International trends in the incidence of malignant melanoma 1953–2008—are recent generations at higher or lower risk?

Authors

  • Friederike Erdmann,

    Corresponding author
    1. Section of Environment and Radiation, International Agency for Research on Cancer (IARC), Lyon, France
    2. Association of Dermatological Prevention (ADP) e.V., Hamburg, Germany
    3. EUROSKIN e.V., Buxtehude, Germany
    • Section of Environment and Radiation, International Agency for Research on Cancer (IARC), 150, cours Albert Thomas, 69372 Lyon Cedex 08, France
    Search for more papers by this author
    • Tel.: +33-(0)4-72-73-84-63, Fax: +33-(0)4-72-73-83-20

  • Joannie Lortet-Tieulent,

    1. Section of Cancer Information, International Agency for Research on Cancer (IARC), Lyon, France
    Search for more papers by this author
  • Joachim Schüz,

    1. Section of Environment and Radiation, International Agency for Research on Cancer (IARC), Lyon, France
    Search for more papers by this author
  • Hajo Zeeb,

    1. Department of Prevention and Evaluation, BIPS—Institute for Epidemiology and Prevention Research, Bremen, Germany
    Search for more papers by this author
  • Rüdiger Greinert,

    1. Association of Dermatological Prevention (ADP) e.V., Hamburg, Germany
    2. EUROSKIN e.V., Buxtehude, Germany
    3. Centre of Dermatology, Buxtehude, Germany
    Search for more papers by this author
  • Eckhard W. Breitbart,

    1. Association of Dermatological Prevention (ADP) e.V., Hamburg, Germany
    2. EUROSKIN e.V., Buxtehude, Germany
    3. Centre of Dermatology, Buxtehude, Germany
    Search for more papers by this author
  • Freddie Bray

    1. Section of Cancer Information, International Agency for Research on Cancer (IARC), Lyon, France
    Search for more papers by this author

Abstract

The incidence of cutaneous malignant melanoma has steadily increased over the past 50 years in predominately fair-skinned populations. This increase is reported to have leveled off recently in several Northern and Western European countries, Australia, New Zealand and in North America. We studied the global patterns and time trends in incidence of melanoma by country and sex, with a focus on and age- and cohort-specific variations. We analyzed the incidence data from 39 population-based cancer registries, examining all-ages and age-truncated standardized incidence rates of melanoma, estimating the annual percentage change and incidence rate ratios from age-period-cohort models. Incidence rates of melanoma continue to rise in most European countries (primarily Southern and Eastern Europe), whereas in Australia, New Zealand, the U.S., Canada, Israel and Norway, rates have become rather stable in recent years. Indications of a stabilization or decreasing trend were observed mainly in the youngest age group (25–44 years). Rates have been rising steadily in generations born up to the end of the 1940s, followed by a stabilization or decline in rates for more recently born cohorts in Australia, New Zealand, the U.S., Canada and Norway. In addition to the birth cohort effect, there was a suggestion of a period-related influence on melanoma trends in certain populations. Although our findings provide support that primary and secondary prevention can halt and reverse the observed increasing burden of melanoma, they also indicate that those prevention measures require further endorsement in many countries.

According to estimates for 2008, there were almost 200,000 new cases of invasive cutaneous malignant melanoma (hereafter called melanoma) globally, and an estimated 46,000 deaths from the disease.1 The vast majority of cases (almost 85%) occur in developed countries, where melanoma ranks as the sixth most frequently diagnosed cancer overall. The incidence of melanoma has steadily increased over the last 50 years in most fair-skinned populations.2–6 Such observations have been well documented in Europe3, 7–12 (particularly in the Nordic countries),2, 3 North America,13, 14 Australia and New Zealand,5, 6 where incidence rates remain the highest reported worldwide. Nevertheless, several studies published during the last decade have reported rates in several of these populations as either stabilizing or declining, including a number of Northern and Western European countries,3 the U.S.,13 Canada,14 Australia5, 15 and New Zealand.6, 16

In general, the increase in melanoma incidence is ascribed to changes in attitudes toward sun bathing, namely from sun-avoidance toward sun-seeking behavior,17, 18 the latter promoted by Coco Chanel in the early 1920s.19 However, recent evidence suggests that the current trend of increased melanoma is predominantly explained by an increasing incidence of thinner melanomas,5, 8–10, 20, 21 as a result of both improved surveillance techniques and earlier diagnosis alongside a growing awareness of skin cancer in the general population.22

