A prospective, population-based study of 40,000 women regarding host factors, UV exposure and sunbed use in relation to risk and anatomic site of cutaneous melanoma

Authors


Abstract

Prospective cohort studies about cutaneous melanoma (CM) risk are still few. Host factor- and UVR exposure data were collected prospectively by questionnaire in this population-based cohort study including 40,000 Swedish born women, aged 25–64 years at enrolment (1990). Risk for CM (Cox regression and Stepwise Cox regression [SCR], hazard ratios [HRs] with 95% Confidence Intervals [CI]) in relation to risk factors, age groups (older or younger than 40 years) and primary site, were analyzed. In 29,520 women with complete follow-up through 2007, 155 invasive and 60 insitu CM were recorded. High numbers of nevi (HR, 2.9; 95% CI, 1.7–5.0) and heredity (HR, 3.7; 95% CI, 2.0–6.8) were associated with risk for CM. SCR analysis added red hair as a risk factor. Sunbed use >10 times/year increased risk for women <40 years (HR, 2.5; 95% CI, 1.0–6.2) and a trend for risk associated with sunbathing vacations (HR, 1.4; 95% CI, 1.0–2.0) was shown for women >40 years. Trunk melanoma showed correlations with high numbers of nevi (HR, 3.0; 95% CI, 1.2–7.3) and heredity (HR, 3.2; 95% CI, 1.1–9.4). Head/neck site was correlated to sunbathing vacations (HR, 2.5; 95% CI, 1.2–5.3) and heredity (HR, 7.6; 95% CI, 1.8–31.8). Our study supports divergent etiologic pathways to CM, with high numbers of nevi correlated to increased risk for trunk CM. Furthermore, it confirms that high numbers of nevi, red hair and heredity for CM are the most important risk factors and frequent sunbed use might be a risk factor for younger women.

The incidence of cutaneous melanoma (CM) has increased during the past decades among fair-skinned populations all around the world and is the most fatal form of skin cancer.1 In Sweden, CM represents a considerable growing public health burden, as the incidence has risen five times since the early sixties and the latest reported average annual incidence increase was 4.2% for women and 4.1% for men.2 The explanation to this rapid increase is not entirely understood but risk factors associated with CM are most certainly both related to the host (genetic and phenotypic factors), the environment [e.g., natural and artificial ultraviolet radiation (UVR) exposure] and interactions there between.1

Most previous studies of risk factors for CM have been retrospective case-control studies and meta-analyses,3–9 study types sometimes criticized for the risk of recall bias. Therefore, there is a need for data from prospective, population-based cohort studies,10, 11 which have the advantage of not introducing interfering recall bias (e.g., for UVR exposure). Further, the evaluation of possible underlying risk factors in relation to both old and new CM etiological hypotheses is important to undertake.

Several authors, including Green and Whiteman12, 13 as well as Masback et al.8 have suggested that CM at different body sites may arise through different causal pathways. A related hypothesis is based on the observation that melanomas on different sites harbor different genetic/molecular profiles.14

In detail, Green and Whiteman12, 13 suggest that (i) CM at chronically UVR exposed sites (e.g., head and neck) are associated with melanocytic stability, with low-risk genes, exogenous proliferative factors and “nevus-resistance” versus (ii) CM at intermittently exposed sites (e.g., trunk), which have been associated with “nevus-proneness,” endogenous proliferative factors, high-risk genes, freckles and melanocytic instability/proliferation.8, 12, 13, 15–17

Our aims of this study were to evaluate how heredity, constitutional host factors, different types of UVR exposure for tanning purposes (artificial and from sun) and sunburn history contributed to CM risk in a large prospective, population-based, Swedish female cohort. Another purpose of the study was to evaluate the risk factors possibly associated with primary tumor sites, based on the theories of divergent pathways to melanoma.8, 12, 13, 15–17

Material and Methods

Participants

A prospective, population-based cohort study [Melanoma Inquiry of Southern Sweden (MISS)] was established (1990–1992) in Southern Sweden (Southern Swedish Health Care Region). The Regional Ethical Review Board of Lund University approved the study (LU-34-92, LU-849-06, LU-190-07). In accordance with the study design the cohort should not include women with any prior diagnosis of invasive cancer. Hence, eligible women (born 1926 to 1965 in Sweden) and randomly chosen from the Population/Census Registry were, before inclusion, controlled in the Regional Cancer Registry. Thus, 1,000 healthy women were selected by randomization from each birth cohort.

