Broad-Spectrum Sunscreen Use and the Development of New Nevi in White Children a Randomized Controlled Trial

ORIGINAL CONTRIBUTION

Broad-Spectrum Sunscreen Use and the Development of New Nevi in White Children A Randomized Controlled Trial

Richard P. Gallagher, MA Context High nevus density is a risk factor for cutaneous malignant melanoma. Me- Jason K. Rivers, MD, FRCPC lanocytic nevi originate in childhood and are largely caused by solar exposure. Tim K. Lee, MSc Objective To determine whether use of broad-spectrum, high–sun protection factor (SPF) sunscreen attenuates development of nevi in white children. Chris D. Bajdik, MMath Design Randomized trial conducted June 1993 to May 1996. David I. McLean, MD, FRCPC Setting and Participants A total of 458 Vancouver, British Columbia, schoolchil- Andrew J. Coldman, PhD dren in grades 1 and 4 were randomized in 1993. After exclusion of nonwhite children and those lost to follow-up or with missing data, 309 children remained for analy- STRONG RISK FACTOR FOR THE sis. Each child’s nevi were enumerated at the start and end of the study in 1996. development of cutaneous malignant melanoma (CMM) Intervention Parents of children randomly assigned to the treatment group (n=222) received a supply of SPF 30 broad-spectrum sunscreen with directions to apply it to in white populations is the exposed sites when the child was expected to be in the sun for 30 minutes or more. presenceA of acquired melanocytic Children randomly assigned to the control group (n=236) received no sunscreen and 1-4 nevi. There is a consistent rise in risk were given no advice about sunscreen use. of CMM with increasing number of nevi Main Outcome Measure Number of new nevi acquired during the 3 years of the in virtually every study that has as- study, compared between treatment and control groups. sessed this relationship.3-6 The presence of remnants of preexisting nevi Results Children in the sunscreen group developed fewer nevi than did children in 7 the control group (median counts, 24 vs 28; P=.048). A significant interaction was in about 50% of CMMs indicates that detected between freckling and study group, indicating that sunscreen use was much acquired nevi are precursor lesions more important for children with freckles than for children without. Modeling of the 8,9 for many, although not all, mela- data suggests that freckled children assigned to a broad-spectrum sunscreen inter- nomas.9 vention would develop 30% to 40% fewer new nevi than freckled children assigned Recent work has focused on the ori- to the control group. gin and etiology of nevi in children, who Conclusions Our data indicate that broad-spectrum sunscreens may attenuate the are, for the most part, born without number of nevi in white children, especially if they have freckles. nevi. Fewer than 2% of children have JAMA. 2000;283:2955-2960 www.jama.com a congenital nevus,10,11 although acquired nevi begin to become clinically velopment of acquired nevi is sunlight With this in mind, we have conducted obvious at an early age.12 Etiologic stud- exposure as measured by sunburn his- a randomized controlled trial to see ies have shown that host and pigmen- tory,13 latitude of residence,19 or re- whether broad-spectrum high–sun pro- tary characteristics (eg, light skin color, ported solar exposure.14,16 tection factor (SPF) sunscreen use freckling, propensity to burn in the sun) Reducing acquired nevi in children might attenuate the number of new nevi that raise adult risk of CMM also pre- may reduce their risk of CMM as adults. that develop in white children. dispose children to develop high nevus density.13-16 Genetic factors also in- Author Affiliations: Cancer Control Research Pro- Columbia (Mr Gallagher and Drs Rivers and McLean), gram, British Columbia Cancer Agency (Messrs Gal- Vancouver. fluence nevus prevalence, with higher lagher, Lee, and Bajdik and Dr Coldman), Depart- Corresponding Author and Reprints: Richard P. Gal- counts of melanocytic nevi in mela- ment of Health Care and Epidemiology, University of lagher, MA, Cancer Control Research Program, Brit- 17,18 British Columbia (Messrs Gallagher and Bajdik), Divi- ish Columbia Cancer Agency, 600 W Tenth Ave, Van- noma-prone families. The princi- sions of Dermatology, British Columbia Cancer Agency, couver, British Columbia, Canada V5Z 4E6 (e-mail: pal environmental risk factor for the de- Vancouver Hospital, and University of British [email protected]).

