Country of origin, age at migration and risk of cutaneous melanoma: A migrant cohort study of 1,100,000 Israeli men

Authors

Errata

This article is corrected by:

  1. Errata: Erratum Volume 134, Issue 4, E2, Article first published online: 9 December 2013

Correspondence to: Hagai Levine, Braun School of Public Health and Community Medicine, Hebrew University-Hadassah Faculty of Medicine, P.O. Box 12272, Kiryat Hadassah, Ein Kerem, Jerusalem 91120, Israel, Tel.: +972-2-6777452, Fax: +972-2-6431086, E-mail: hlevine@hadassah.org.il

Abstract

Cutaneous melanoma (CM) is a common cancer with increasing incidence in many parts of the world where light-skinned populations live. We conducted a large-scale nationally representative migrant cohort study to assess country of origin and age at migration as predictors of CM, controlling for possible confounders. Data on 1,086,569 Israeli Jewish males, who underwent a general health examination before compulsory military service at ages 16–19 between the years 1967–2005, were linked to Israel National Cancer Registry to obtain incident CM up to 2006. Cox proportional hazards was used to model time to event. Overall, 1562 incident cases were detected during 19.3 million person-years of follow-up. Origin was a strong independent predictor of CM. Incidence was higher for European (hazard ratio [HR] = 4.08, 95% confidence interval [CI]: 3.55–4.67) and Israeli origin (HR = 2.92, 95% CI: 2.25–3.79) compared to N. African/Asian origin, adjusted for year of birth, years of education, residential socio-economic position, rural residence and body surface area (or height). Among those of European origin, the adjusted risk was significantly lower for those who immigrated after the age of 10 years (HR = 0.58, 95% CI: 0.45–0.73) but not for younger ages (HR = 1.02, 95% CI 0.84–1.23) compared to Israeli born. The high rates of CM among men of European origin and the almost twofold lower risk among those immigrating after age 10 provide solid support for the deleterious role of childhood sun exposure as a risk factor for melanoma. These findings will serve in directing public health and research efforts.

Abbreviations
BSA

body surface area

CI

confidence interval

CM

cutaneous melanoma

HR

hazard ratio

IARC

international agency for research on cancer

ICD

international classification of diseases

INCR

Israel national cancer registry

NOS

Not otherwise specified

ROC

receiver operating characteristic

SEP

socio-economic position

UV

ultra-violet

Cutaneous melanoma (CM) is a common and potentially lethal cancer with increasing rates in many parts of the world, where light-skinned populations live.[1] The rise in incidence is thought to be linked to changing sun exposure patterns, although many aspects of the etiology of CM are not understood or are poorly quantified.[2] Both genetic and environmental factors are related to the pathogenesis of melanoma. Although not all melanomas are sun related, there is clear and convincing evidence that sun exposure, and more specifically ultra-violet (UV) exposure, is a major environmental cause of melanoma, especially in high-risk populations.[3] Melanoma incidence and mortality among Caucasians correlate with latitude of residence and dose of UV radiation, the highest rates being nearest the equator.[3] In areas as geographically diverse as the United States, New Zealand and Australia, the incidence of melanoma is greater at latitudes closer to the equator. However, in the United States, data from 1992 to 2001 demonstrate that the latitude gradient applies only to non-Hispanic whites.[4] Varying CM incidence rates (highest in Australia, New Zealand, white populations in North America/Europe and the Jewish population in Israel) are most likely due to the combination of levels of sun exposure and innate susceptibility.[5] Known risk factors for CM include multiple or atypical nevi, family history, skin phenotype (such as skin pigmentation, hair and eye color), premalignant and skin cancer lesions and actinic damage indicators.[3, 6] A few studies reported associations with anthropometric measures, including body surface area (BSA) and height, although there is conflicting evidence.[7, 8]

Migrant studies are a well-established methodology to study risk factors for cancer.

Previous ambient exposure studies have reported that fair-skinned peoples born and raised in environments of low solar irradiation exhibited significantly lower risks of CM than people of similar complexion born and raised in sunny environments.[9]

Massive immigration to Israel took place since the establishment of the state, mainly of Jews from Europe, North Africa and Western Asia. At the same time, Israel underwent transition from a developing country to a modern, western life-style, industrial state.[10] Geographically, the state of Israel is located at a latitude of 29°–33° north and is characterized by a subtropical and semiarid climate. Striking differences in risk of CM by origin have been noted previously in Israel.[11-15] This unique setting in Israel and our extensive dataset allowed us to conduct a large-scale migrant cohort study to identify high risk populations and critical periods of sun exposure during childhood, which would aid in directing public health and research efforts. Our objective was to determine the associations of origin and age at immigration with the risk of CM and examine the extent to which these associations are explained by other risk factors such as birth cohort and socio-demographic and anthropometric characteristics.