The final cohort of 40,000 women, aged 25–64 years, were invited by letter to answer a questionnaire about risk factors for CM at enrolment. All primary non-responders were sent one reminder. All participating women gave written informed consent.

The present study period was defined as the time from inception of the study (1990) to the end of follow-up (December 31, 2007). Follow-up time (person-years) was measured from inception of the study until diagnosis of first primary CM, death or the end of the study period.

Questionnaire

The utilized standardized validated18 questionnaire (constructed in the late eighties, including 42 questions in total) mainly concerned suspected risk factors for CM, e.g., heredity for CM, sun tanning behavior, sunburn history, sunbed use, sunscreen use, travelling habits and phenotypic/host trait information as hair color, eye color, freckles and nevi count on the left arm. It further covered questions about life style and sporting habits, parity, smoking habits, alcohol consumption, prescribed drugs as well as questions about marital status, educational level and residence during life. Exactly the same questions, e.g., with an instructed, self-reported nevus count, had successfully been used in several former Swedish case-control studies8, 9, 19 and therefore this specific questionnaire was applied in this prospective study. The participants were encouraged to contact the authors of the questionnaire if questions about how to interpret the questions would arise. In this report questions regarding host factors and UVR exposure were included and evaluated (see Supporting Information III).

Registries and additional diagnoses

The Swedish personal identification number enabled linking of all participants to both the Regional and National Cancer Registries as well as to the National Cause of Death Registry, for information about the incidence of CM and other malignancies, time of death and cause of death. CM included both in situ as well as invasive tumors. Vital status and registry matches were determined up to the end of the study period.

Histopathology

All available CM cases until December 2006 were histopathologically re-examined by one experienced pathologist (A.M.) thus avoiding interobserver variation. Data concerned histopathological type, tumor thickness (Breslow), level of invasion (Clark) and ulceration. The additional histopathological data on the further CM was obtained from the original pathological reports.

Statistical analyses

Frequencies were analyzed (Chi-squared test or two-sided Fisher's exact test when expected values were very low) for comparison of proportions between groups and presented by Odds Ratios (ORs) and 95% Confidence intervals (CIs). Multivariate Cox proportional hazard regression analyses [Hazard ratios (HR)], were applied to assess factors independently associated to the risk for CM. These analyses were performed on (i) the total cohort, (ii) on the cohort subdivided in two age groups: (a) women older or (b) younger than 40 years at the time of enrolment, i.e., women born before or after 1951, and also (iii) on anatomical (site) subgroups. The rationale for breaking up the cohort in two age groups were based on that several Swedish and Norwegian studies as well as recent documentations from the International Agency for Research on Cancer (IARC) have shown correlations (especially in females) between younger age and an increased risk for CM related to artificial UVR exposure.9–11, 19–21 To assess the robustness of the primary Cox proportional hazards regression analyses, forward Stepwise Cox regression (SCR) analyses, based on the significant findings in the univariate analyses, were performed.

p-Values ≤0.05 were considered significant, with the minor exception for the forward SCR model where a p-value of ≤0.1 were allowed, to reveal results close to p = 0.05. Calculations were performed using Statistical Package for the Social Sciences (SPSS 11) software (SPSS Inc., Chicago, IL).