MATERIALS AND METHODS to the sunscreen group received a bottle counted by 2 of the 3 physicians and of SPF 30 broad-spectrum sunscreen 17 (4%) were counted by all 3 physi- Study Design and Data Collection near the end of each school year in June cians. Assuming the variance among the The study was approved by the British 1993, 1994, and 1995. Parents were in- duplicate and triplicate counts was typi- Columbia Cancer Agency and Univer- structed to apply the sunscreen in cal, the proportion of variance in whole- sity of British Columbia research eth- amounts they usually used to all sun- body nevus counts attributable to the ics committees. Six Vancouver elemen- exposed sites on the enrolled child effect of the counter was less than 5%. tary schools with the largest proportion whenever he/she was expected to be in Data were used only if students com- of white children were selected for the the sun for 30 minutes or more. Par- pleted the whole protocol, defined as study. School principals were ap- ents were specifically asked to use the the intake and exit nevus counts, the proached for permission to conduct the particular bottle of sunscreen only on intake questionnaire, and at least 2 of study within their schools. After secur- the enrolled child. At the end of July the 3 summer sun update, Christmas ing permission from the Vancouver each year, a second bottle of sun- break, and spring break question- School Board, the principals released screen was sent. Parents were then naires. If 1 of the summer sun updates names of all children in grades 1 and 4 asked to measure and report how much was not completed, mean values from (aged 6-7 and 9-10 years, respec- of the original bottle had been used by the other 2 such questionnaires were tively) and their parents to the study. marking what remained in the first substituted. The same procedure was Parents were sent a letter explaining the bottle on an actual-size diagram of the followed for missing Christmas and study and were asked for written per- sunscreen bottle. Parents were in- spring break questionnaires. mission to examine each child and en- structed to use the second bottle of sun- roll the child and a parent in the 3-year screen on the index child for the re- Data Analysis investigation. mainder of the summer and the next It is customary in clinical trials to con- At enrollment, each student was ex- Christmas and spring breaks. Parents duct an analysis based on intent-to- amined by either a dermatologist whose children were randomized to the treat. In this study, no intermediate (J.K.R.) known for his expertise in control group were given no advice as nevus counts were taken between ran- childhood nevus studies19-21 or by a phy- to sunscreen use, and no placebo was domization and conclusion. It is there- sician specially trained by him. All nevi, provided. Because of the level of gen- fore not possible to conduct an intent- regardless of size, were counted using eral education about sun exposure, to-treat analysis based on imputed techniques outlined in the Interna- however, use of sunscreen was sub- end-point values for subjects who were tional Agency for Research on Cancer stantial in the control group. lost to follow-up during the course of counting protocol.21 The scalp, geni- At the end of each summer vaca- the study. tal area, and buttocks were not exam- tion, solar exposure during the previ- Several measures of sun exposure ined, nor was the breast area in girls. ous 3 months was determined for chil- were calculated. Minimal erythemal Degree of freckling on the face, dren in each study group using an dose (MED) information for clear sky shoulders, and arms was estimated us- activity-based questionnaire. Cloth- conditions by latitude and month of the ing a chart13 with good observer repro- ing preference and sunscreen use dur- year were obtained from Diffey and El- ducibility. Height and weight of each ing outdoor activities were assessed on wood.23 Vacation solar exposure in child were taken to allow calculation a semiquantitative basis. Similar instru- MEDs during the 3 years was assessed of body surface area.22 Skin reflec- ments were used to evaluate solar ex- using location, latitude, and month of tance on a non–sun-exposed site (up- posure during the Christmas and spring the vacation, assuming that vacation ex- per inner arm) was measured using a breaks each year. As Vancouver is a rela- posure took place during peak, day- reflectance spectrophotometer set to tively low-sunlight area and records light UV-B exposure hours. Total UV 680 nm. Parents of each child com- high temperatures only in the sum- exposure from vacation and recre- pleted a detailed questionnaire, assess- mer, evaluating summer exposure plus ational activities in MEDs, adjusted for ing the child’s ethnic origin, sun sen- the other 2 school holiday periods each clothing worn while outdoors, was also sitivity, sunburn history, and holiday year captures most solar exposure in calculated by anatomic subsite. Be- sunlight exposure to the time of ran- children. cause it was not possible to directly domization. In May 1996, all children retained in measure whole-body exposure, MED Children were individually random- the study were reexamined by physi- values for each of the 4 anatomic sub- ized by the study statistician (A.J.C.) to cians, and their nevi were enumerated sites were simply summed in each sub- the sunscreen (intervention) or the am- once again. Physician-counters did not ject to get the whole-body score. bient use (control) group. The statis- know to which study group children Whole-body nevus counts from 1993 tician had no contact with the physi- had been assigned. To ensure that ne- were subtracted from 1996 counts for cians counting nevi or with the study vus counts were concordant among each child, giving the number of new subjects. Parents of those randomized counters, 69 (15%) of the students were nevi. All nevi regardless of size were in-