Material and Methods

Study design, setting and population

Israel is one of the few Western countries where military service is mandatory. All Israeli Jewish adolescents are obligated to present themselves at age 17 for a medical board examination before military service (even if exempted later from service). Consequently, use of these data provides a generally representative sample of the young Jewish population, particularly of males. Our study population included Jewish males who were examined between 1967 and 2005 and were aged 16 to 19 years when examined. They were followed up by data linkage for cancer incidence until the end of 2006. We excluded men with any cancer diagnosis before examination (since 1960) based on the cancer registry (1,101 such cases) as well as those of Ethiopian origin (10,265 males born in Ethiopia and 1,620 Israeli born whose father/grandfather was born in Ethiopia; 0 melanoma cases) who differ significantly from the rest in their skin pigmentation. Females were not included because their baseline data were available only for a more recent period. The study protocol was approved by the Institutional Review Board of the IDF Medical Corps.

Data Sources and Variables

Risk factor data

Origin, socio-demographic and anthropometric data were recorded during the obligatory medical board examination. Subjects who were barefoot and wearing only a shirt and underwear had weight and height measurements taken by trained medical personnel using a beam balance and stadiometer. Origin was defined as European or N. African/Asian according to country of birth for immigrants to Israel, or by paternal country of birth for the Israeli born, or grandfather's country of birth if both the participant and father were Israeli-born. Israeli origin was defined for the relatively uncommon fourth generation (or more) Israelis, who are a mixture of European (mostly), North African and Asian ancestries. European origin also included countries of emigration from Europe (the Americas, Australia and Southern Africa). Immigration status (separately for European and for N. African/Asian origin) was classified as Israeli born, immigration at ≤10.0 years of age and at >10.0 years of age (hereafter, this variable is referred to as “age at migration”). In an additional analysis, age at migration was grouped into 5 year categories.

BSA was calculated according to Mosteller's formula as the square root of weight in kilograms multiplied by height in centimeters and divided by 3,600.[16] We grouped BSA into quintiles of our study sample: Q1, <1.630; Q2, 1.630–1.713; Q3, 1.714–1.790; Q4, 1.791–1.893; Q5, >1.893. We grouped height in centimeters into quintiles: Q1, <169; Q2, 169–172; Q3, 173–175; Q4, 176–179; Q5, >179.

Year of birth was dealt with as a continuous variable. Years of schooling were grouped into ≤9, 10, 11 and ≥ 12 years. Residential socio-economic position (SEP) was based on the town/city of residence, according to a national classification of 10 clusters by geographical units, which we categorized as low (1–4), medium (5–7) and high (8–10).[17] Place of residence was classified as urban or rural (< 2,000 inhabitants). Data were complete for anthropometric measures and origin and nearly complete for SEP, years of schooling and place of residence (0.6%, 0.1% and 1.0% missing, respectively).

Outcome data

Cases of CM in our cohort were identified by linkage to the Israel National Cancer Registry (INCR) based on gender, birth date and the personal identification number given to all Israeli citizens at birth or immigration. The INCR, a population-based registry in operation since 1960, meets internationally accepted requirements for the coding system (ICDO-Version 3) and completeness of data. Reporting is mandatory by Israeli law since 1982 and completeness of the registry is considered to be about 95% for all types of solid cancer and has maintained high coverage since inception of the registry in the 1960s.[18] The INCR data include personal identification, the date of diagnosis, site affected, ICD code of the tumor and histologic description of the tumor. Only histologically verified CM (the only definite method of diagnosing CM), registered between 1967 and 2006, the last year with data at the time of linkage, were included, and grouped as invasive melanoma or melanoma in situ. The tumor skin site was grouped as axial (head and neck or trunk), distal (lower or upper limbs) or other (skin NOS and overlapping lesion of skin).