Combining and relabeling variables

In the univariate analyses (Supporting Information Table S2) and in the forward SCR analyses (Table 2), risk factors with formerly four categories were, for statistical reasons, combined and relabelled. The two higher categories (of each tested risk factor) were relabelled as one “high” variable (e.g., “high risk sunburns” was a combination of the categories “3–9 sunburns/year” and “10+ sunburns/year”). In all analyses the continuous nevus category was subdivided into 0, 1–3, >3, where 1–3 nevi was defined as “low numbers of nevi” and >3 nevi was defined as “high numbers of nevi.” Moreover, brown and black hair colors were combined in “brown/black” and the two variables “dark blond” and “medium blond” were combined and relabelled “blond,” to be discriminated from light blond. “Other hair colors” was a combination of greyish, grey, white and unclassified colors.

Concerning the variables sunbed/sun lamp use in the primary Cox regression analyses (Table 1), the two middle categories “1–3 times/year” and “4–10 times/year” were combined (medium use) and the top category >10 times/year was referred to as “highest sunbed/sun lamp use.”

Table 1. Melanoma risk in relation to all women and to age subgroups
inline image
Abbreviations

CI: confidence interval; CM: cutaneous malignant melanoma; HR: hazard ratio; OR: odds ratio; UVA: ultraviolet radiation A; UVB: ultraviolet radiation B; UVR: ultraviolet radiation

Results

Characteristics of participants and melanomas

Twenty-seven women were not possible to contact, giving 39,973 as the initial study cohort receiving the questionnaire. The answering frequency was 74% (n = 29,520). From 1990 to December 2007 (giving 458,950 person-years of follow-up time in total), 2,045 women died and 215 cases of CM were reported in 206 women: 155 invasive and 60 in situ tumors. One hundred forty-nine women (out of 18,331) in the older age group (>40 years) were diagnosed with melanoma, compared to 57 cases out of 11,189 women in the younger group (<40 years). Of all histopathologically re-examined CM only 2% of the diagnoses were changed compared to the original pathology report. The median Breslow thickness for first primary, invasive CM was 0.75 mm [interquartile range (IQR): 0.51–1.43 mm]. Eight (4%) of the 206 women had multiple primary CM, seven of these had two primaries each and one had three primary CM. An univariate analysis indicated that the following factors were associated with increased risk for CM; heredity, freckles, high numbers of nevi, red hair color, high numbers of sunbathing vacations, high numbers of childhood sunburns, high numbers of adolescent sunburns and high use of sun lamps. Detailed melanoma characteristics (including tumor site and histological type) and the univariate analysis are presented as Supporting Information (Supporting Information Tables S1 and S2).

Risk factors for all women and for the two age groups

In the whole cohort, as well as in both age subgroups reported heredity was found to be associated with an increased risk for CM (HR 3.7/4.2/3.1; Table 1). Furthermore, in the whole cohort both low numbers of nevi (HR, 1.8; 95% CI, 1.1–3.2) and high numbers of nevi (HR, 2.9; 95% CI, 1.7–5.0; compared to no nevi) were associated with increased risks for CM.

We found similarly increased risks (approximately threefold) for both low and high numbers of nevi, as well as an association (HR, 1.4; 95% CI, 1.0–2.0) between ever going on sunbathing vacations (versus never) and risk for CM in older women (>40 years; Table 1).

For younger women (<40 years) there was again a strong association between high numbers of nevi (HR, 2.5; 95% CI, 1.2–5.3) and increased risk for CM, which also was the case for the highest degree of sunbed use (HR, 2.5; 95% CI, 1.0–6.2; Table 1). Further, when adjusting also for frequent sunbathing events, the risk associated with highest degree of sunbed use was reduced, but still doubled compared to baseline risk (data not shown).

No other analyzed variables showed significant associations to diagnosis of CM in this model. The risks, in correlation to all different subcohorts, are summarized in Table 1. The applied forward SCR analyses confirmed the results above and in addition, associations between CM diagnosis and red hair color were observed for the whole cohort (HR, 2.9; 95% CI, 1.4–6.0) and for younger women (HR, 4.0; 95% CI, 1.4–11.1) but not for older women (Table 2).