2956 JAMA, June 14, 2000—Vol 283, No. 22 (Reprinted) ©2000 American Medical Association. All rights reserved. SUNSCREEN USE AND DEVELOPMENT OF NEW NEVI cluded in the counts. Comparisons be- enrolled in the study and randomized to ous studies, as shown in TABLE 1. Skin tween study groups were based on me- either the sunscreen or control group. At reflectance value at 680 nm did not dians, and differences were assessed the completion of the trial 3 years later, demonstrate a significant relationship using the Kruskal-Wallis test. 393 (86%) remained. The children in the with nevus frequency. A linear regression model to ac- study were largely white (323 [82%]); Analysis of the number of new nevi count for the number of new nevi was Chinese Canadian and other Asian Ca- revealed that children in the sunscreen fitted, using the following predictor nadian students (37 [9%]) made up the group developed significantly fewer new variables: treatment group, school grade second-largest ethnic group. The num- (equivalent to age), sex, skin reflec- ber of Asian Canadian and dark- Figure 1. Trial Profile tance value, facial freckling, hair color, skinned subjects was small, and, as they 20 skin reaction to sunlight, family his- acquire few new nevi with age and are 696 Schoolchildren Eligible tory of skin cancer, sunburn history to at low risk of eventual cutaneous mela- 354 in Grade 1 age 5 years, sunburn history during noma,24 they were eliminated from con- 342 In Grade 4 1993 through 1996, hours spent out- sideration prior to beginning the analy- doors during 1993 through 1996, va- sis. Six grade 1 and 8 grade 4 students 458 Enrolled cation sun exposure during 1993 with missing nevus-counter identifica- through 1996, and total sun exposure tion were excluded, leaving 309 white 458 Randomized during 1993 through 1996 adjusted for children for the final analysis (FIGURE 1). clothing. The variables sex, grade, skin The median nevus counts at intake 222 Received Sunscreen 236 Did Not Receive reaction to sunlight, treatment group, were 41 for grade 1 students (aged 6-7 (Intervention) Sunscreen (Control) 108 in Grade 1 117 in Grade 1 and hair color (dark brown, light years) and 68 for grade 4 children (aged 114 in Grade 4 119 in Grade 4 brown, blond, red) were modeled as 9-10 years). The distribution of nevi at categorical variables and all others as intake was skewed positively, with a few 0 Withdrew 0 Withdrew 39 Excluded (Nonwhite) 31 Excluded (Nonwhite) continuous variables. children having very high counts. No 32 Lost to Follow-up 33 Lost to Follow-up The baseline model included sex, child had a count of zero. 6 Missing Information 8 Missing Information grade, hair color, and treatment group. Factors such as hair color, skin re- Additional variables were added to the action to sunshine, facial freckling, and 145 Completed Trial and 164 Completed Trial and Data Analyzed Data Analyzed baseline model using a forward- sunburn score in the first 5 years of life 67 in Grade 1 87 in Grade 1 selection algorithm, with inclusion re- demonstrated associations with nevus 78 in Grade 4 77 in Grade 4 stricted to factors with a significance counts similar to those seen in previ- level of PϽ.10. An inspection of plot- ted data suggested potential interac- Table 1. Intake Nevus Count (1993), Pigmentation Characteristics, and Sunburn Score tions between treatment group and Among White Students other predictor variables. Conse- No. of Subjects quently, in the initial stages of multi- Median Sunscreen Control Nevus P variate analysis, variables were added Group Group Total Count Value* to the model as a combined main ef- Skin reflectance fect and interaction-with-treatment- Dark 48 56 104 55.5 group effect. Significance of these vari- Medium 47 56 103 53.0 .72 ables was assessed according to the P Light 50 52 102 50.5 value of the interaction effect rather Hair color Dark brown 46 58 104 59.0 than the main effect. Light brown 41 45 86 53.5 After including variables with sig- Ͻ.001 Red 10 10 20 27.0 nificant interaction effects, subse- Blond 48 51 99 51.0 quent modeling was performed to test Freckles on face for the significance of the remaining in- Few or none 52 62 114 42.0 dependent variables. Residual plots Moderate 36 53 89 53.0 Ͻ.001 were used to confirm the indepen- Dense 57 49 106 66.5 dence, normal distribution, and con- Sunburn score, first 5 years stant variance of the errors. Grade 1 Low 28 51 79 38.0 .07 RESULTS High 39 36 75 43.0 Grade 4 A total of 696 children (354 in grade 1 Low 39 41 80 62.0 .02 and 342 in grade 4) were ascertained in High 39 36 75 78.0 the 6 schools. Of these, 458 (66%) were *Kruskal-Wallis test for difference in medians.

