Interventional study in 1,232 young German children to prevent the development of melanocytic nevi failed to change sun exposure and sun protective behavior

Authors


Abstract

Sunscreens have been proposed as protective measures to inhibit the development of melanocytic nevi in childhood and to decrease the long term risk for cutaneous melanoma development. Our present study investigates the influence of sunscreen use and education on the number of incident melanocytic nevi. A total of 1,812 children in 78 public nursery schools in 2 German cities were randomized to 3 study arms: (i) parents were informed on study purpose and sun protection measures only at an initial educational meeting; (ii) parents received educational material 3 times yearly; and (iii) education and 800 ml free broad spectrum sunscreens with sun protection factor 25 provided on a yearly basis. Final assessment after 3 years follow-up included 1,232 children (68%). Changes of sun protection habits including sunscreen use were sparse, without any differences attributable to the intervention efforts. As a consequence, there were no significant differences between the 3 study arms for the main outcome measure, the number of incident melanocytic nevi. Analysis of the sunscreen use in the entire cohort irrespective of our study arms did not show any impact on incident nevus numbers in bivariate or multivariate analysis. In conclusion, intervention with educational letters and free sunscreens seemingly had no additional effect on sun-protection for German children. High prevalence of sunscreen use at study commencement, social desirability, and inadequate application of sunscreens might have partially covered their effect. © 2005 Wiley-Liss, Inc.

Exposure to ultraviolet light in childhood as well as in adulthood is the central environmental risk factor for cutaneous melanoma (CM).1, 2, 3, 4 The use of sunscreen is intended to avoid sunburn as well as to reduce the risk of CM. Numerous case-control studies focusing on the effect of sunscreen use and the risk of CM reported contradicting results reaching from a reduced risk,5, 6, 7, 8 an unchanged risk,9 to an increased risk in persons using sunscreen.10, 11, 12, 13, 14, 15, 16, 17 Most of these case-control studies were hospital-based, might not have been controlled comprehensively for exposure to sunlight and skin type and were most likely influenced by recall bias, as they involved melanoma patients. Meta-analyses of studies published recently on the relationship of sunscreen use and the risk of CM could not find any association.18, 19, 20, 21

The number of melanocytic nevi (MN) is known to be the strongest risk factor for CM. Nevi are precursor lesions of a substantial proportion of 20–60% of CM.22 Humans develop MN from an early age onward and numerous studies have investigated the development of MN in children.1, 23, 24, 25, 26, 27, 28, 29, 30 Studying sunscreen use in children and its relationship to MN should be less influenced by recall bias. Thus far, however, investigations of sun protective behavior in children have not shown conclusive evidence for either a protective or harmful effect of sunscreen.31, 32, 33, 34 The dynamic development of MN in children allows detection of risk factors and effects of sun protection within a short study period of 3 years.24, 33

To our knowledge, this is the second study of a randomized controlled trial investigating the effect of an intervention with education or education plus free sunscreens on the development of melanocytic nevi in young children. The aim of our study was to ascertain if children receiving education or education and free sunscreen would develop significantly less incident MN during 3 years of follow-up when compared to a control group.

Material and methods

Primary hypothesis

The primary hypothesis to be tested was if children receiving education or education and free sunscreen would develop significantly less incident MN during 3 years of follow-up when compared to a control group. Secondary hypotheses to be addressed were if significantly reduced levels of sun exposure and improved sun protection habits could be achieved in the intervention groups compared to the control group.

Participants

From March to October 1998, 1,887 German children between 2–7 years of age were included in the present randomized controlled trial on the prevention of MN (Fig. 1). Children were recruited from 49 public nursery schools in Stuttgart (latitude: 48° 46 min), and 29 public nursery schools in Bochum, (51° 29 min), Germany, which were selected randomly from a total of 242 public nursery schools in Stuttgart and 169 in Bochum in different suburbs of both cities. Only 3 public nursery schools that had been approached refused participation. In the 78 participating public nursery schools >80% of the parents agreed to enter our study. Children 2–7 years of age with skin Type I–IV according to Fitzpatrick35 were eligible.

Figure 1.

Flow-chart of progress of children through the randomized controlled trial. *Reasons for children not to be randomized: 3 public nursery schools declined participation (n = 81), no parental consent (n = 436), parents did not attend educational session or child ill or on holidays (n = 36). **Reasons for children to be excluded: no physical examination available, skin Type V or VI, missing information for age.