Table 2. Forward stepwise cox regression analysis
inline image

Risk factors in relation to anatomical sites of melanoma

We found that ages at diagnosis varied by anatomical site of CM, with median age 52.0 years for trunk, 60.0 years for CM on extremities and 66.5 years for head and neck CM. Reported heredity for CM was strongly associated with increased risks for all three anatomical sites (Table 3). An association with a high nevus count (HR, 3.0; 95% CI, 1.2–7.3) and a risk for trunk CM was found, while increased risk for head/neck CM was associated with sunbathing vacations (HR, 2.5; 95% CI, 1.2–5.3), even after adjusting for frequent sunbathing events (data not shown). Sunburn blisters were associated with a decreased risk for truncal CM (HR, 0.4; 95% CI, 0.2-1.0).

Table 3. Risk factors for melanomas in relation to anatomic sites
inline image

High numbers of nevi (HR, 2.3; 95% CI, 1.0–5.4) as well as freckles (HR, 1.9; 95% CI, 0.9–3.9) showed tendencies to associations with CM on the extremities. No other analyzed risk factor showed significant association to a specific anatomical site of CM in our model (Table 3).

Discussion

This prospective, population-based cohort study about environmentally and host related risk factors for CM in Swedish women has a high answering frequency and shows that heredity for CM, high numbers of nevi and red hair color are strongly associated with increased risk for CM.

Moreover, sunbathing vacations as a proxy for intense intermittent UVR exposure were associated with an increased CM risk in older women, whereas frequent sunbed use was associated with an increased risk for CM in younger women. Truncal CM were strongly associated with heredity and a high nevus count while head/neck CM were associated with sunbathing vacations.

Heredity and nevi

Our findings of reported heredity (9.5% of the cases and associated with an increased risk for CM in all age groups and for all anatomical sites) are consistent with earlier studies, where approximately 10% of CM cases have reported heredity.22, 23 Cho et al.15 similarly reported that “family history” was associated with increased risks for CM at all sites.

Furthermore, we showed that “high numbers of nevi,” was associated with increased risk for CM24–26 in all age groups but also more specifically with increased risk for truncal CM. The latter is a result in line with the findings by Chang et al.27 who recently showed that an abnormal nevus phenotype was associated mostly with CM on intermittently UVR exposed sites.

The tendency to an association between high numbers of nevi and an increased risk also for limb CM is consistent both with Chang et al.27 and with a recent meta-analysis of Caini et al.,28 thus further supporting the theory of divergent pathways to CM.16 A limitation of our study regarding nevus counts might be that whole body nevus counts were not performed; instead the instructed, self-reported number of nevi/moles on the left arm was used as a proxy for the total body nevus counts. The participants may have found it difficult to differ nevi from sunburn freckles, pigmented seborrhoeic keratoses or solar lentignes and thus under- or over-estimated the number of nevi as no detailed instruction about how to discriminate a nevus from, e.g., a solar lentigo was presented within the questionnaire. However, due to the prospective study design, there is no reason to believe that nevus counts differed among melanoma cases and controls. Total body nevus counts have shown to be strongly correlated to nevus count of one arm.29, 30 Therefore we have considered our results to be both reliable and generalizable.