Statistical Methods

Chi-square tests were used to assess differences in baseline and cancer characteristics by origin, and the t-test to assess differences in age at diagnosis. Cox proportional-hazards regression models were used to assess the association between origin and time to CM diagnosis adjusting for year of birth, BSA, years of schooling, residential SEP and rural/urban dwelling (Table 3, Fig. 1a). Due to the strong association between height and BSA (Pearson correlation = 0.64, p <0.001), each variable was separately included in the multivariable model. For the origin and education variables, the largest groups were chosen as the reference category: N. African/Asian origin and ≥ 12 years of education. As year of birth was associated with the incidence of CM (adjusted hazard ratio [HR] 1.04; 95% confidence interval [CI] 1.03–1.05 for each additional year), we introduced year of birth into all single predictor (“univariate”) models as well as the multivariable analyses. An additional analysis restricted to European (or N. African/Asian) origin was used to assess the association between age at migration and time to CM diagnosis adjusting for the same covariates (Table 4, Fig. 1b). All analyses were repeated with exclusion of melanoma in situ. Multivariable analyses were carried with all variables that were statistically significant in the univariate analyses. A stepwise forward likelihood ratio procedure was used with p to enter <0.05 and p to exit ≥ 0.10 to include the covariates into the model. 95% CIs were computed for the HRs and nominal p values are presented. Inspection of log minus log plots for each variable verified the assumption of proportionality of the hazards. We tested the area under the receiver-operating-characteristic (ROC) curve (C statistics) to evaluate the discriminatory ability of the multivariable models to predict CM (Figs. 2a and 2b). Analyses were performed with IBM SPSS statistics version 19.

Figure 1.

(a) Cumulative adjusted incidence of CM according to origin among a cohort of Israeli men examined from 1967 to 2005 and followed up to 2006. (b) Cumulative adjusted incidence of CM according to age at migration*, among European origin (adjusted HR =1.77; 95% CI 1.33–2.38, p < 0.001 for a direct comparison of those who immigrated at ≤10.0 years of age with those at ages >10.0 years of age). *The curve for immigration at age ≤10 years almost coincides with the Israeli born.

Figure 2.

ROC curve for prediction of CM among a cohort of Israeli men examined from 1967 to 2005 and followed up to 2006. (a) Prediction by origin (adjusted for birth cohort) for the entire cohort (area under ROC curve 0.829; 95% CI 0.819–0.838, p < 0.001). (b) Prediction by age at migration (adjusted for birth cohort) for European origin only (area under ROC curve 0.791; 95% CI 0.780–0.801, p < 0.001).

Table 1. Baseline characteristics by origin of the study cohort of Israeli men examined from 1967 to 2005 and followed up to 2006
 n (%) or mean ± SD
CharacteristicAll participants (N = 1,086,569)aEuropean (N = 464,441)N. African/Asian (N = 566,836)b
  1. a

    The small Israel origin group (n = 55,292) is not shown separately but is included in the total.

  2. b

    All characteristics differ between European and N. African/Asian origin (p < 0.001 for each).

Age at examination, mean ± SD (Range in years)17.4 ± 0.4 (16.0–19.9)17.4 ± 0.5 (16.0–19.9)17.3 ± 0.4 (16.0–19.9)
Anthropometrics
Height in centimeters173.6 ± 6.8174.7 ± 6.8172.6 ± 6.7
Body surface area (Mosteller's formula)1.77 ± 0.171.79 ± 0.171.74 ± 0.16
Origin
N. Africa/Asia566,836 (52.2) (100)
Europe464,441(42.7)(100) 
Israel55,292 (5.1)  
Socioeconomic position
Low289,989 (26.8)95,077 (20.6)178,668 (31.7)
Medium554,260 (51.3)243,163 (52.8)287,056 (50.9)
High235,809 (21.8)122,563 (26.6)98,544 (17.5)
Years of schooling
≤9117,248 (10.8)24,410 (5.3)88,091 (15.5)
1099,705 (9.2)34,199 (7.4)61,536 (10.9)
11282,045 (26.0)116,878 (25.2)148,429 (26.2)
≥12586,739 (54.0)288,424 (62.2)268,498 (47.4)
Place of residence
Urban959,654 (89.2)396,938 (86.6)514,944 (91.5)
Rural115,935 (10.8)61,562 (13.4)47,538 (8.5)
Table 2. Cancer site and morphology of 1,562 incident cases of cutaneous melanoma by origin in a cohort of 1,086,569 Israeli men examined from 1967 to 2005 and followed up to 2006
 n (%)
CharacteristicTotalaEuropeanN. African/Asianb
  1. a

    The small Israel origin group (n = 55,292, 74 Melanoma cases) is not shown separately but is included in the total.