Exclusion criteria

Children of parents who did not attend to the first educational session, who did not provide consent to our study, or children who were ill or on holidays at the time of the baseline examination could not participate in our study. Children with skin Type V or VI35 (non-Caucasian; n = 40), with missing age information (n = 4), with immunosuppression after organ transplantation or due to chemotherapy of cancer or who did not allow physical examination (n = 31) had to be excluded.

Ethical approval

Ethical clearance was given by the Ethical Committees of the Faculty of Medicine, University of Tübingen and Bochum. As proposed by the ethical committees, at commencement of our study all parents including the control group received comprehensive information on the risks of sun exposure and sunburn, sun protective measures, sunscreen use (proper application and reapplication), and the aim and design of our study. This educational session took 3 hr. Parents provided written consent to our study and children gave oral assent.

Baseline assessment

Initial assessment was carried out several weeks after the first educational session. Parents answered a questionnaire and physical examination of all participating children was carried out. Parental interview comprised questions about sun exposure of their child playing at home, duration and destination of holiday, history of sunburns, sunscreen use and education and ethnicity of parents. Physical examination was conducted by 2 dermatologists. Evaluation of the reliability of MN counts in 49 Australian children showed a very high concordance between one of the authors (J.B.) and an expert in MN in children who did not participate in the present study (Simone Harrison, Townsville, Australia). The concordance correlation coefficient36 was 0.97 (95% confidence interval [CI] = 0.95, 0.98) (unpublished data). Children were examined wearing only underpants in a well-illuminated room. A standard protocol was used to define and count MN.37 Melanocytic nevi were defined as brown to black pigmented macules or papules of any size, darker in the color than the surrounding skin. Small MN were distinguished from freckles by their darker color and if present by their elevated appearance. In all cases of doubt, pigmented lesions were not counted. No attempt was made to differentiate lentigo simplex from junctional MN. Halo nevi, nevi spili, congenital nevi and blue nevi were counted separately and were included in the total count of MN. The severity and distribution of freckling was recorded separately.

Intervention procedures

After parents provided consent for our study, children were randomized into 3 study arms: (i) control group: after the initial educational session no more information or educational sessions were provided; (ii) educational group: parents received an educational letter 3 times yearly (Easter, Pentecost, and the summer holidays) with more detailed information on proper sunscreen use and sun protection than the educational session provided at study commencement; they also received information brochures from public melanoma prevention campaigns with detailed information; and (iii) education and sunscreen group: parents received educational material like the educational group and additionally 800 ml of free broad-spectrum sunscreen with sun protection factor 25 yearly. Parents were instructed to apply sunscreens during the times of intensive solar radiation from Spring to Autumn on the sun exposed body parts, i.e., on face and arms of the child when wearing clothes and on the whole body whenever at the beach or outdoor swimming pool. They were instructed to closely cover the skin with a layer of sunscreen several times a day. Parents were asked to buy more sunscreen if they had used up their free sunscreen.

Randomization process

For feasibility reasons and to avoid interaction between the groups, children were allocated into one of two intervention groups or the control group by randomizing the entire respective day-care centre. Randomization was conducted by using random allocation from a computer program.

Follow-up and final assessment

Baseline assessment was conducted in Stuttgart in Summer 1998 and in Bochum in Autumn 1998. Final assessment took place in Stuttgart in Summer 2001 and in Bochum in Autumn 2001. The follow-up interval was 3 years on average. The final examiner was not aware of the randomization and thus blinded. The second and final parental interview and physical examination of the child were conducted in the same way as at the initial assessment.

Outcome variable

The number of newly developing (incident) MN as observed during follow-up time was the outcome variable.

Influencing characteristics

The following variables were considered as potential independent predictors of the outcome variable: (i) intervention; (ii) basic demographic and pigmentary characteristics of the child, such as age, gender, skin color, eye color, hair color, skin type according to Fitzpatrick, and freckling; (iii) parental features, such as ethnicity, child's family history of melanoma, number of MN on arms of mother and father, and education; (iv) sun protection of the child at baseline, that is previous use of sunscreen and clothing habits at the beach or outdoor swimming pool; and (v) sun exposure of child as assessed by history of holidays in sunny climates, activity score at home, and number and severity of sunburns. Newly experienced sunburns refer to sunburns during the 3 years of follow-up and previous sunburns to the time before. To further quantify the reported holidays a score of the countries where holidays were spent from 1998–2001 was created. The countries were recoded into: “0”, “no holiday”; “1”, holiday in “Northern Europe or USA”; “2”, holiday in “Northern Mediterranean”; “3”, holiday in “Countries in Northern Africa”; and “4”, holiday in “Tropics.” The 4 variables relating to the years 1998–2001 were added up, creating a score ranging from 0–16. For the sun exposure at home a home activity score was calculated considering the following activities: playing ball, sun bathing, swimming outdoors, playing outdoors, walking, riding the bike, and being outdoors in general. Each activity scored “0” for less than once per week and “1” for once per week or more often. These scores were added up to gain an overall activity score at home ranging from 0–7. Further details of the variables considered have been published previously.27