Sunbeds and sunlamps

The use of artificial UVA/UVB for tanning purposes might possibly result in an increased risk for melanoma, however etiologically difficult to dissect from different types of solar UV exposure. In our prospective study design we wanted to test if artificial UVR exposure was associated with an increased risk for CM, as not only the recent review from IARC21 concluded that also artificial UVR, e.g., from sunbeds and sun lamps, is carcinogenic. Recent studies4, 7 as well as older studies9, 10, 19, 20 have shown that artificial UVR exposure could be associated with increased risk for CM. Nevertheless, limitations of several published studies have been the difficulty to properly adjust for natural UVR (sun) exposure in the statistical analyses. Lazovich et al.7 as well as Cust et al.4 did, however, try to avoid this limitation by rigorously adjusting for sun exposure, and both found an association between CM and UVR exposure from sunbeds. The precise pathogenesis of CM is still unknown, but the carcinogenicity of UVB have been ascribed to direct DNA damage, while the probable carcinogenicity of UVA have been explained by damage to DNA by indirect effects (reactive oxygen species, photoreactions and photoproducts).31, 32 The possibly carcinogenic effect of UVR from sunbeds does most likely not differ from that of natural UVR exposure.

In our cohort the overall use of sunbeds was high (∼50%), as has been noted in earlier studies of Swedish study participants.33, 34 The use of sun lamps was lower, possibly due to that the early sun lamps [mainly emitting Ultraviolet B radiation (UVB)] were restricted after 1982 by Swedish law35 and thereafter only sunbeds/solaria of “UV type 3,”36 with an excess of Ultraviolet A radiation (UVA), have been allowed in tanning parlors and sold for non-medical purposes.37

As sun lamps could represent UVB, as a risk factor for CM, this usage was separated from (UVA-rich) sunbed use in our analyses. The univariate analyses indicated in fact a correlation between sunlamp use and risk for CM, but the multivariate analyses could not confirm this, possibly explained by low numbers.

In contrast, frequent sunbed use was correlated to an increased risk of CM in younger, but not in older, women. The strong relation between frequent usage of sunbeds and an increased risk for CM in younger women has previously been reported in Sweden.9, 19 In addition and further supporting our findings, a large prospective Scandinavian study,10 the reports from IARC20, 21 and a recent study by Cust et al.4 showed associations between sunbed use and the risk for CM, especially evident in younger women.

The reasons for these observations might be that clamshell-like sunbeds expose almost 100% of the body surface to UVA, which consequently and instantly gives very high amounts of intermittent UVA in a short time, compared to the much lesser skin area exposed for solar UVR during common outdoor activities.37 Most modern sunbeds emit mainly UVA and the emission is reported to be more than three times higher compared to the UVA in natural sunshine on a summer day.38 In addition to UVA, most sunbeds nowadays emit a small component of UVB radiation, to intensify the tanning effect and therefore theoretically they may harbor a sunburning capacity. Nonetheless, the younger women observed in our study had exposed themselves to very high amounts of UVA already at an early age, possibly indicating UVA carcinogenicity and/or that the specific time/age when UVA/sunbed exposure takes place might be important. The finding could also be a proxy for an early sun-seeking behavior9, 34 and/or, as suggested by Lazovich et al.,7 a marker for cumulative UVR exposure, thus reflecting a very high total UVR exposure, regardless of source. Unfortunately no specific question in the questionnaire addressed how often the sunbed exposure was associated with sunburn(s), and thus this possible explanation to the detected increase in risk could not be evaluated.

If sunbed exposure might be a proxy for a positive attitude to overall high UVR exposure in different ages this could also probably explain why frequent usage of sunbeds in older women showed no significantly increased risk for CM in our study. The explanations for this observation might possibly be that the relative impact on risk for CM from sunbed use in older women is attenuated with age, due to a relatively stronger impact on risk for CM from added, accumulated sun exposure, as suggested by Cust et al.4