  2. b

    Cancer characteristics did not differ between European and N. African/Asian origin (p > 0.05 for each).

Site Total1,562 (100)1,201 (100)287 (100)
Axial (head & neck or trunk)811 (51.9)631 (52.5)144 (50.2)
Distal (lower or upper limbs)475 (30.6)366 (30.5)89 (31.0)
Other276 (17.5)204 (17.0)54 (18.8)
Morphology
Invasive melanoma1,163 (74.5)887 (73.9)223 (77.7)
Melanoma in situ399 (25.5)314 (26.1)64 (22.3)
Table 3. Cox regression hazard ratios (HR) of cutaneous melanoma with 95% confidence interval (CI), univariate and multivariable analyses, among Israeli males (N = 1,074,493, 1,531 cases)a
 Univariate analysisbMultivariable analysisc
CharacteristicHR (95% CI)pHR (95% CI)p
  1. a

    12,076 (1.1%) males [including 31 cutaneous melanoma cases (2.0%)] with missing covariate data were excluded for this analysis.

  2. b

    All “univariate” analyses were in fact adjusted for year of birth introduced as a continuous model.

  3. c

    Mutually adjusted for all variables in the multivariable analysis. Two models were run: one including body surface area (BSA) without height, and the second including height without BSA, due to the strong correlation between the two anthropometric variables. The covariate coefficients shown are those from the BSA model.

Origin <0.001 <0.001
N. Africa/Asia1.0 1.0 
Europe5.24 (4.61–5.97)<0.0014.08 (3.55–4.67)<0.001
Israel3.41 (2.64–4.40)<0.0012.92 (2.25–3.79)<0.001
Year of birth1.04 (1.03–1.05)<0.0011.04 (1.03–1.05)<0.001
Body surface area (Mosteller's formula) <0.001 0.001
1st quintile (<1.630)1.0 1.0 
2nd quintile (1.630–1.713)1.60 (1.30–1.86)<0.0011.24 (1.04–1.49)0.018
3rd quintile (1.714–1.790)1.84 (1.55–2.19)<0.0011.29 (1.08–1.54)0.005
4th quintile (1.791–1.893)2.16 (1.82–2.56)<0.0011.36 (1.14–1.62)0.001
5th quintile (> 1.893)2.54 (2.14–3.02)<0.0011.46 (1.23–1.74)<0.001
Height in centimeters <0.001 <0.001
1st quintile (<169)1.0 1.0 
2nd quintile (169–172)1.33 (1.13–1.56)0.0011.06 (0.90–1.25)0.50
3rd quintile (173–175)1.63 (1.38–1.93)<0.0011.18 (0.99–1.40)0.06
4th quintile (176–179)1.84 (1.57–2.17)<0.0011.23 (1.04–1.45)0.016
5th quintile (> 179)2.42 (2.06–2.83)<0.0011.40 (1.19–1.65)<0.001
Socioeconomic position <0.001 0.001
Low1.0 1.0 
Medium1.60 (1.39–1.84)<0.0011.24 (1.07–1.42)0.03
High2.09 (1.80–2.43)<0.0011.39 (1.19–1.62)<0.001
Years of schooling <0.001 <0.001
≤90.23 (0.19–0.29)<0.0010.46 (0.37–0.57)<0.001
100.48 (0.40–0.58)<0.0010.71 (0.59–0.86)<0.001
110.65 (0.54–0.77)<0.0010.79 (0.67–0.95)0.011
≥121.0 1.0 
Place of residence <0.001 0.002
Urban1.0 1.0 
Rural1.51 (1.31–1.74) 1.25 (1.09–1.44) 
Table 4. Cox regression hazard ratios (HR) of cutaneous melanoma with 95% confidence interval (CI), multivariable analysis, among male Israelis of European origin (N = 457,832, 1,178 cases)
CharacteristicHR (95% CI)p value
Age at immigration <0.001
Israeli born1.0 
Immigrated ≤10 years of age1.02 (0.84–1.23)0.87
Immigrated > 10 years of age0.58 (0.45–0.73)<0.001
Year of birth1.03 (1.02–1.04)<0.001
Body surface area (Mosteller's formula) 0.08
1st quintile (<1.630)1.0 
2nd quintile (1.630–1.713)1.17 (0.94–1.45)0.16
3rd quintile (1.714–1.790)1.24 (1.01–1.53)0.043
4th quintile (1.791–1.893)1.24 (1.01–1.52)0.042
5th quintile (> 1.893)1.34 (1.09–1.65)0.005
Socioeconomic position 0.02
Low1.0 
Medium1.24 (1.05–1.48)0.01
High1.29 (1.07–1.55)0.007
Years of schooling <0.001
≤ 90.52 (0.39–0.68)<0.001
100.78 (0.63–0.97)0.03
110.87 (0.71–1.07)0.18
≥ 121.0 
Place of residence
Urban1.0 
Rural1.31 (1.13–1.52)<0.001