Findings from seminal migrant studies indicate the importance of the interaction between environmental, genetic and individual host factors in the etiology of melanoma occurrence.23, 24 Important risk factors in the development of melanoma include the number of common nevi and atypical nevi, skin phenotype, a family history of melanoma, actinic damage, a history of sunburns and exposure to ultraviolet (UV) radiation (especially in childhood).23, 25–28 It is not clear at present whether sunburn is simply an indicator of high intermittent sun exposure or whether it is an additional, independent risk factor for melanoma.26 Intermittent sun exposure, mainly during childhood, is thought to be a causal factor for melanoma.18, 23, 29 In older patients, however, the disease is thought be more related to chronic exposure, given the higher frequency of chronically exposed body sites in this age group, implying divergent causal pathways for melanoma.30, 31 However, sun exposure patterns might be of less importance than the individual response to exposure to UV radiation.22, 26, 27

Although a steady improvement in survival among melanoma patients has been reported over the last decades,2, 20, 32 the mortality rates from melanoma continue to rise in several European countries as a result of increasing incidence,3, 20, 32–34 most pronounced in older patients. In contrast, mortality curves are stable or have even begun to fall in Australia,34, 35 the U.S.,34 Scandinavia2 and the UK.22 Primary and secondary prevention activities are of major public health relevance in the reduction of the incidence of, and mortality from melanoma, even if the former can take several decades until effects are realized at the population level.19, 36

To better understand the evolution of melanoma occurrence at the national level, this descriptive study provides a comprehensive overview of the patterns and trends in incidence rates in 39 countries across four continents, based on published data from high-quality population-based cancer registries. Temporal variations in melanoma incidence are presented for broad age groups and by sex, with an emphasis on trends within the last 10-year calendar period and among recent birth cohorts. We discuss trends in the context of identifying population subgroups at an increased risk of melanoma, as a means to define target groups for future primary prevention and early detection strategies.

Material and Methods

To examine time trends in the incidence rates of melanoma (ICD-10 C43), regional or national population-based cancer registry, data were extracted from Cancer Incidence in Five Continents (CI5) Volume I to IX.37 The specific inclusion requirement was the availability of at least 9 consecutive years of data and compilation in the last volume of the CI5 series (Volume IX), a criterion indicative of each registry's data quality, given that the editorial process involves a detailed assessment of the comparability, completeness and validity of the incidence data. To increase the timeliness of the information (the last year of diagnosis available in Volume IX of CI5 is 2002), the registry was supplemented with more recent data available online from various cancer registry-specific websites. Of the 39 countries studied, we obtained national data for 23 countries. For the remaining countries, regional registry data were aggregated to obtain a proxy of the (unknown) national incidence (55 regional registries in total, see Table 1). The varying start-up and overall years available for each registry within a given country led to a pragmatic selection of registries that sought to maximize both the number of included registries and the length of period of study. In addition, we obtained data by race (U.S. Blacks and U.S. Whites) from the SEER program (9 registries). The time span of observations at the country level varied from 9 to 56 years in the final dataset. Corresponding population data were obtained from the same sources as the incidence data.

Table 1. Populations included in the analysis, time span, mean annual number of cases and person-years, mean annual incidence rate and M:F ratio over the 2000–2002 period
inline image

Cases were stratified by sex and 5-year age-group. Age-standardized incidence rates (ASR) per 100,000 were estimated using the world standard population and cumulative risk of a diagnosis of melanoma (up to age 74) used as a measurement of lifetime risk, assuming no other causes of death were in operation.38 A correction factor was applied to the ASR in Colombia to take into account the cases of unknown age.39 To examine time trends by age (25–44, 45–64 and 65 and over), we selected 16 countries, chosen on the basis of having sufficiently long time series and number of cases available to allow interpretation of the observed stratum-specific rates. Furthermore, we analyzed time trends by birth cohorts in eight selected populations, which were specially selected to represent international generational trends of melanoma with sufficient length of the available time series, elevated numbers of cases or level of risk, adequate population size and representation of a given region or population at the global level. To graphically summarize the direction of the trends, locally weighted regression (lowess) curves were fitted to provide smoothed lines through scatterplots of ASRs by calendar period. A bandwidth of 0.3 was used, for example, 30% of the data were used in smoothing each point. Rates were plotted on a semilog scale.

Trends were quantified by calculating the estimated annual percentage change (EAPC) and the corresponding 95% confidence interval (CI) over the last 10 available years on fitting calendar year as a continuous variable to the incident counts using Poisson regression, offset by the corresponding logarithm of the person-years. EAPCs were calculated only for those countries with at least 10 years of data, a last year of available data within the period 2004–2008 and with annual incidence rates per annum of ≥1 per 100,000 person-years.

