 1 receptor gene (MC1R), a G-protein coupled receptor that binds to α-melanocyte stimulating hormone (α-MSH). Loss-of-function polymorphisms in MC1R cause a shift of melanogenesis from the photoprotective eumelanin to pheomelanin, resulting in a phenotypic spectrum of red hair colour, pale skin and freckles. A pooled analysis of nine MC1R variants showed a RR of 2·44 (95% CI 1·72–3·45) for CM in patients carrying the ‘red hair variants’ (polymorphisms D84E, R142H, R151C, R160W, D294H) compared with a RR of 1·1 (1·1–1·51) for those with the ‘non-red hair’ MC1R variants. Despite its role in pigmentation and red hair, two well-known risk factors for melanoma, there is evidence supporting the involvement of MC1R in nonpigmentary pathways through enhanced DNA damage and oxidative damage. A series of genome-wide association studies on pigmentary phenotypes and skin cancer risks have implicated further genetic risk loci affecting pigmentation (ASIP, TYR), naevi proliferation (PLA2G6, MTAP, IRF4), DNA repair pathways and apoptosis [TERT/CLPTM1L, TIPARP (formerly PARP-1), ATM, CASP8] and other loci of yet unknown significance.[70-73] A meta-analysis of genetic association studies in melanoma, including genome-wide association studies, showed that the loci connected with the most compelling risk-effect estimates are correlated pigmentary traits, including MC1R. Recently, a rare functional nonsynonymous variant (E318K) within the MITF gene that alters the gene's transcriptional activity was identified and associated with a significant odds ratio for melanoma ranging from 2·19 (1·41–3·45) to 4·78 (2·05–11·75).[75, 76] The E318K MITF variant was also associated with a higher naevus count, a nonblue eye colour and an increased incidence of renal cancer.
Additional risk factors
Additional well-established risk factors of melanoma include a family history of the disease, a personal history of NMSC and immunosuppression related to organ transplantation, lymphoproliferative disease or human immunodeficiency virus infection/AIDS. Lately, several novel risk associations have been reported without conclusive evidence. In a recent large study of early-onset melanomas, O'Rorke et al. reported a more-than-double increased risk for melanoma in infants with higher birthweight, suggesting that factors which lead to intrauterine fetal growth may impact on melanoma risk. Exposure to heavy metals and insecticides and recent experience of stressful events have also been reported, but their results should be interpreted with caution due to their observational design.[78-80] An inverse association between smoking and melanoma risk, especially of melanomas on the head and neck, was identified in a recent meta-analysis. The authors suggested that nicotine has an anti-inflammatory effect that protects melanocytes from the inflammatory reaction induced by long-term UVR exposure. Additionally, smoking downregulates gene expression of the Notch pathway, which has been reported to enhance the growth of melanoma cells.
Several pharmaceutical agents and their association with melanoma have recently been investigated. A study of all skin cancer cases diagnosed in northern Denmark from 1999 to 2009 revealed a cumulative and dose-dependent protective effect of nonsteroidal anti-inflammatory drugs (NSAIDS) (including aspirin), other nonselective NSAIDs and older COX-2 inhibitors for melanoma and SCC [incidence rate ratios (IRR) 0·87 (95% CI 0·80–0·95) for melanoma and 0·85 (95% CI 0·76–0·94) for ever use (more than two prescriptions) vs. nonuse (fewer than two)]. Moreover, a protective effect of aspirin use in postmenopausal women has been reported, with the greater protection detected in women with ≥ 5 years of aspirin use (hazard ratio 0·70; 95% CI 0·55–0·94). A recent meta-analysis of 10 studies including 490 322 patients, however, showed a negative association of melanoma risk with the use of aspirin or nonaspirin NSAIDs with RRs of 0·96 (95% CI 0·89–1·03) and 1·05 (0·96–1·14), respectively. An increased incidence of melanoma with an odds ratio of 1·88 (95% CI 1·08–3·29) has been reported in patients with inflammatory bowel disease treated with tumour necrosis factor (TNF)-α inhibitors (compared with nonuse). The risk depended on the length of treatment: patients using long-term anti-TNFs had an odds ratio of 3·93 (1·82–8·5), whereas no risk was observed in those who received anti-TNF treatment for < 120 days.
Novel disease associations
History of cancer in childhood
A sixfold increase in the risk of a subsequent malignancy has been reported in patients with a history of cancer in childhood. A recent review of childhood cancer survivors regarding their risk of melanoma found a standardized incidence ratio of 2·42 (95% CI 1·77–3·23), with soft tissue and bone sarcoma, lymphoma/leukaemia and central nervous system tumours preceding the development of melanoma.
Parkinson disease and melanoma
Parkinson disease is a common degenerative disorder of the central nervous system characterized by a loss of melanin-positive, dopamine-secreting cells in the pars compacta region of the substantia nigra. There is significant epidemiological evidence suggesting that patients with Parkinson disease carry a ‘biological advantage’ against a great number of malignancies, as most cancer rates are lower in those patients compared with the general population. This cancer-preventive effect of Parkinson disease does not seem to include melanoma, as suggested by several well-designed epidemiological studies. A recent prospective study of 2106 patients with Parkinson disease showed a sevenfold increased age- or sex-adjusted RR for melanoma compared with the overall incidence according to the American Academy of Dermatology skin cancer screening programmes. Growing evidence also indicates that patients with melanoma have an increased risk for subsequently developing Parkinson disease. The underlying mechanism of this correlation remains unidentified. The association between dopaminergic therapy (levodopa) for Parkinson disease and melanoma development has been strongly disputed.[89, 90] Various pathogenic pathways have been proposed, including genetic risk factors, changes in melanin or melanin synthesis enzymes and defects in autophagy. Clearly, further research is needed to clarify the association of these two distinct diseases, one characterized by cell degeneration and the other by uncontrolled cell proliferation, with melanin playing a central role in both. In view of this intriguing association, patients with Parkinson disease should be more closely monitored for the early detection of possible skin malignancies.
The ongoing trend of rising melanoma incidence rates in most white populations is projected to continue over the next two to three decades. This rise is accentuated in older people aged > 60 years and has lately been observed in younger age groups, possibly marking the effects of an increased recreational exposure to UVR, particularly indoor tanning exposure. Even though mortality rates have stabilized in several countries, as a result of earlier detection and screening, the incidence of thick potentially fatal melanomas remains constant, and is perhaps even increasing, calling for continuous surveillance and rigorous prevention programmes, particularly in high-risk patients. It is now evident that the epidemiology of melanoma is a multifaceted issue reflecting the complexity and heterogeneity of the disease itself. It is hoped that the recent advances in the biology and genetic basis of melanoma that have ushered in the exciting era of targeted treatment for patients with advanced disease will further elucidate the pathogenesis of the disease, generating more efficient and personalized approaches in the diagnostic and preventive setting.
We are indebted to Professor Alan Geller (Department of Social and Behavioral Sciences, Harvard School of Public Health, Boston, MA, U.S.A.) for his valuable comments in the preparatory phase of the manuscript.