Results

Sample characteristics

Table 1 presents the baseline characteristics of all study participants and by origin. Altogether, 1,086,569 Israeli males were eligible for the study and were followed for an average of 17.8 years, constituting 19.3 million person-years (not shown in tables). The mean age at examination was 17.4 years (standard deviation [SD] = 0.43); 84.3% were examined at age 17. The origin was split between N. Africa/Asia (52.2%) and Europe (42.7%) and a minority being of Israel origin (5.1%). All covariates differed (p <0.001) by origin (Europe vs. N. Africa/Asia).

During the follow-up, 1,562 CM cases were diagnosed. Mean age at diagnosis (not shown in tables) was 38.9 (SD = 9.6), higher for European origin (39.3) than for N. African/Asian origin (37.7, p = 0.04) and Israeli origin (36.4, p = 0.03). There were no significant differences in cancer site (p = 0.70) or morphology distribution (p = 0.20) between those of European versus N. African/Asian origin (Table 2).

Association with covariates

Cases had a higher BSA (mean difference = 0.25, 95% CI 0.17–0.33, p <0.001) and were taller (mean 1.2 cm (95% CI 0.9–1.6, p <0.001) than the rest of the cohort. The positive association of BSA with the risk of CM found in the “univariate” model, persisted, albeit attenuated, in the multivariable model, as did residential SEP, years of schooling and rural residence (Table 3). In a separate model, when height was included instead of BSA, height showed a similarly graded positive association with CM incidence, and results for origin and other covariates were similar to the model with BSA.

Association with origin

The strong association of origin with CM in the “univariate” model was only modestly attenuated after multivariable adjustment for the covariates: European origin (adjusted HR 4.08; 95% CI 3.55–4.67) >Israeli origin (adjusted HR 2.92; 95% CI 2.25–3.79) >N. African/Asian origin (reference category) (Table 3, Fig. 1a). The incidence was significantly lower for Israeli origin (i.e., 4th generation immigrants) compared to European origin in a direct comparison (adjusted HR = 0.73; 95% CI 0.57–0.92, p = 0.008). Overall, the multivariable model was highly predictive of CM (area under ROC curve 0.829; 95% CI 0.819–0.838, p <0.001), whereas the model adjusted solely for year of birth was nearly as predictive (area under ROC curve 0.818; 95% CI 0.808–0.828, p <0.001) (Fig. 2a), implying that the additional contribution of the covariates was minor.

Association with age at immigration

To explore the association of age at immigration with CM incidence, the analysis was restricted to males of European origin which comprises Israeli-born (N = 349,177), immigrants at ages ≤10 years (N = 55,139) and above 10 years (N = 60,125). The mean age at immigration for immigrants born in Europe was 10.2 years (SD = 5.6). There were no significant differences in age at diagnosis, cancer site or morphologic distribution by age at migration among the European origin group. Compared to Israeli-born males of European origin, the risk was much lower for those who immigrated after the age of 10 years (adjusted HR = 0.58; 95 CI% 0.45–0.73, p <0.001), but not for those who immigrated at ages ≤10 years (adjusted HR = 1.02; 95% CI 0.84–1.23). (Table 4, Fig. 1b). A direct comparison of those who immigrated at ≤10.0 years of age with those at ages >10.0 years of age showed substantial excess risk (adjusted HR = 1.77; 95% CI 1.33–2.38, p <0.001). The other risk factor associations were generally similar among the Europeans and the full cohort. We repeated the analysis by further stratifying by age at immigration in 5-year age bands with the same covariates in the model. Compared to Israeli born, the risk by age of immigration was significantly lower for ages 10.1–15.0 (HR = 0.60; 95% CI 0.43–0.84, p = 0.003) and 15.1–19.9 (HR = 0.55; 95% CI 0.39–0.77, p = 0.001), but not for ages 0–5.0 (HR = 1.03; 95% CI 0.79–1.34, p = 0.84) or 5.1–10.0 (HR = 1.01, 95% CI 0.78–1.29, p = 0.97) (data not shown in tables). Overall, the year of birth adjusted model in Europeans was highly predictive of CM (area under ROC curve 0.791; 95% CI 0.780–0.801, p <0.001), with the multivariable model adding very little (area under ROC curve 0.797; 95% CI 0.786–0.807, p <0.001) (Fig. 2b).