Ten-year cohorts were obtained on subtracting the mid-points of 5-year age-group from the corresponding mid-years of 5-year calendar time. The trends are presented as rates versus birth cohort by age. Assuming the incidence rates were constant within the 5-year age classes a and 5-year periods of diagnosis p, an age-period-cohort (APC) model was fitted. We assumed the number of new cases followed a Poisson random variable with the logarithm of the person-years at risk specified as an offset:

equation image

Thus, birth cohorts c are computed as c = pa. The effects were estimated and presented using the full APC model. The relative straightforwardness of fitting APC models is, however, at odds with the difficulties in providing an informed presentation of the model parameters, given the irresolvable issue of nonidentifiability. To provide a unique and nonarbitrary solution, we constrained the linear component of the period effect to have zero slope, and therefore assumed that the linear changes in melanoma incidence were a result of cohort-related factors. Such assumptions are credible if the suggested changes in exposure to UV radiation can be considered the result of societal or peer-related influences that put men and women at higher (or lower) risk of melanoma in successive generations. As the solution presented is entirely dependent on our choice of allocation of the overall time trend (drift), caution should be applied when interpreting the results. The model analysis and presentation was performed using APCfit40 in Stata. We used the default number of internal knots (five) for each of the spline bases for the three variables (A, P and C), and hence, the internal knots were placed at each quartile.

Results

Geographical variations by sex

Table 1 presents the country and sex-specific ASR for the period 2000–2002. Incidence rates of malignant melanoma in men and women are similar in most countries, although there are notable exceptions, with a higher incidence observed in men in high-risk Australia, New Zealand and the U.S. Whites as well as in low-risk populations including Japan and Philippines; women had higher rates than men in a number of Northern and Western European countries (Table 1).

There is a clear relationship between the lifetime risk—the estimated risk of getting melanoma before the age of 75 years—and the respective country's predominant skin type and geographical location (Fig. 1). In Australia and New Zealand, the incidence rates tend to be two to three times higher than they are observed anywhere else over the study period; a person living in these countries has a lifetime risk of 3.6%, compared with 1.9% in the U.S. Whites, 1.1% in Canada and ranging from 0.3 to 1.6% in European countries (lifetime risk for both sexes combined, data not shown).

Figure 1.

Cumulative risk of 0–74 years in 2000–2002. Bar chart of cumulative risk of the incidence of melanoma by country by sex, sorted in descending order of cumulative risk in men. Time period is 2000–2002, except for Spain (1998–2000). Cumulative risks based on data from regional registries and groups of regional registries are marked with *. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

In Europe, the highest incidence rates are observed in the Nordic countries of Norway, Denmark and Iceland with the lifetime risk ranging from 1.3 to 1.6%, with lifetime risk also high in Switzerland (1.6%), the Netherlands (1.2%) and the Czech Republic (1.0%). In contrast, rates in Mediterranean countries tend to be lower as well as in the Baltic countries and the Eastern European countries under study. The lifetime risk of developing melanoma in Asian countries (other than Israel) is very low in comparison.

Time trends: rates versus calendar period

All-ages adjusted incidence rates of melanoma in men and women are presented in Figure 2. The incidence rates among populations that are predominantly Caucasian have increased over the past decades. The rising incidence is pronounced and steady in most countries in Europe (Fig. 2b). The baseline incidence (ca. 1955) is evidently low in the Nordic countries (where registration is longstanding), while a minor deceleration in the trends in both sexes can be observed from the early 1990s, in Sweden and more notably in Norway, where rates have been rather stable since 1995, particularly among women. In contrast, rates in Finland and particularly in Denmark show no indication of leveling off.

Figure 2.

Age-standardized incidence trends by sex in (a) the Americas, Asia and Oceania; in (b) Europe. Time trends in melanoma incidence rates (age standardized, world standard) by country and geographic area, by sex, depicted using on a semilog scale. Rates <0.2 per 100,000 person-years are not plotted. The curves are locally weighted regressions (lowess) of the age-standardized rates by calendar period. The closed circles represent the observed values. Countries are ordered by geographical region, using the United Nations classification (Eastern, Northern, Southern and Western Europe). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

There are indications that the incidence rates of Australia, New Zealand, U.S. Whites, Canada and Israel have stabilized (Fig. 2a). A leveling off in some European populations is observed among women in Lithuania, Estonia, Slovakia, Ireland and Scotland and among men in Lithuania and Switzerland. Nevertheless, the changes are too recent to be considered unambiguous evidence of changing risk of melanoma. Similarly, there are the first signs of a decline in incidence rates in France and notably (although subject to considerable random variation) in Icelandic women with an EAPC of −8.4% (−22.0, 5.2) from 2004 through 2008 (data not shown). In contrast, the incidence rates continue to climb in other regions of Europe, including much of Southern and Eastern Europe.