Statistical analysis

Depending on the distribution, numerical variables were described by mean values and standard deviations (±SD), or median values and interquartile ranges (IQR). Bivariate comparisons between control and intervention groups of potentially influencing characteristics, of outcome variables at baseline, and with respect to the number of incident MN, to the changes in sun exposure characteristics, and to sun protection habits between 1998–2001 were conducted using χ2 tests, analysis of variance, and non-parametric Kruskal-Wallis tests as appropriate. Non-parametric Wilcoxon tests, χ2 test statistics, and Fisher's exact test were used for comparisons of 2 study arms at a time. Statistical analysis was conducted using SPSS for Windows. Throughout the analysis a significance level of 0.05 was assumed.

Multivariate analysis of the intervention

Multivariate linear regression analysis was used to assess the impact of the interventions on the number of incident melanocytic nevi adjusted for confounding variables. All influencing characteristics assessed during our study were considered as potential confounding variables. In preparation for multivariate linear regression analysis, the outcome variable was log-transformed to acquire an approximately normally distributed variable and all categorical variables were dummy-coded. The multivariate linear regression model included: age, gender, skin type according to Fitzpatrick, hair color, freckling in face, parental level of education, ethnicity of parents, number of moles on mothers' arms, number of moles on fathers' arms, score of countries where holidays were spent between 1998–2001, activity score at home, history of sunburn and extent and severity of sunburns.

Analysis of the effect of sun protection over the entire cohort

As the intervention showed neither an impact on sun protective behavior and sun exposure of the participants nor on the number of incident melanocytic nevi, we additionally analyzed the effect of sun protection on the number of incident melanocytic nevi over the entire cohort. For this purpose, the incident numbers of MN were dichotomized into below or equal vs. above the median number of MN for each year of age separately. Within a multivariate logistic model (adjusted for age, gender, skin type according to Fitzpatrick, hair color, freckling of the child, parental characteristics and sun exposure) sun-protective behaviors were assessed.

Results

Description of trial profile

At commencement of our study, 2,440 children in 81 public nursery schools were eligible. Three public nursery schools (81 children) declined participation. Parents of 436 children did not give consent to our study. Thirty-six children were ill or on holidays at baseline examination or their parents did not attend the educational session. A total of 1,887 children remained for randomization. Twenty-seven public nursery schools with 637 children were randomized to the control group, 26 public nursery schools with 624 children to the educational group, and 25 public nursery schools with 626 children to the education and free sunscreens group. A total of 75 children had to be excluded from the analysis because they had either no physical examination, a non-Caucasian skin type or information about age was missing, leaving 1,812 children.

Description of children lost to follow-up

There were 580 (32.0% of 1,812) children lost to follow-up (Table I). Comparisons of the characteristics of children lost to follow-up with children with complete follow-up was based on the data collected during baseline assessment in 1998 (n = 1,812). The results showed that the children lost to follow-up were less likely to have a fair skin complexion (p < 0.0001) and had fewer MN (p = 0.0002) compared to children with complete follow-up. Loss to follow-up was different for the 3 intervention arms (p < 0.0001). The parents of the children lost to follow-up were on average less educated (p < 0.0001), less likely to be both German descent (p < 0.0001) and had fewer MN on their arms (p < 0.0001) compared to the parents of children with complete follow-up. The children lost to follow-up had spent fewer holidays in sunny climates (p < 0.0001), and had previously experienced less sunburns (p = 0.0086) compared to children with complete follow-up. The children lost to follow-up had used sunscreen previously less often (p < 0.0001) and were less likely to wear at least 2 pieces of protective clothing at the beach or at the swimming pool (p < 0.0001). In conclusion, children prone to develop more MN were more likely to complete follow-up whereas children lost to follow up had a tendency to be in the “low risk group.”