©2000 American Medical Association. All rights reserved. (Reprinted) JAMA, June 14, 2000—Vol 283, No. 22 2957 SUNSCREEN USE AND DEVELOPMENT OF NEW NEVI nevi than those in the control group site. The majority of study subjects in regression lines for subjects in the 2 (median counts, 24.0 vs 28.0; P=.048). both the sunscreen and control groups study groups. A comparison using mean values reported zero episodes of trunk expo- To further assess the difference in showed an even greater difference (28.8 sure unprotected by sunscreen, creat- number of new nevi between subjects vs 34.6). A few children had a lower ne- ing artificially low medians. Mean val- with and without facial freckles, models vus count in 1996 than in 1993. ues provide more credible estimates and were constructed separately for grade 1 TABLE 2 compares measures of show an excess of unprotected epi- children and grade 4 children, with sub- sunlight exposure in the 2 treatment sodes in the control group compared jects dichotomized into 2 groups: those groups. Time spent outdoors from with the sunscreen group (7.7 vs 5.2). with Յ10% freckling (no freckles), and 1993-1996 was very similar among the A model of the effects of the inde- those with Ͼ10% freckling density students in each study group. No dif- pendent variables on the whole-body (freckles). The model predicted that ference in vacation solar exposure from number of new nevi is presented in grade 1 children who had freckles would 1993-1996 in MEDs was seen be- TABLE 4. Total sunlight exposure, ad- have about 40% fewer new nevi after 3 tween the sunscreen and control justed for clothing, school grade (age), years when randomized to the sun- groups, and no major difference was the interaction term for sunscreen screen group rather than the control seen in total sunlight exposure ad- group, and degree of facial freckling, ap- group. Grade 4 children with freckles ran- justed for clothing coverage for whole- pears to predict nevus counts. The in- domized to the sunscreen group would body or anatomic subsite (Table 2). teraction between being randomized to have about 30% fewer new nevi than if Use of sunscreen was assessed by the broad-spectrum sunscreen group they were randomized to the control anatomic subsite (TABLE 3). When sun- and degree of freckling is statistically group. Children with no freckling in screen was used, exposure was defined the strongest predictor of new nevi. grades 1 and 4 would have little advan- as protected and when not used, as un- FIGURE 2 shows that the importance of tage when randomized to the sunscreen protected. Median number of episodes being randomized to the sunscreen group compared with the control group. of protected and unprotected exposure group increases with increasing de- Finally, if sunscreen attenuates the during the observation period showed gree of freckling. Removing subjects development of new nevi, it might be a greater proportion of unprotected epi- with the greatest number of new nevi expected that, after control for freck- sodes in the control group at each sub- had little effect on the divergence of the ling, subjects in the intervention group who used the most sunscreen would have the fewest new nevi. FIGURE 3 Table 2. Solar Exposure Variables by Randomization Group* (grade 4 children presented; grade 1 Sunscreen Group Control Group graph similar) also demonstrates that this is the case, and, although the dif- No. of No. of UV Variable Subjects Median Subjects Median ferences are not statistically signifi- Time spent outdoors, 1993-1996, h 144 357.0 154 361.5 cant, there is an inverse relationship be- Vacation solar exposure, 1993-1996, MEDs† 131 962.5 140 962.5 tween sunscreen use and new nevi. Total sunlight‡ exposure adjusted for clothing coverage, 1993-1996, MEDs 131 142 Face, neck, and ears 421.7 420.8 COMMENT Trunk 148.2 149.0 To our knowledge, this is the first ran- Upper limbs 391.4 391.5 domized trial of the use of sunscreen as Lower limbs 275.0 276.0 a chemopreventive agent for attenuat- Whole body 1252.2 1214.3 ing nevi in children. Strengths of the *Minimal erythemal dose (MED) information from the tables of Diffey and Elwood.23 †Calculated from number of weeks of holiday, latitude, and number of clear sky MEDs per day. study include individual rather than ‡Based on reported outdoor activity each month, assumed to occur between noon and 1 PM. Clothing adjustment based on type of activity and clothing preference of each subject. Whole body index composed of values for the 4 group randomization and blinding of the anatomic sites summed for each subject. nevus counters to the status of the chil-