Rates in Asian countries (with the exception of Israel) tend to be low and possibly stable, whereas rates are only slightly higher in the South American countries examined, with rates among U.S. Blacks intermediate. Their interpretation is somewhat hampered by the small numbers in the numerator of the rates in many of these populations, resulting in a high degree of random variation inherent in the visual description of the trends.

The incidence levels in Northern Europe, particularly in Iceland, are higher among women over time, although the direction and magnitude of the trends in rates generally do not appear to differ much between the sexes in Europe. In Australia, Canada and U.S. Whites, incidence rates in men were constantly higher than in women since mid of the 1980s, while in New Zealand, rates in women surpassed those of men by 1993.

Time trends: rates versus calendar period by age

Figure 3 displays trends in incidence by age (25–44, 45–64, 65+ years) for males (Fig. 3a) and females (Fig. 3b) in 16 higher-risk countries; Figure 4 describes the EAPC in melanoma in the last 10 years available across all ages and in the youngest age group.

Figure 3.

Age-standardized incidence trends by age: (a) males and (b) females. Time trends in melanoma incidence rates by age group (age-standardized and world standard), by country and geographic area and by sex on a semilog scale. Rates <1 per 100,000 person-years are not plotted. The curves are locally weighted regressions (lowess) of the age-standardized rates by calendar period. The closed circles represent the observed values. Iceland: started plot in 1976 to overcome excessive random variations in the previous years. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Figure 4.

Estimated annual percentage change in melanoma incidence in the last 10 years available (all ages and ages 25–44) with corresponding 95% CI. Statistics are presented for registries with at least 10 years of data, last year of data in 2004–2008 and with annual incidence rates ≥1 per 100,000 person-years. Countries are sorted in decreasing order of EAPC in men. Countries are sorted in decreasing order of EAPC in men. For the names of registries presented, see caption of Table 1. For the periods presented, see Table 2. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Uniformly increasing trends across age groups are seen over the last decades in all study populations, with the highest rates of increase in melanoma seen among older males and females (diagnosed aged 65 and over). The male rates in the oldest age group tend to be higher than their female counterparts, whereas rates in the youngest age group appear more elevated among women.

A key observation in the age-specific trends—most notably in the youngest age group (ages 25–44)—is an indication of stability or slight declines in trends in Australia, New Zealand, Canada, U.S. White (in men), Israel, Norway and Czech Republic (in women). Very recently, rates also in U.S. White women have a tendency to level off and rates in Iceland and France (in women) even show a declining trend (Figs. 3 and 4).

In women, rates in Australia, New Zealand, Canada, Czech Republic and Norway were stabilized in the 1980s, while a slight decrease was observed in Australia by 1998. A leveling off, plateau or even a slight decline in melanoma among females aged 25–44 was observed in the last decade in Australia (−1.8% per annum [−2.4, −1.1]), New Zealand (0.5% [−1.6, 2.7]), Canada (0.1% [−0.9, 1.1]), Norway (1.0% [−1.3, 3.4]), and among men in Australia (−1.2% [−2.0, −0.3]), in New Zealand (−0.2% [−2.6, 2.2]), in Canada (−1.0% [−2.3, 0.4]), in U.S. Whites (0.1% [−0.8, 1.1]) and in Norway (−0.4% [−3.9, 3.1]) (Table 2). However, only the minor declines observed in Australia and Icelandic women reached statistical significance. Rates in U.S. White males aged 25–44 appear rather stable and contrast with the increasing rates observed in older U.S. men and in U.S. White women irrespective of age. In general, the slightly decreasing tendencies are more apparent among men than women in the countries.

Table 2. Estimated annual percentage change in melanoma incidence in the last 10 years available and its 95% CI (all ages and ages 25–44). Statistics are presented for registries with at least 10 years of data and with annual incidence rates ≥1/100,000 person-years.
inline image

Other than Norway, Iceland and France, incidence rates continue to climb in all age group in most European populations studied.

Time trends: Rates versus calendar period and birth cohort by age

Figure 5 compares rates versus birth cohort and rates versus calendar period by 5-year age groups of diagnosis for men and women in eight countries, which were selected on the basis of their sufficient time series and number of cases. The parallelism exhibited in the rates versus birth cohort on the semilog scale is indicative of a strong cohort effect and a generational influence at the population level. Uniform increases in incidence rates of melanoma are seen in successive male and female birth cohorts in all eight countries, up to the late 1940s. This observation is followed by a stabilization or decline in rates among more recently born cohorts in selected countries: in Australia, New Zealand, the U.S. (Whites), Canada and Norway. Other than in Norway, no decline or stabilization is apparent in the incidence rates in the other European countries under study.