Table 1. Comparison of Baseline Characteristics in 1998 Between Children Lost to Follow-Upand Children with Complete Follow-UP
 OverallLost to follow-upComplete follow-upp-value
(n = 1,812)(n = 580)(n = 1,232)
Basic characteristics of children
 % Skin Type I10.1%6.9%11.6%<0.001
 % Blonde-fair hair color52.6%44.7%56.3%<0.001
 % No freckles in face79.0%85.7%75.9%<0.001
 Median number of nevi, [IQR]8 [4,13]7 [4,11]8 [5,14]=0.002
 % From town Stuttgart52.0%42.9%56.3%<0.001
 % Randomized in education and sunscreen group33.9%25.9%37.7%<0.001
Characteristics of parents and family
 % Mothers with university degree19.9%16.2%21.4%<0.001
 % Fathers with university degree35.4%25.8%39.5%<0.001
 % Both parents of German decent79.5%70.9%83.6%<0.001
 Median number of nevi on mothers' arms, [IQR]27 [10,50]20 [6,44]30 [14,52]<0.001
 Median number of nevi on fathers' arms, [IQR]20 [5,43]13 [2,35.5]21, [7,46]<0.001
History of sun exposure of children
 % On holidays in sunny climates70.7%62.9%74.4%<0.001
 Median weeks on holidays in sunny climates, [IQR]4 [0,8]3 [0,8]4 [0,8]<0.001
 Median hr spent in the sun per day during holidays &&num;131;in sunny climates, [IQR]3 [0,5]2 [0,4]3 [0,5]<0.001
 % Playing more than 4 hr per day outside at home &&num;131;during summer23.7%26.5%22.5%=0.015
 Median number of previous sunburns, [IQR]0 [0,1]0 [0,1]0, [0,2]=0.086
History of sun protection of children
 % Ever used sunscreen94.4%87.9%97.5%<0.001
 Median age when sunscreen was first used, [IQR]1 [1,1]1 [1,1.05]1 [1,1]=0.029
 % Almost always using sunscreen76.4%69.2%79.8%<0.001
 % Wearing swimming trunks at the beach94.9%91.7%96.4%<0.001
 % Wearing T-shirt at the beach73.0%67.6%75.5%=0.004
 % Wearing at last 2 of swimming trunks, T-shirt or &&num;131;shorts at the beach71.9%67.5%73.9%<0.001

Description of children with complete follow-up and results of comparison of intervention and control groups at baseline

Follow-up was completed for a total of 1,232 children from 78 public nursery schools (Table II). Age at baseline ranged from 2–7 years and 51.4% were boys. At baseline, the median number of MN was 8 (IQR = [5,14]) and 97.5% of parents answered that they had used sunscreen previously to protect their children. At baseline there were no statistically significant differences between intervention and control groups.

Table 2. Comparison of Baseline Characteristics between Control and Intervention Groups1
 Overall(n = 1,232)Control group(n = 398)Education group(n = 369)Education and sunscreen group (n = 465)p-value
  • 1

    SD, standard deviation; IQR, interquartile range.