Table 3. Use of Sunscreen During Episodes of Summer Recreational Activity Sunscreen Group Control Group

Median No. Median No. of Median No. of Median No. Median No. of Median No. of No. of of Protected Unprotected Total Sun No. of of Protected Unprotected Total Sun Anatomic Site Subjects Sun Episodes Sun Episodes Episodes Subjects Sun Episodes Sun Episodes Episodes Face, neck, and ears 141 186.0 46.0 236.0 159 161.0 61.0 238.0 Trunk 141 208.0 0.0 216.0 159 207.0 1.0 216.0 Upper limbs 141 176.0 54.0 236.0 160 142.0 71.0 236.0 Lower limbs 141 108.0 41.5 153.0 160 95.8 53.0 145.5

2958 JAMA, June 14, 2000—Vol 283, No. 22 (Reprinted) ©2000 American Medical Association. All rights reserved. SUNSCREEN USE AND DEVELOPMENT OF NEW NEVI dren. The pattern of association be- short period of follow-up; there are no relationship28-30 between sunscreen use tween phenotypic factors and nevus clear data on the duration of solar expo- and melanoma risk. Furthermore, re- counts at induction gives reasonable as- sure needed to initiate nevus forma- sults from a recent cross-sectional study surance that the subjects are similar to tion. Limited data are available from the of nevi in European children suggest those recruited for previous studies of study by Harrison et al12 demonstrating that sunscreen is not effective at pre- nevi.13,14,16 Subject retention was excel- the presence of new nevi within 1 year venting the appearance of new nevi.31 lent, and parents and children exhib- of birth in a cohort of Australian chil- The authors suggested that this is be- ited a high degree of compliance in pro- dren. Thus, the initiation period for new cause use of high-SPF sunscreen pro- viding data on solar UV exposure during nevi was thought to be short, and the motes extended duration of sun expo- the trial. Parents in the sunscreen group 3-year follow-up was anticipated to be sure.32 An increase in new nevi among provided high-quality information on long enough to see differences develop children using sunscreen regularly was the volume of sunscreen used. between the 2 intervention groups. also seen in a German study.33 Both of No information was collected on chil- Another potential problem is the pos- the European studies were well con- dren who elected not to participate in the sibility that randomization to the sun- ducted but assessed sun exposure and trial, and it is possible that there may be screen group sensitized parents of these sunscreen use retrospectively, in some some differences between participants children. If this were the case, the chil- cases 3 to 5 years afterward. In addi- and nonparticipants. Another potential dren’s parents might have been at- tion, it was not clear whether the sun- drawback to the study is the relatively tuned to the potential benefits of sun screen used in the German study was avoidance and might have restricted so- broad-spectrum and attenuated UV-A Table 4. Parameter Estimates for Variables lar exposure in the children. How- and UV-B or was primarily designed to Predicting Number of New Nevi ever, reported hours spent outdoors screen out UV-B radiation. Finally, as in Vancouver Schoolchildren* during the 3 years of the study were in any retrospective investigation of Independent Estimate P similar in the 2 groups. Finally, within nevi, incomplete control of host- Variable (SE)* Value the intervention group, there was an in- susceptibility factors must be consid- Sunscreen group −0.89 (4.07) .83 Grade 6.82 (3.21) .04 verse relationship between quantity of ered a potential explanation for these Sex −1.51 (3.21) .64 sunscreen used and number of new findings. Freckling 0.12 (0.13) .34 Sunscreen group–freckling nevi, suggesting that sunscreen use was In our study, the reduced number of interaction −0.38 (0.17) .03 the factor of consequence in the study. new nevi in the broad-spectrum sun- Total sunlight exposure adjusted for clothing, A recent study in women has dem- screen group suggests that these agents 1993-1996, MEDs 0.007 (0.003) .05 onstrated a protective effect of chemi- may be useful in preventing transfor- Sunburn score to age 25 5 years 0.30 (0.17) .09 cal sunscreen against CMM. Most pre- mation of normal melanocytes into *Estimates generated using linear regression model vious studies, however, have suggested nevi, at least in children who freckle. controlling for grade, sex, and hair color. either no association26,27 or a positive As 1993 and 1996 counts were both