Figure 5.

Incidence rates versus birth cohorts and period among men (a) and women (b). Melanoma incidence rates versus birth cohorts and rates versus period, by country and age (25–74) for eight selected countries. Rates are displayed on a semilog scale. For each graph, the age-specific lines depict, from bottom to top, the rates in the age-groups 25–29, 30–24, …, 70–74, respectively. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

An analysis of the goodness of fit of the APC model and the two-factor alternatives revealed that the full APC model fitted the data in 12 of the 16 populations (eight populations by sex), and that both nonlinear period and nonlinear cohort effects made a statistically significant contribution in explaining the variation over and above the model with age and drift in 13 and 16 of the 16 populations, respectively (data not shown).

On the basis of the full APC model and a parameterization that allocates drift to birth cohort (Fig. 6), an increasing trend in the incidence rate ratios (IRRs) can be seen in the Czech Republic, Denmark, England (in both sexes) and in White U.S. women. In contrast, generational declines or stabilizations among men are seen among birth cohorts from Australia, New Zealand, Canada and (perhaps less evidently) in Norway since the 1950s. A similar picture emerges for women, although in Norway, a leveling off or decrease in the IRR in postwar cohorts is less evident. Rather, along with an increasing IRR in cohorts born 1950 onward, there appears to be a period-related decline from the 1980s (also apparent in Denmark but to a lesser extent). On inspection of the observed rates by calendar period in Norway, the period decrease appears most evident in women aged under 40 (Fig. 5b).

Figure 6.

Melanoma incidence rate ratios per 100,000 person-years and corresponding 95% CIs in eight selected countries for men and women aged 25–74, by birth cohort and by period. The default is to make the reference points at the median value (with respect to the number of cases) for the period and cohort variable. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Discussion

This study presents a comprehensive and up-to-date analysis of time trends in the incidence rates of cutaneous malignant melanoma worldwide, highlighting the major sex-specific variations in risk of melanoma between countries and the divergent trends according to the three components of time (age, calendar period and birth cohort). In global terms, the highest melanoma incidence rates are by far those observed in New Zealand, Australia and the U.S. Whites. In Europe, rates are elevated in the Nordic populations, Switzerland, the Netherlands and the Czech Republic, while Southern European populations tend to have a lower incidence. The burden of melanoma among U.S. Black as well as in the Asian and Latin American countries under study is low and rather constant over time, as is consistent with previously reported observations.41 This pattern confirms a relationship between the predominant skin type of the inhabitants of a country and the geographic latitude and is in line with findings from previous studies.2, 3, 15, 18, 42, 43

The all-age melanoma trends by calendar period confirm the uniform increases in incidence rates of predominantly fair-skinned populations reported over the past decades.2–7, 9–15 In general, such trends have been attributed to socioeconomically linked changes in lifestyle and social behavior involving an increasing amount of UV exposure. Inspired by Coco Chanel in the 1920s, tanned skin became a symbol of health and wealth in the 1930s, triggering a radical change in attitudes from sun avoidance to sun-seeking behavior. Various lifestyle changes including an increasing number of holidays spent in sunny destinations (and subsequent intermittent and cumulative sun exposure), more revealing fashion trends (e.g., the bikini in the 1960s) and more outdoor leisure activities in general have led to an increasing body surface area exposed to UV radiation.17–19, 44 In the 1960s, artificial UV systems were introduced and indoor tanning became increasingly popular over the following decades,19, 28 in Europe particularly in the Northern countries, even among children and adolescents.28, 42, 45 It has been estimated that the effect on the skin via solar and artificial UV radiation is equivalent,18, 28 and in 2009, the International Agency for Research on Cancer classified UV radiation (UVA, UVB and UVC) as well as the use of tanning devices as “carcinogenic to humans” (Group 1).46 High intermittent sun exposure as a result of high prevalence of sunbed use, outdoor leisure activities as well as traveling abroad to sunny destinations may thus have led to the elevated rates in the Nordic countries and to the observed north–south gradient in Europe.

Besides a UV radiation-related true increase in incidence, much of the recent rises in many European countries can probably be attributed to an increasing proportion of thinner tumors. Studies according to Breslow thickness have identified rapid rises in the incidence of thin melanomas in various European countries8–11, 20, 21, 47 as well as in Queensland in Australia,5 whereas the proportion of thin melanomas is currently rather stable in Australia (at the national level),43 New Zealand16 and the U.S.48 Furthermore, no increasing trend of thick melanomas in either sex has been reported for France, Italy, Northern Ireland and Germany.8, 11, 21, 47 Changes in coding or registration practices may have perhaps contributed less to the reported increases49 than an increasing awareness of skin cancer among health professionals50 and the general public.