Basic characteristics of children
 Mean age (±SD)4.3 (±1.1)4.3 (±1.1)4.3 (±1.1)4.4 (±1.1)=0.661
 % Boys51.4%51.0%50.7%52.3%=0.888
 % Born in Germany98.5%98.7%98.9%98.1%=0.547
 % Skin Type I11.6%11.1%10.3%13.1%=0.145
 % Blonde-fair hair color56.3%54.5%53.1%60.4%=0.067
 % No freckles in face75.9%77.9%75.6%74.4%=0.494
 Median no. of MN [IQR]8 [5,14]8 [4,14]8 [4,14]9 [6,14]=0.101
 % From the town Stuttgart56.3%52.3%59.3%57.4%=0.118
Characteristics of parents and family
 % Mothers with university degree21.4%20.3%21.8%22.1%=0.267
 % Fathers with university degree39.5%36.7%44.1%38.2%=0.319
 % Both parents of German decent83.6%85.4%81.6%83.6%=0.548
 Median number of nevi on mothers' arms [IQR]30 [14,52]30 [12,55]28.5 [12.75,49]30 [15,54]=0.462
 Median number of nevi on fathers' arms [IQR]21 [7,46]20 [5,45]24.5 [7,49]21 [8,44]=0.162
History of sun exposure of children
 % With history of holidays in sunny climates74.4%74.1%74.5%74.5%=0.989
 Median weeks on holidays in sunny climates [IQR]4 [0,8]4 [0,8.25]4 [0,8.5]4 [0,8]=0.626
 Median hr in the sun per day during holidays&&num;131;in sunny climates [IQR]3 [0,5]4 [0,5]4 [0,5]3 [0,5]=0.712
 Median score of outdoor activities at home&&num;131;during summer [IQR]5 [4,6]5 [4,5.25]5 [4,5.5]5 [4,6]=0.673
 % Playing more than 4 hr per day outside at home during summer22.5%24.7%22.6%20.5%=0.822
 Median number of previous sunburns [IQR]0 [0,2]0 [0,2]0 [0,2]0 [0,2]=0.797
History of sun protection of children
 % Ever used sunscreen97.5%97.2%96.7%98.3%=0.350
 Median age when sun-screen was first used [IQR]1 [1,1]1 [1,1]1 [0.8,1]1 [1,1]=0.882
 % Almost always using sunscreen79.8%79.6%80.4%79.3%=0.829
 % Wearing swimming trunks at the beach96.1%96.2%96.5%95.7%=0.833
 % Wearing T-shirt at the beach75.2%73.4%78.0%74.6%=0.298
 % Wearing shorts at the beach15.8%14.8%16.3%16.4%=0.795
 % Wearing swimming trunks, T-shirt and shorts&&num;131;at the beach13.9%13.1%14.6%14.0%=0.719

Incidence of MN during the 3 years of follow-up

The median number of new MN was 26 (IQR = [17,40]; range = 1–186) (Table III). There was no statistically significant difference between the 3 comparison groups with respect to incident MN(p = 0.7793). This non-significant result was confirmed by the multivariate linear regression analysis.

Table 3. Comparison of Incident Numbers of Melanocytic Nevi, Changes of Sun Exposure and of Sun Protection Habitsbetween 1998–2001 Between Control and Intervention Groups1
 Overall(n = 1,232)Control group(n = 398)Education group(n = 369)Education andsunscreen group(n = 465)p-value
  • 1

    IQR, interquartile range; SD, standard deviation.

  • 2

    Fisher's exact test.

Incident melanocytic nevi developed between 1998–2001
 Median [IQR]26 [17,40]27 [17,40]26 [16,41]27 [18,40]=0.779
Sun exposure 2001 and changes between 1998–2001
 Median weeks on holidays in sunny climates [IQR]5 [2,8]5 [2,8]6 [2,8]4 [2,7.5]=0.021
 Median score of country of holiday [IQR]4 [3,6]4 [3,6]4 [3,6]4 [3,6]=0.009
 Median difference in hr/day in sun during&&num;131;holidays in sunny climates (1998–2001) [IQR]0 [−1,1]0 [−1,1]0 [−1,1]0 [−1,1]=0.061
 Median difference in home activity&&num;131;score (1998–2001) [IQR]0 [−1,1]0 [−1,1]0 [−1,1]0 [−1,1]=0.836
 Mean difference of hr/day outside at home&&num;131;(1998–2001) (±SD)0.18 (±1.11)0.24 (±1.09)0.14 (±1.13)0.15 (±1.12)=0.353
Sunburn experience between 1998–2001
 % With sunburn experience22.2%23.2%21.5%22.0%=0.844
 Median number of newly experienced sunburns [IQR]0 [0,1]0 [0,1]0 [0,1]0 [0,1]=0.604
Habits of sun protection and changes between 1998–2001
 % Use of sunscreen since 199899.0%98.0%99.7%99.4%=0.0332
 % “Almost always” using sunscreen since 199885.6%83.1%84.8%88.4%=0.079
Changes in the use of sun protective clothing while on the beach or at a swimming pool between 1998–2001
 % Change to use T-shirts12.3%13.1%10.1%13.4%=0.532
 % Change to use shorts12.4%11.8%13.0%12.3%=0.989
 % Change to use trunks and T-shirt and shorts11.6%10.8%12.0%11.9%=0.981
 % Change to use hat7.7%7.0%7.3%8.7%=0.625