Figure 2. Appearance of New Nevi Among Children by Facial Figure 3. Appearance of New Nevi in Grade 4 Children by Amount Freckling of Sunscreen Used

200 100 Sunscreen Group Control Group 80 150

100 40

20 50 Control Group Sunscreen Group 0 No. of New Nevi, 1993-1996 No. of New Nevi, 1993-1996

0 –20

–50 –40 0 1020304050 60 70 80 90 100 0 100 200 300 400500 600 700 Freckles on Face, % Coverage Sunscreen Use, mL

©2000 American Medical Association. All rights reserved. (Reprinted) JAMA, June 14, 2000—Vol 283, No. 22 2959 SUNSCREEN USE AND DEVELOPMENT OF NEW NEVI conducted prior to the sunny months smaller degree of solar UV damage in In summary, our findings indicate of the year, it is unlikely that the effect children who freckle than in those who that broad-spectrum sunscreens may seen in freckled children was due to dif- do not freckle. If sunscreens protect the attenuate the development of nevi in ferential misclassification of freckles and melanocytes, the importance of such children and perhaps ultimately nevi between the 1993 and 1996 counts. protection against the formation of new reduce their risk of developing mela- Both nevi1-3,34 and freckling35-37 are nevi would be more important among noma. known to increase risk of CMM, and those who freckle. evidence suggests that nevi and freck- An alternative explanation for the Funding/Support: This study was supported by the Brit- ling together have a synergistic effect greater importance of being in the sun- ish Columbia Health Research Foundation grant 161(91- 3,6 1), by the Medical Research Council of Canada, grant on risk. This may indicate that sub- screen group for those who freckle may MA-13255, and by Health Canada through the Action jects who freckle and develop nevi have be found in the limited duration of the Plan for Health and the Environment. Mr Bajdik is a re- an underlying instability in their me- intervention. It is possible that the study search student of the National Cancer Institute of Canada, supported with funds provided by the Terry Fox Run. lanocytes. If so, these melanocytes saw a short-term result only in those Sunscreen (PreSun Ultra 30) for the trial was provided might be more likely to evolve into a subjects most sensitive to the develop- by Westwood-Squibb Pharmaceuticals, Buffalo, NY. Disclaimer: The authors have no financial relation- clone that ultimately becomes a vis- ment of nevi. In a trial of longer dura- ship with any of the funding agencies for the study ible nevus under the influence of solar tion, a more clear-cut protective effect or with Westwood-Squibb Pharmaceuticals. Acknowledgment: We thank Sharon Gee, BSc, for co- UV radiation. Furthermore, forma- might be seen for all subjects, regard- ordinating the study and Heather Neidig for admin- tion of a nevus may take place with a less of freckling status. istrative support.