In a parallel analysis restricted to the N. African/Asian origin group, the association with CM did not differ significantly by age at migration.

Additional analyses

We repeated all analyses excluding the melanoma in situ cases (N = 399) [see Supporting Information Tables 1 and 2]. Overall, the risk factors coefficients for invasive melanoma were remarkably similar to those shown in Tables 3 and 4, although the CIs were expectedly slightly wider. Those of European origin had higher risk for invasive melanoma compared to the N. Africa/Asia origin (adjusted HR 3.98; 95% CI 3.40–4.66). Compared to Israeli-born of European origin, the risk was substantially lower for those of who immigrated after the age of 10 years (adjusted HR = 0.53; 95 CI% 0.40–0.71, p <0.001), but not for those who immigrated at ages ≤10 years (adjusted HR = 1.06; 95% CI 0.85–1.31).

Discussion

In this cohort of over 1 million Israeli male adolescents followed prospectively for 19.3 million person-years up to 2006, over 1,500 incident cases of CM were observed. After adjusting for a number of possible confounders, the association of origin with CM persisted strongly, with the incidence being lowest for North African and Asian origin, intermediate for the Israeli origin, and highest for European origin. Among those of European origin, higher risk was evident for the Israeli-born and for immigrants before the age of 10, compared to those who immigrated at an older age. The incidence was positively associated with later years of birth, BSA, height, markers of higher SEP and rural dwelling. Overall, origin and age at immigration (for European origin) were strongly predictive of CM incidence.

CM is a skin cancer occurring almost exclusively in white populations, whereas its incidence remains very low in populations of African or Asian origin with darker pigmentation.[1] Based on the INCR, age-adjusted invasive melanoma incidence rates per 100,000 Jewish males in 2008 were: 14.6 for Israel born, 17.2 for European/American born, 6.4 for Asian born and 5.9 for North African born.[19] In comparison, similarly adjusted IARC data showed sharper contrast with rates per 100,000 of 7.6 for populations in Europe, 15.8 in North America, 1.5 in West Asia and 0.4 in North Africa.[20] As expected, we found a strikingly different incidence of melanoma by origin, with low rates among the North African and Asian origin group and highest rates among those of European origin, similar to previous studies in the Israeli population.[11-15] In our cohort, we were able to explore for the first time the extent to which these differences were confounded or mediated by other risk factors such as birth cohort, height/BSA and socio-demographic characteristics. The strong association of origin with CM risk was only moderately attenuated after adjustment for these covariates. These findings, in addition to the lack of association of CM among N. African/Asian origin with age at migration, support the claim that the differing risk by origin is predominantly due to the protective effects of darker skin color. The small Israeli origin group (fourth generation), which had intermediate risk for CM, was characterized by lower years of education and lower residential SEP compared to European origin, pointing to the ethnic admixture of this group.

Migrant studies have been used to examine sun exposure and melanoma risk by comparing incidence in populations who have migrated between different geographical areas. Previous migrant studies provided some evidence of childhood being a critical time for future melanoma development, that is, an increased risk for individuals who spent their childhood in sunny geographical locations.[9, 21] In a case–control study in Australia, migrants arriving before age 10 years appeared to have a risk similar to that of native born Australians, whereas the estimated incidence in those arriving after age 15 years was around one quarter of the native born rate.[22, 23] Similar findings were found in a case–control study involving European patients among whom arrival before age of 10 years to a sunny location of residence (such as the Mediterranean) conferred a fourfold increased risk of developing melanoma.[24] However, a previous systematic review considered case–control studies to constitute of lower-quality evidence due to the inherent difficulties of measuring historical sun exposure.[9] Earlier studies in Israel reported a one-third lower risk among European migrants to Israel compared to Israeli-born Jews.[12-14, 25] A more recent study of melanoma occurring (in both sexes) up to the age of 30 did not find a significant difference between immigrants from Europe and those of Israeli origin (Israeli-born, father born in Israel),[15, 18] whereas the risk for the Israeli-born of European origin parents was significantly higher than for European migrants. However, age at migration was not explored.