Incidence rates are still uniformly increasing in most European countries, including almost all Southern and Eastern European countries studied. However, trends in several high-risk countries including Australia and New Zealand, North America (U.S. Whites and Canada), Israel and Norway (and perhaps very recently in France and Iceland) suggest a recent leveling off or slight decline in melanoma incidence rates, whereas the observed trends reach statistical significance only in Australia (at ages 25–44 years) and Iceland (among women, all ages). This finding, more readily apparent among young persons aged 25–44 years of age, is in accordance with earlier observations of stable incidence trends among young people in Australia (Queensland,5, 15 New South Wales51), Canada,52 the U.S.13 and some Northern European countries3 and may be associated with changing UV radiation exposure patterns following a growing awareness of skin cancer and its risk factors.

On visual examination (Figs. 5a and 5b), the recent changes in the rates over time on a semilogarithmic scale appear more related to birth cohort (the parallelism indicative of a generational effect) than calendar period. Incidence rates in the aforementioned countries—Australia, New Zealand, North America and Norway—increased among successive birth cohorts up until the late 1940s after which rates tended to stabilize or decline among consecutive generations born in the 1960s and 1970s. The observation lends weight to the hypothesis that a change in generational UV exposure is a key contributor to the recent promising changes in melanoma risk in these higher-risk populations, and that childhood may represent a critical time window for melanoma risk at later ages.23, 31 Although it can be speculated that more recent generations already benefit from public health campaigns of the last decades, particularly introduced in Australia, New Zealand and the U.S. but also in several European countries,19, 53, 54 the time lag between the changing prevalence in UV exposure and resulting stabilizations or declines in incidence rates might be too short to be the direct result of such interventions.22 Nevertheless, in equivalent period terms, declines in the IRRs (Fig. 6) are observed in Denmark and Norway during the 1990s, suggesting changes in behavior, possibly through the impact of interventions, may have had an impact at a specific point in calendar time, irrespective of age. Further work might be warranted to ascertain if and when changes in trends by calendar period are observed in different countries, either by estimating and testing local curvature effects or by comparing the slopes of the linear trends a priori using predefined contrasts.

In contrast, incidence rates continue to climb even in the young age group in most European regions, for example, in Denmark, England, Slovenia and Spain. Findings of some European studies indicate that the rise among young people is in part attributable to an increasing number of thin lesions.8, 10, 20 This hypothesis is most strongly supported by the observed trends in the 25- to 44-year-old age group in countries exhibiting either stable rates (e.g., in New Zealand and the U.S.) or slightly decreasing rates (e.g., in Australia) that mirror the constant proportions of thin melanoma among young people reported in the U.S.48 and New Zealand16 and a declining rate of thin melanomas in Australia.43 Studies examining the incidence of melanoma by body site have reported a rising incidence of melanoma on the trunk or limbs, locations that are usually less exposed to sunlight and for which melanomas are more commonly diagnosed in young people.8, 10, 20, 42, 55 Interestingly, a rising proportion of melanomas diagnosed at the trunk is also reported for young Icelandic42 and U.S.55 women. The most common (and most rapidly increasing) histological type for melanoma is superficially spreading melanoma,7–10, 18, 20 a type which is associated with thinner tumors and intense, intermittent UV exposure.20, 31 The body site distribution and most common type of melanoma among young people lead to the assertion of a high intermittent UV exposure in this age group. This observation contrasts somewhat with the observed trends seen here in young women from both the U.S. and Iceland.

The key strengths of this study are the comprehensive nature of this comparison: geographical variations were examined in 39 countries in four continents; overall trends were examined in 39, trends by age in 16 and trends by calendar period and birth cohort in eight countries. We have used the most recent data of high quality from 84 population-based cancer registries, and our study may represent the first truly comprehensive global assessment of melanoma incidence trends over half a century. Our study also has limitations. We had, for example, no information on tumor stage, thickness, body site and histological subtype. The monitoring of trends in incidence by tumor thickness could be used to evaluate early detection efforts4 and is a useful indicator of morbidity and mortality generally, given that the survival is related to the tumor thickness at diagnosis.4, 9, 18 Melanoma diagnosed at different body sites and histological types provide important information on the probable patterns of UV radiation exposure,4, 30, 31 and would offer more insight into time trends,31 particularly stratified analyses of birth cohort effects. Furthermore, as with many studies based on routine data, our analysis lacks information on individual risk factors such as skin type and sunbathing behavior. Several cancer registries already collect detailed information on melanoma characteristics56; and in terms of future research, a more exhaustive dataset would help clarify the risk factors responsible for the observed patterns and trends in melanoma rates.