REFERENCES 1. Holman CD, Armstrong BK. Pigmentary traits, eth- ley RP, Peck IM. Benign pigmented nevi in children. ship of cutaneous malignant melanoma to individual nic origin, benign nevi, and family history as risk fac- Arch Dermatol. 1992;128:1201-1206. sunlight-exposure habits. J Natl Cancer Inst. 1986; tors for cutaneous malignant melanoma. J Natl Can- 15. Coombs BD, Sharples KJ, Cooke KR, Skegg DC, 76:403-414. cer Inst. 1984;72:257-266. Elwood JM. Variation and covariates of the number 27. Elwood JM, Gallagher RP. More about: sun- 2. Holly EA, Kelly JW, Shpall SN, Chiu SH. Number of of benign nevi in adolescents. Am J Epidemiol. 1992; screen use, wearing clothes, and number of nevi in melanocytic nevi as a major risk factor for malignant 136:344-355. 6- to 7-year-old European children [letter]. J Natl Can- melanoma. J Am Acad Dermatol. 1987;17:459-468. 16. English DR, Armstrong BK. Melanocytic nevi in cer Inst. 1999;91:1164-1166. 3. Marrett LD, King WD, Walter SD, From L. Use of children, I: anatomic sites and demographic and host 28. Autier P, Dore JF, Schifflers E, et al, for the EORTC host factors to identify people at high risk for cuta- factors. Am J Epidemiol. 1994;139:390-401. Melanoma Cooperative Group. Melanoma and use of neous malignant melanoma [published correction ap- 17. Newton-Bishop JA, Bataille V, Pinney E, Bishop sunscreens: an EORTC case-control study in Ger- pears in CMAJ. 1992;147:1764]. CMAJ. 1992;147: DT. Family studies in melanoma: identification of the many, Belgium, and France. Int J Cancer. 1995;61: 445-453. atypical mole syndrome (AMS) phenotype. Mela- 749-755. 4. Grob JJ, Gouvernet J, Aymar D, et al. Count of be- noma Res. 1994;4:199-206. 29. Westerdahl J, Olsson H, Masback A, Ingvar C, Jons- nign melanocytic nevi as a major indicator of risk for 18. MacKie RM, McHenry P, Hole D. Accelerated de- son N. Is the use of sunscreens a risk factor for malig- nonfamilial nodular and superficial spreading mela- tection with prospective surveillance for cutaneous ma- nant melanoma? Melanoma Res. 1995;5:59-65. noma. Cancer. 1990;66:387-395. lignant melanoma in high-risk groups. Lancet. 1993; 30. Beitner H, Norell SE, Ringborg U, Wennersten G, 5. Green A, MacLennan R, Siskind V. Common ac- 341:1618-1620. Mattson B. Malignant melanoma: aetiological impor- quired naevi and the risk of malignant melanoma. Int 19. Kelly JW, Rivers JK, MacLennan R, Harrison S, tance of individual pigmentation and sun exposure. J Cancer. 1985;35:297-300. Lewis AE, Tate BJ. Sunlight: a major factor associated Br J Dermatol. 1990;122:43-51. 6. Osterlind A, Tucker MA, Hou-Jensen K, Stone BJ, with the development of melanocytic nevi in Austra- 31. Autier P, Dore JF, Cattaruzza MS, et al, for the Engholm G, Jensen OM. The Danish case-control study lian schoolchildren. J Am Acad Dermatol. 1994;30: European Organization for Research and Treatment of cutaneous malignant melanoma, I: importance of 40-48. of Cancer Melanoma Cooperative Group. Sunscreen host factors. Int J Cancer. 1988;42:200-206. 20. Gallagher RP, Rivers JK, Yang CP, McLean DI, use, wearing clothes, and number of nevi in 6- to 7. Skender-Kalnenas TM, English DR, Heenan PJ. Be- Coldman AJ, Silver HK. Melanocytic nevus density in 7-year-old European children. J Natl Cancer Inst. 1998; nign melanocytic lesions: risk markers or precursors Asian, Indo-Pakistani, and white children: The Van- 90:1873-1880. of cutaneous melanoma? J Am Acad Dermatol. 