Changes of sun exposure and sun protection habits duringthe follow-up

Analysis of sun exposure and protection habits showed few significant differences between the groups after the intervention (Table III). Yet these changes did not follow a uniform pattern toward fewer holidays in sunny climates in the intervention groups and more sunscreen use in the education and free sunscreen group. The median number of weeks spent on holidays in sunny climates was overall significantly different between control and intervention groups (p = 0.0213), with most weeks on holidays in the educational group, followed by the control group, and lowest number in the educational and sunscreen group. The median score concerning the countries were participating families spent their holidays between 1998–2001 was significantly different between control and intervention groups (p = 0.0088). There was no significant difference between control and educational intervention group (p = 0.8732), whereas the educational and sunscreen intervention group was more likely to holiday in countries further away from the equator than the control group (p = 0.0064) and than the educational intervention group (p = 0.0122). The percentage of children reported to have used sunscreen since 1998 was different for control and intervention groups (p = 0.0325), although actual differences were minimal. The use of sunscreen since 1998 was higher in the educational intervention group compared to the control group (p = 0.0392), whereas the differences between control and the educational and sunscreen intervention group and between the 2 intervention groups were not statistically significant (p = 0.1246, p = 0.6340, respectively).

Effect of sun-protection on the number of incident MN analyzed over the entire cohort

Bivariate analysis showed that children who had equal or below the median number of incident MN for their age were more likely to usually wear swimming trunks, T-shirt and shorts at the beach or outdoor swimming pool (p = 0.003) (Table IV). All variables of sunscreen use were not significantly associated with incident melanocytic nevus number in bivariate analysis. Within the multivariate model (adjusted for age, gender, skin type, hair color, and freckling of the child, parental characteristics and sun exposure) sun protective behavior such as the use of sunscreen (p = 0.9851), the extent of the use of sunscreen on different body sites (p = 0.5236), as well as using sun protective clothing at the beach or outdoor swimming pool (swimming trunks, T-shirt and shorts, p = 0.156) failed to reach statistical significance.

Table 4. Effect of Sunscreen Use and Clothing on The Number of Incident MN Analyzed Over The Entire Cohort
 Overall (n = 1,232)≤ Median MN (n = 636)> Median MN (n = 596)p-value
  1. Results of comparisons of children with a number of incident melanocytic nevi (MN) less or equal the median for their year of age with children above the median.

Sun protection habits of children as assessed in 2001
 % Ever used sunscreen99.0%98.6%99.5%=0.103
 % Almost always using sunscreen85.6%84.7%86.6%=0.355
 Median score of extent of sunscreen use [IQR]10 [8,10]10 [8,10]10 [9,10]=0.003
 % Wearing swimming trunks at the beach97.2%98.0%96.5%=0.110
 % Wearing T-shirt at the beach73.1%71.4%74.9%=0.168
 % Wearing shorts at the beach19.4%21.7%16.9%=0.032
 % Wearing swimming trunks, T-shirt and shorts at the beach17.2%19.5%14.7%=0.003

Discussion

The present randomized controlled trial investigated the effect of education and free sunscreens on the number of MN in 1812 German children 2–7 years of age. After 3 years of follow-up time 1,232 children (68% of randomized children) were available for the final evaluation. For the main outcome measure, the number of incident MN, there was no significant difference between the 3 study arms (control group, educational group and educational and free sunscreen group). This result might be explained by the fact that sunscreen use was common before the intervention and increased for all 3 groups without significant differences and the intervention failed to improve sun protective behavior or reduce sun exposure. Analyzing the longitudinal data of the entire cohort irrespective of study arms the multivariate model showed neither a protective nor a harmful effect for children using more sunscreen or clothing.

Sunscreens are widely used to protect children from harmful effects of sunlight,38, 39, 40 and marketing suggests a “safe suntan.”41 The most important problem analyzing the protective or harmful effects of sunscreen use is the positive correlation identified previously of higher sun protection factors with increased sun exposure42, 43, 44 as well as with sunburn.39 Studies investigating the effect of sunscreen are potentially confounded by sun exposure. Moreover such studies are influenced by the inappropriate application of sunscreen.44, 45, 46, 47, 48, 49, 50 Conflicting results have been reported on the impact of sunscreens on melanocytic nevus counts in children. Sunscreen use was found to be associated with higher nevus counts in cross-sectional studies from Europe,31, 51, 52 Israel32 and Australia,53 and in a longitudinal study from Australia.30 An interventional study in 309 schoolchildren from Vancouver found a protective effect for sun-protection factor 30 sunscreens, especially in freckled children.33 The inconsistent results of these studies might be explained in part by undetected confounding in non-randomized studies30 and randomized studies should give better control of potential confounding.