Although the selection of the countries for the subanalyses was performed in terms of a sufficient time series and number of cases, we cannot exclude an over-representation of populations with high quality and longstanding registries that are more likely to have long-term melanoma prevention strategies and perhaps as a result, more favorable incidence trends. In such an instance, our summary of recent trends might be skewed toward overoptimism with respect to the future patterns of melanoma burden.

In the context of APC models, it is not possible to identify the linear trends attributable to period or birth cohort, and we made a crude (but given a priori evidence, nonarbitrary) assumption that the trend (drift) was due to generational changes so as to allow a unique presentation of the IRRs, as in Figure 6. We cannot exclude the possibility of an underlying linear period-based trend in some countries, perhaps driven by a gradually increasing awareness of melanoma among both the general public and among health professionals in the last decades; such an effect would tend to attenuate the model-based cohort-specific increases and accentuate the declines.

The considerable variations in melanoma trends are likely to involve the interplay of several factors. Country-specific variations are presumably related to interactions between UV radiation exposure and skin type,23, 24 postulated divergent pathways to melanoma (intermittent vs. cumulative exposure)29, 31 and critical age periods of exposure.23, 26 The different prevention activities in the respective countries may have also influenced the trends. Prevention activities were first initiated in Australia from the 1960s, and later, coordinated primary prevention campaigns from the early 1980s (Slip! Slop! Slap!) were introduced, with the objective to encourage individuals to reduce their exposure to UV radiation.19, 54 Finally in 1988, the more comprehensive SunSmart skin cancer prevention program was implemented and is still operating.54

Comparable activities (education programs and public awareness campaigns) were implemented in the mid-1980s in the U.S.34 and also in several European countries,18, 53 for example, in Sweden12 and the UK.9 However, the impact of primary prevention measures on incidence rates of melanoma is unlikely to be seen in the near future.15, 53

Currently, available data may lead to conflicting interpretations in this respect. On one hand, the slowly decreasing incidence rates in the younger age groups may be an indication that primary prevention is having an impact by changing attitudes to UV among recent generations toward better sun protection or avoidance. On the other hand, the body site distribution of melanoma among younger people implies rather frequent sunbathing behavior in younger generations.

The reduction in the incidence of melanoma observed in several high-risk countries among younger persons and in recent birth cohorts are encouraging, and underscore the importance of targeted solar and artificial radiation awareness campaigns as well as secondary prevention activities to ensure the future reduction in incidence of, and mortality from the disease.19, 36 Important target groups for primary prevention activities are children, adolescents and their parents.19 The need for the further development and implementation of such population approaches to primary prevention is perhaps most apparent in Europe, where uniformly increasing incidence rates of melanoma continue in many countries.