One could also hypothesize that older women possibly might have a more diversified UVR exposure, based on a higher proportion midday solar/outdoor UVR exposure compared to sunbed exposure. If so, solar UVR includes the a higher proportion of UVB, with vitamin D-production properties, while UVA is known to be correlated with decreased levels of Vit D3.39 This might be important, as vitamin D is suggested to have anti-cancer capacities (as involved also in the innate immune system and in regulating cell growth and cell differentiation40), which indirectly could support our findings. UVA mediated DNA damage has also been suggested to activate p53 less effectively than UVB, thus inducing less apoptosis and retaining mutagenic photoproducts in the exposed cells.32 Theoretically this could explain why it would take less time for frequent, high-dose (sunbed-)UVA to induce a CM31, 41 or to “prepare the way”39 for other CM inducing processes, compared to the time needed for CM development after predominantly sun exposure (UVA+a higher proportion of UVB). With our limited data on hand about the frequency, time and amount of solar/outdoor UVR exposure (not only for tanning purposes) this is difficult to evaluate and this is a limitation of this prospective study as our questionnaire was constructed long before it was evident how important it would be to be able to differentiate between types of UV exposure. By re-analyzing the data, adjusting for frequent sunbathing events as a possible proxy for a sun seeking behavior, the association between frequent sunbed use and CM in younger women remained, but was somewhat attenuated. The questionnaire was not compatible with further sun exposure analyses. Recent studies, carefully adjusting for sun exposure, have however reported little confounding effect of sun exposure on the increased risk for CM correlated to sunbed use.4, 7 Other studies3, 10 alike ours with limited information about sun exposure, also indicated that sun exposure probably had little confounding effect on increased risk for CM from sunbed use.

Thus, our findings might truly reflect an increased risk for CM associated with early, frequent sunbed exposure or be a marker for a, so far unknown, confounding factor young women have been exposed to.

Sunburns

Our further measures of intermittent exposure, by proxy of sunburns, were inconclusive. Our univariate analyses indicated that high numbers of sunburns in childhood, adolescence and adulthood were more common in CM cases compared to controls. Thus, our inconsistent results for sunburns in the multivariate analyses were somewhat unexpected and chance findings could probably not be excluded. In addition, our strict definition of sunburn makes it also difficult to compare our results to previous studies with different study designs, using less strict definitions of sunburn. The results observed by us were not in line with a recent meta-analysis by Dennis et al., which showed an association between CM risk and sunburns regardless of when they occurred and further showed a dose-response relationship between number of sunburns and risk of CM.42

As there is no reason to believe that cases and controls would report the number of sunburns differently, due to the prospective study design, our findings must be regarded as reliable. Instead our simultaneous adjustment to several risk/confounding factors in our analyses might possibly explain our non-significant findings due to an overall rather low event rate. Unfortunately our questionnaire did not differentiate sunburn(s) due to solar exposure/during sun bathing vacations from sunburn(s) due to sunbed exposure, giving us difficulties to speculate whether the crucial factors for some of the other observed increased risks in fact were related to sunburns.

Interestingly, sunburn blisters were associated with a decreased risk for truncal CM. This finding might be the result of chance or be due to the small numbers but it could also represent a result of a changed attitude to UVR exposure after severe sunburn.

Other constitutional host factors

In our primary multivariate models, most constitutional host factors (except nevi) showed no associations to increased risk for CM, which is inconsistent with a meta-analysis of Gandini et al.,23 where the importance of phenotypic characteristics was examined and showed RRs of 2.1 for high density of freckles, 1.47 for blue versus dark eyes and 3.64 for red hair color versus dark, respectively. An explanation to our findings might be the current limited number of cases in each group, which however will become larger and gain power in later follow-up studies.

Again, our primary multivariate analyses were simultaneously adjusted to several risk factors as nevi, pigmentary host factors, age and UVR exposure, which might influence and numb the risk evaluation, compared to other studies. However, after constructing the stepwise Cox regression model only red hair color did influence the results, indicating that, e.g., eye color and number of freckles are less important risk predictors in a Scandinavian population, a finding that correlates well with the recent prospective study by Veieröd et al.10

Pathways for melanoma

The hypothesis of divergent pathways for CM development8, 12, 13, 15–17 (see Introduction) was partly supported by our study. We clearly showed that high numbers of nevi was a risk factor for truncal CM and that freckles showed a tendency to a correlation to limb CM, in line with previous studies.8, 15, 27

The found tendency for an association between low numbers of nevi and an increased risk for head/neck CM could further indicate a support for the divergent pathways hypothesis, as could possibly the median age at diagnosis (66.5 years) for head/neck location in contrast to the age at diagnosis for trunk CM (52.0 years).