The major limitation of the present randomized controlled trial was the fact that intervention could not influence the reported levels of sun exposure and sun protection habits toward a consistent pattern with reduced sun exposure and more sunscreen use in the intervention groups. This might be explained by the fact that the unexpected high prevalence of previous sunscreen use (98% of all children) and of children almost always using sunscreen when in the sun (79%) impeded any major improvement due to the intervention. Additionally the one single educational session conducted with parents of all groups at study initiation as required by the ethical committee might have further reduced the effect of the intervention measures. Parents had to be instructed about the aims of our study and how to protect their children from the harmful effects of sunlight by seeking shade, avoiding highest ultraviolet radiation intensity about noon, proper application and reapplication of sunscreens and avoidance of sunburns. Intervention with educational letters and provision of free sunscreens seemingly had no additional effect on sun-protection in the highly motivated study population.

Actual levels of sunscreen use and sunburns as well as data provided in the questionnaire might have been influenced by social desirability covering the true effect of the intervention. Similarly an Australian interventional study aiming to create sun-smart schools found that an intervention mailing information material was ineffective to increase sun-protective behavior.54 The results of our present study might have been influenced by loss to follow-up of children. Children prone to develop more melanocytic nevi, i.e., children with fair skin, more nevi at baseline assessment and with parents with higher nevus counts on their arms, were more likely to complete follow-up. This selection most likely reflects their parents' higher awareness of the negative implications of sun exposure for their children. It can be argued that because of the occurring selection the 3 study groups became overall more homogenous, in particular also with respect to their sun protection practices. It is most likely that our study is underestimating the effect the intervention might have on the development of MN.

Because the intervention showed no effect on the number of incident MN and failed to influence sun exposure and sun protection we analyzed the sunscreen use over the entire cohort irrespective of our study arms. This analysis could not demonstrate any significant protective or harmful effect of sunscreens on MN counts in the cross-section of the first assessment55 as well as in the multivariate analysis of the longitudinal data. Because our present study was thoroughly controlled for confounders, this might indicate that sunscreens did not effectively prevent development of MN. In performing this analysis, however, the advantage of randomization was lost. As noted previously, parents might have overstated sunscreen use due to social desirability. This information bias was uncontrolled in our present study but is likely to be non-differential and thus toward the null, covering the true effect of sunscreen. Inadequate application of sunscreens might have reduced the efficacy of sunscreens, which seems to be difficult to control in large epidemiological studies. Future studies should try to get more objective data on the actual sunscreen use (e.g., by weighing used sunscreen bottles). The number and weight of brought back bottles might be influenced by social desirability. The scoring system used to quantify holiday sun exposure in a simplified way might have introduced a certain bias to the analysis because it was based on an arbitrary classification according to the countries where holidays were made instead of using latitudes. This score could be much too distorted as compared to the reality, to detect changes.

Another important means of sun protection is protective clothing. A cross sectional study on 631 children from Canada found significantly lower nevus counts for the highest level of wearing clothes while in the sun.31 A longitudinal study in Australian adolescents could not observe an effect for wearing protective clothing, whether at school, on holidays or during childhood.30 An interventional study in 1,398 Australian school children showed slightly lower nevus counts in the intervention groups (education, education and offered low-cost sun-protective swim wear) compared to the control group, but the differences were not statistically significant.34 Cross-sectional multivariate analysis of our present study showed a protective effect of clothing at first assessment,55 which was confirmed by the longitudinal analysis of the entire cohort irrespective of our study arms by bivariate analysis. The protective effect of clothing was not observed on a significant level by multivariate analysis. Because our present study did not include clothing in the intervention measures these findings should not be over-interpreted and can not justify recommendations concerning sun protection by clothing. Special ultraviolet protective clothing remains uncommon in Germany and might be more effective.

In conclusion, the present randomized controlled trial in young German children could not demonstrate a significant change of sun exposure and sun protection habits due to an intervention with education and free sunscreens compared to a control group with a single educational session for all parents at study initiation. Analyzing the entire cohort irrespective of study arms neither a significant protective nor a harmful effect of sunscreens was found. High prevalence of sunscreen use at study commencement, social desirability and inadequate application of sunscreens might have partially covered their effect. As long as no proof of sufficient efficacy of sunscreens has been established, public education should aim primarily on ultraviolet radiation avoidance and sunscreens should be used additionally. A combination of sun protective measures seems advisable.

Ancillary