In our study, taking advantage of the population-based examination of nearly all adolescent male Jews in Israel, we were able to implement a cohort design, examining age at immigration in addition to origin and other pertinent variables. Direct comparison of Israeli-born males of European origin to European-born immigrants allowed us to isolate the effect of exposure to the sunny Israeli environment by age at migration. We found age at migration to be highly predictive of CM risk, with a nearly twofold lower risk among those who immigrated after the age of 10 years, compared to Israeli-born males of European origin and compared to the European-born who immigrated before the age of 10 years. Thus, age at immigration was strongly associated with the risk of CM, independent of birth cohort, and anthropometric and socio-demographic characteristics. We did not find differences in anatomical distribution or morphology by origin group or age at migration. Elaboration of the risk by age at immigration by refinement of the age categories showed that the risk was similar for immigration at ages 0–5.0 or 5.1–10 and higher than at ages 10.1–15.0 or 15.1–19.9, which showed similar risk. Our findings point to the susceptibility of fair-skinned children immigrating to sunny countries before the age of 10 years, supporting early childhood sun exposure to be a critical period in terms of future melanoma risk.

Identifying childhood sun exposure as a critical period for CM risk carries significant implications for public health efforts. Recent global analysis has shown that while incidence rates continue to rise in most European countries (primarily Southern and Eastern Europe), rates have stabilized in recent years in Australia, New Zealand, the United States, Canada, Norway and Israel.[26] The authors concluded that their findings provide support that primary prevention can halt and reverse the trend of an increasing burden of melanoma. They also indicate that such prevention measures require further endorsement in many countries. Our study, together with previous studies, direct attention to the need for health promotion efforts specifically directed at children and their parents and care-takers. These efforts are mostly needed for populations with fair skin born in or migrating at a young age to sunny countries. The United States Preventive Services Task Force recommends counseling children, adolescents, and young adults aged 10 to 24 years who have fair skin about minimizing their exposure to UV radiation to reduce risk for skin cancer.[27] The American Cancer Society advises on the importance of protecting children from the sun because of the increased risk for cancer resulting from severe sunburns in childhood.[28]

Recommended interventions at the community level include policy and regulation to increase preventive behaviors (such as covering up, using shade, avoiding the sun during peak UV hours) among populations in specific settings, including primary school and outdoor recreational settings.[29] Recommendations referring to the using of sunscreen may be also beneficial, although further evidence for their efficiency is needed, especially for children.[30] Our study serves to stress the importance of origin and age at migration as factors to consider when evaluating personal risk as well as targeting populations for health promotion activities. A plethora of activities to reduce sun exposure and prevent skin cancer have been undertaken in Israel since 1992, initiated by the Israel Cancer Association.[31] Encouraged by the recent stabilization of CM incidence trends in Israel, such efforts should be continued.

Strengths of our migrant cohort include its prospective nature, population-based design reflecting mandatory health examinations among 17-year olds in our population, large sample size, long follow-up, high degree of completeness of the cancer registry data throughout study period and the ability to adjust for possible confounders such as anthropometric, birth cohort and SEP factors.

This analysis also has certain limitations. We did not have direct information and could not adjust for skin phenotype or family history of melanoma or for sun exposure and related health behavior such as sunscreen use. Hence, we cannot determine whether the lower risk found among immigrants from Europe at older ages is due to lower susceptibility at older ages or lower cumulative exposure. Our setting of a baseline examination in adolescence did not allow us to assess whether risk continues to decline for immigrants at older ages. We had data on country of birth and paternal country of birth, but not on maternal origin. However, maternal and paternal origins in our population have previously been shown to be highly associated.[32] We had shorter follow-up periods for the more recent birth cohorts; therefore, our analysis better represents risks for CM manifesting at young adulthood, rather than lifetime risk. Lifetime risk can be better assessed in the future by continued follow-up of our cohort in the coming years.

In conclusion, we found higher rates of CM among men of European origin, but significantly lower rates among those who emigrated from Europe after age 10 years. These findings provide solid support for the importance of childhood sun exposure, particularly in light skinned people, as a preventable risk factor for CM and can aid in directing public health and research efforts.

Ancillary