Furthermore, most studies about intermittent UVR exposure have shown an increased risk especially for CM on anatomical sites with “low chronic exposure” or “high occasional exposure” (e.g., trunk).5, 6, 8, 15, 43–45 We could however not confirm a statistically significant association between our different variates of intermittent sun exposure and trunk CM.

In contrast to the divergent pathway theory, intermittent exposure has also been shown to be associated with risk for CM on frequently sun exposed sites (e.g., head/neck),28 which is in line with some of our findings, where sunbathing vacations increased the risk for head/neck CM. Our finding might of course be a chance finding, as the number of cases in this sub-cohort was low. It might nevertheless possibly reflect an overall high UVR exposure lifestyle/sun-seeking behavior with both intermittent (sunbathing vacations) and chronic UVR exposure. Sunbathing vacations abroad might involve UVR exposure with a higher sunburning effect, with a higher impact on risk, compared to the (presupposed) lower risk for regularly exposed sites as head/neck. Unfortunately, we were prohibited to examine both chronic exposure and sunburns specifically associated to sunbathing vacations, as no proper questions addressing chronic UVR exposure and the specific time of sunburn/s were available in the questionnaire. In Sweden, however, it is quite uncommon that women are spending most of their working hours outdoors (as one possible proxy for chronic sun exposure) and thus the expected rate of chronic sun exposure/outdoor workers in the present cohort of women (excluding senior citizens) would be low.

Further strengths and limitations

A strength with our report is that a validation study about the reproducibility of our self-administered questionnaire for assessment of melanoma risk has been performed, using identical questionnaires/questions as in the present study. In the validation study, Westerdahl et al. found no obvious problems with recall bias and fair to good test–retest reliability was presented.18

The data in the present study was collected in 1990–1991 when media coverage still was less intense than it has been in the last years. Participants could nevertheless be aware of the possible relationship between UVR exposure and CM development and report their exposure in an incorrect way.46 However, the prospective study design ought most likely minimize the possible risk of recall bias, but this is difficult to estimate.47 Variables concerning total UVR exposure would in that case be particularly affected.47 In our study no such variable was present.

A limitation of the study might be that it comprises white-skinned women only and the results may not be generalizable to other ethnic groups or to men. Another limitation of the study was the design of the questionnaire, with few variates evaluating different types of sun exposure, e.g., leisure time spent outdoors and chronic sun exposure. Instead most variates concerned more intense and intermittent sun exposure such as winter sport vacations, sunbathing vacations, history of presupposed deliberate suntans and sunburns. One explanation to this is that the questions were constructed in the late eighties, when the full extent of the possible risk differences connected to the different types of UVR exposure were unknown. Thus the questionnaire limited us to fully discriminate the independent differences in risk of melanoma for sunbed use and, e.g., the risk from the variate sun exposure. Several studies indicate however that the confounding effect of sun exposure on sun bed exposure risks is low.3, 10, 21

Future implications

The cohort participants are still quite young and the follow-up period is still short in relation to CM development. With a longer follow-up time the future data will be more reliable and a second questionnaire was sent to the cohort in 2000–2001 including most of the former questions in addition to more thoroughly asked questions about physical activities, weight, skin phototype and skin reaction to UVR exposure. The latter data will be presented in future reports with the excellent possibility to evaluate changes in behavior over time and to validate recall bias.

Conclusion

This is a unique population-based, prospective study of 40,000 Swedish women, where we have recorded unbiased risk factors for CM, before diagnosis. The study supports the theory of divergent pathways to CM. It also indicates that heredity, high numbers of nevi and red hair color are the most important factors associated with increased risk for CM. Finally, it confirms that frequent use of sunbeds increase the risk for CM in younger women.

Ancillary