1995; couver Mole Study. J Am Acad Dermatol. 1991;25: 32. Autier P, Dore JF, Negrier S, et al. Sunscreen use 33:1000-1007. 507-512. and duration of sun exposure: a double-blind, ran- 8. Kruger S, Garbe C, Buttner P, Stadler R, Guggen- 21. English DR, MacLennan R, Rivers JK, Kelly J, Arm- domized trial. J Natl Cancer Inst. 1999;91:1304- moos-Holzmann I, Orfanos CE. Epidemiologic evi- strong BK. Epidemiological studies of melanocytic naevi: 1309. dence for the role of melanocytic nevi as risk markers protocol for identifying and recording naevi. IARC In- 33. Luther H, Altmeyer P, Garbe C, et al. Increase of and direct precursors of cutaneous malignant mela- ternal Report No. 90/002. Lyon, France: Interna- melanocytic nevus counts in children during 5 years noma. J Am Acad Dermatol. 1992;26:920-926. tional Agency for Research on Cancer; 1990. of follow-up and analysis of associated factors. Arch 9. Weinstock MA, Colditz GA, Willett WC, et al. Moles 22. Mosteller RD. Simplified calculation of body sur- Dermatol. 1996;132:1473-1478. and site-specific risk of nonfamilial cutaneous malig- face area [letter]. N Engl J Med. 1987;317:1098. 34. White E, Kirkpatrick CS, Lee JA. Case-control study nant melanoma in women. J Natl Cancer Inst. 1989; 23. Diffey BL, Elwood JM. Tables of ambient solar ul- of malignant melanoma in Washington State, I: con- 81:948-952. traviolet radiation for use in epidemiological studies stitutional factors and sun exposure. Am J Epide- 10. Rivers JK, Frederiksen PC, Dibdin C. A preva- of malignant melanoma. In: Gallagher RP, Elwood JM, miol. 1994;139:857-868. lence survey of dermatoses in the Australian neo- eds. Epidemiological Aspects of Cutaneous Malig- 35. Elwood JM, Gallagher RP, Hill GB, Spinelli JJ, Pear- nate. J Am Acad Dermatol. 1990;23:77-81. nant Melanoma. Boston, Mass: Kluwer Academic Pub- son JC, Threlfall W. Pigmentation and skin reaction 11. Kroon S, Clemmensen OJ, Hastrup N. Incidence lishers; 1994:81-105. to sun as risk factors for cutaneous melanoma: West- of congenital melanocytic nevi in newborn babies in 24. Parkin DM, Whelan SL, Ferlay J, Raymond L, ern Canada Melanoma Study. BMJ (Clin Res Ed). 1984; Denmark. J Am Acad Dermatol. 1987;17:422-426. Young J. Cancer incidence in five continents. Vol 7. 288:99-102. 12. Harrison SL, MacLennan R, Speare R, Wronski I. IARC Scientific Publication No. 143. Lyon, France: In- 36. Holly EA, Aston DA, Cress RD, Ahn DK, Kris- Sun exposure and melanocytic naevi in young Aus- ternational Agency for Research on Cancer; 1997. tiansen JJ. Cutaneous melanoma in women, II: phe- tralian children. Lancet. 1994;344:1529-1532. 25. Holly EA, Aston DA, Cress RD, Ahn DK, Kris- notypic characteristics and other host-related fac- 13. Gallagher RP, McLean DI, Yang CP, et al. Sun- tiansen JJ. Cutaneous melanoma in women, I: expo- tors. Am J Epidemiol. 1995;141:934-942. tan, sunburn, and pigmentation factors and the fre- sure to sunlight, ability to tan, and other risk factors 37. Chen YT, Dubrow R, Holford TR, et al. Malig- quency of acquired melanocytic nevi in children. Arch related to ultraviolet light. Am J Epidemiol. 1995;141: nant melanoma risk factors by anatomic site: a case- Dermatol. 1990;126:770-776. 923-933. control study and polychotomous logistic regression 14. Pope DJ, Sorahan T, Marsden JR, Ball PM, Grim- 26. Holman CD, Armstrong BK, Heenan PJ. Relation- analysis. Int J Cancer. 1996;67:636-643.