Acknowledgements

The work reported in this paper was undertaken while Ms. Erdmann was hosted by the International Agency for Research on Cancer. No specific funding was received for this study. The Association of Dermatological Prevention (ADP) e.V. supported Ms. Erdmann's stay at the International Agency for Research on Cancer. The authors thank the following cancer registries (Director in parentheses) who are participating investigators, having contributed their data that served as a basis for this article: Australia—Australian Capital Territory Cancer Registry (Dr. Halliday), New South Wales Central Australia Cancer Registry (Tracey), Northern Territory of Australia Cancer Registry (Dr. Condon, Dr. Guthridge), Queensland Cancer Registry (Dr. Aitken), South Australian Cancer Registry (Dr. Luke) and Tasmanian Cancer Registry (Prof. Venn); Austria—Austrian Cancer Registry (Zielonke); Belarus—Belarusian Cancer Registry (Prof. Piliptsevich, Levin and Dr. Grakovich); Belgium—Belgian Cancer Registry (Dr. Van Eycken); Brazil—Cancer Registry of Goiânia (Prof. Edesio); Bulgaria—Bulgarian National Cancer Registry (Dr. Dimitrova); Canada—Alberta Cancer Registry (Russell), British Columbia Cancer Research Centre (Tamaro), Manitoba Cancer Registry (Dr. Turner, Musto), New Brunswick Provincial Cancer Registry (Leonfellner, Dr. Balram, Dr. Kumar), Newfoundland Cancer Registry (Ryan), Nova Scotia Cancer Registry (McIntyre, Dr. Dewar), Northwest territories Cancer Registry (Dr. Clearsky, Dr. Kandola), Ontario Cancer Registry (Dr. Marrett, King), Prince Edward Island Cancer Registry (Dr. Vriends, Dr. Dryer) and Saskatchewan Cancer Registry (Stuart-Panko); China—Hong Kong Cancer Registry (Dr. Law) and Shanghai Cancer Registry (Dr. Zheng); Colombia—Cali Cancer Registry (Dr. Bravo); Costa Rica—Costa Rica National Tumor Registry (Dr. Ortiz Barbosa); Croatia—Croatian National Cancer Registry (Dr. Znaor); Czech Republic—Czech National Cancer Registry (Dr. Holub, Dr. Srb and Prof. Abrahamova); Denmark—Danish Cancer Registry (Dr. Larsen and Gjerstorff); Ecuador—National Cancer Registry of Tumores (Dr. Yepez, Dr. Cueva Alaya); Estonia—Estonian Cancer Registry (Dr. Mägi); Finland—Finnish Cancer Registry (Prof. Hakulinen); France—Calvados General Cancer Registry (Dr. Guizard), Doubs Cancer Registry (Dr. Woronoff), Isere Cancer Registry (Dr. Colonna), Bas-Rhin Cancer Registry (Prof. Velten), Somme Cancer Registry (Dr. Lapôtre-Ledoux and Prof. Ganry) and Tarn Cancer Registry (Dr. Grosclaude); Iceland—Icelandic Cancer Registry (Prof. Tryggvadottir and Prof. Jonasson); India—Chennai Metropolitan Tumour Registry (Dr. Shanta and Dr. Swaminathan); Ireland—National Cancer Registry Ireland (Dr. Comber); Israel—Israel National Cancer Registry (Dr. Barchana); Italy—Ferrara Province Cancer Registry (Dr. Ferretti), Latina Province Cancer Registry (Pannozzo), Tumour Registry of Modena (Prof. Federico), Parma Tumor Registry (Dr. Michiara), Romagna Tumor Registry (Dr. Falcini), and Veneto Tumor Registry (Dr. Zambon); Japan—Osaka Cancer Registry (Dr. Tsukuma), Nagasaki Prefectural Cancer Registry (Dr. Soda), Miyagi Prefectural Cancer Registry (Dr. Nishino), Yamagata Prefectural Cancer Registry (Dr. Shibata and Dr. Matsuda); Lithuania—Lithuanian Cancer Registry (Dr. Smailyte); The Netherlands—Netherlands Cancer Registry (Dr. Hoefsmit); New Zealand—New Zealand Cancer Registry (Hanna); Norway—Cancer registry of Norway (Dr. Ursin); Philippines—Manila Cancer Registry (Dr. Baut and Dr. Laudico), DOH Rizal Cancer Registry (Dr. Esteban and Dr. Mirasol-Lumague); Poland—Cracow City and District Cancer Registry (Dr. Rachtan), Kielce Regional Cancer Registry (Dr. Gozdz, Dr. Mezyk) and Warsaw Cancer Registry (Dr. Zwierko); Republic of Korea—Republic of Korea Cancer Registry (Dr. Sohee Park); Singapore—Singapore Cancer Registry (Prof. Shanmugaratnam, Prof. Lee, Dr. Luming Shi and Dr. Khuan Yew Chow), Slovakia—Slovakia National Cancer Registry (Dr. Chakameh Safaei Diba); Slovenia—Cancer Registry of Republic of Slovenia (Prof. Primic Žakelj); Spain—Granada Cancer Registry (Dr. Sánchez Perez), Murcia Cancer Registry (Dr. Navarro Sánchez), Navarra Cancer Registry (Dr. Ardanaz Aicua), Tarragona Cancer Registry (Dr. Galceran) and Zaragoza Cancer Registry (Dr. Pilar Rodrigo); Sweden—Swedish Cancer Registry (Klint); Switzerland—Geneva Cancer Registry (Prof. Bouchardy) and St. Gallen-Appenzell Cancer Registry (Dr. Ess); Thailand—Chiang Mai Cancer Registry (Dr. Songphol Srisuhkho and Dr. Yupa Sumitsawan), Khon Kaen Provincial Registry (Dr. Supannee Sriamporn and Dr. Surapon Wiangnon); United Kingdom—Scotland (Information Services Division), United Kingdom—England (Office for National Statistics); U.S.A.—Connecticut Tumor Registry, Hawaii Registry (Dr. Goodman and Dr. Hernandez), State Health Registry of Iowa (Dr. Lynch and Dr. Platz), New Mexico Tumor Registry (Williams and Dr. Key), Utah Cancer Registry (Dibble and Dr. Stroup), San Francisco-Oakland Registry (Dr. Glaser), Detroit Registry (Dr. Schwartz), Seattle-Puget Sound Registry (Dr. Schwartz and Dr. Li), Atlanta Registry (Dr. Ward and Dr. Young).

Ancillary