eAppendix. Trained Interviewers Administered a Structured Questionnaire by Telephone Using a Residence History Calendar to Aid Recall
eTable 1. Adjusted Odds Ratios for Melanoma With Sunscreen Use in Population Subgroups Defined by History of Sunburn, Age at Diagnosis of Melanoma, and Nevus Density
eTable 2. Adjusted Odds Ratios for Melanoma With Total Sun Exposure Inversely Weighted and Unweighted for Sunscreen Use in Population Subgroups Defined by Pigmentation Score, Use of Sunscreen to “Stay Outside Longer,” and Nevus Density
Customize your JAMA Network experience by selecting one or more topics from the list below.
Watts CG, Drummond M, Goumas C, et al. Sunscreen Use and Melanoma Risk Among Young Australian Adults. JAMA Dermatol. 2018;154(9):1001–1009. doi:10.1001/jamadermatol.2018.1774
Does an association exist between the use of sunscreen in childhood and early adulthood and risk of melanoma before age 40 years, and what factors are associated with sunscreen use among Australian children and young adults?
In this population-based, case-control family study of 603 patients and 1088 controls using Australian Melanoma Family Study data, childhood sunscreen use and lifetime sunscreen use were significantly associated with a decreased risk of cutaneous melanoma among young adults. Sex, age, ancestry, educational level, skin pigmentation, and sunburn were factors associated with sunscreen use.
Regular sunscreen use may reduce the risk of melanoma among young adults.
There are limited data among young adults on sunscreen use during childhood and adulthood and on the association of sunscreen use with melanoma risk.
To assess correlates of early-life sunscreen use and the association between sunscreen use and risk of cutaneous melanoma before age 40 years.
Design, Setting, and Participants
This population-based, case-control family study analyzed Australian Melanoma Family Study data for persons with questionnaire data on sunscreen use collected by interview from 2001 to 2005 across 3 states in Australia, representing two-thirds of the country’s population. Case participants (aged 18-39 years) had confirmed first primary melanoma. Siblings of case participants were included, and case participants without a sibling control were excluded. Unrelated controls (aged 18-44 years) were recruited from the electoral roll or were a spouse, partner, or friend nominated by case participants. Data analyses were conducted from October 2017 to February 2018.
Self- and parent-reported sunscreen use, sun exposure, and other candidate risk factors during childhood and adulthood.
Main Outcomes and Measures
Logistic regression analyses adjusted for potential confounders were used to estimate odds ratios (ORs) for melanoma and for correlates of sunscreen use.
Participation was 629 of 830 contactable cases (76%) (629 of 1197 overall [53%]), 240 of 570 contactable controls (42%) from the electoral roll (240 of 1068 overall [23%]), and 295 of 371 nominated spouse or friend controls (80%); analysis incuded 603 cases and 1088 controls. The median (interquartile range) age was 32 (28-36) years for 603 cases, 35 (30-38) years for 478 unrelated controls, and 34 (29-38) years for 610 sibling controls. There were more women than men (range, 57%-62%) in all groups, approximately 40% (range, 39%-43%) of participants had a university education, and most participants (range, 58%-73%) had British/northern European ethnicity. Risk of melanoma was less with higher use of sunscreen in childhood (OR for highest vs lowest tertiles, 0.60; 95% CI, 0.42-0.87; P = .02 for trend) and across the lifetime (OR, 0.65; 95% CI, 0.45-0.93; P = .07 for trend). Subgroup analyses suggested that the protective association of sunscreen with melanoma was stronger for people reporting blistering sunburn, receiving a diagnosis of melanoma at a younger age, or having some or many nevi. Total lifetime sun exposure was unrelated to melanoma risk (OR for highest vs lowest tertile, 0.97; 95% CI, 0.66-1.43; P = .94 for trend). By contrast, total sun exposure inversely weighted by sunscreen use (as a measure of sun exposure unprotected by sunscreen) was significantly associated with melanoma risk (OR, 1.80; 95% CI, 1.22-2.65; P = .007 for trend) and appeared stronger for people having lighter pigmentation or some or many nevi or using sunscreen to stay longer in the sun. Regular users of sunscreen were more likely to be female, younger, and of British or northern European ancestry and to have higher educational levels, lighter skin pigmentation, and a stronger history of blistering sunburn.
Conclusions and Relevance
Our findings provided evidence that regular sunscreen use is significantly associated with reduced risk of cutaneous melanoma among young adults and identified several characteristics associated with less sunscreen use.
The association of sun exposure and sunburn, particularly in childhood, with melanoma risk is well established.1-5 Despite sunscreen being widely available and recommended for sun protection, optimizing the use of sunscreens remains a challenge, and controversies continue to surround its use.6,7
Meta-analyses of observational studies, most conducted more than 15 years ago, have shown null or positive associations of sunscreen use with melanoma risk.8-10 This finding has generally been assumed to be due to uncontrolled confounding of factors associated with sunscreen use, including host factors, such as skin phototype, and behavioral factors, such as sun exposure (using sunscreen to extend time in the sun) and inadequate application of sunscreen.8 More recent analyses suggest a reduced risk of melanoma with sunscreen use.11-13 The only randomized clinical trial, conducted in Australia, found that regular sunscreen use was associated with a lower risk of invasive melanoma (hazard ratio, 0.27; 95% CI, 0.08-0.97; P = .05).12 A population-based cohort study in Norway showed that use of sunscreens with a sun protection factor (SPF) of at least 15 was associated with a reduced risk of melanoma compared with sunscreens with a lower SPF (hazard ratio, 0.67; 95% CI 0.53-0.83).11 A population-based case-control study in Minnesota observed a lower risk of melanoma with regular sunscreen use across 2 decades relative to nonuse (odds ratio [OR], 0.44; 95% CI, 0.23-0.86; P = .18 for trend).13
No studies have specifically examined the association between use of sunscreen in childhood and early adulthood and risk of melanoma before 40 years of age or investigated the factors associated with sunscreen use in childhood and early adulthood. To address these evidence gaps, we analyzed data from 603 cases and 1088 controls in the Australian Melanoma Family Study,14 a population- and family-based case-control study designed to examine risk factors for melanoma among young adults.
The Australian Melanoma Family Study was conducted in Victoria, New South Wales, and Queensland, states that together comprise two-thirds of the Australian population. Case participants were aged 18 to 39 years and had received a confirmed first primary cutaneous melanoma diagnosis between July 2000 and December 2002. Unrelated controls aged 18 to 44 years were recruited from the electoral roll (registration is compulsory for Australian citizens aged 18 years or older) or from a spouse, partner, or friend nominated by the patients. Members of the case and control families were also recruited,15 and siblings of case participants were analyzed as sibling controls in secondary analyses. Participation was 629 of 830 contactable cases (76%) (629 of 1197 overall [53%]), 240 of 570 contactable controls (42%) from the electoral roll (240 of 1068 overall [23%]), and 295 of 371 nominated spouse or friend controls [80%]. The research and ethics committees of Cancer Councils in Queensland, New South Wales, and Victoria and at The University of Sydney and The University of Melbourne, all in Australia, approved this study. All participants provided written informed consent.
Data were collected from 2001 through 2005. Participants provided information about their demographics, ethnicity, family history of melanoma, and sun bed use. Participants self-reported hair and eye colors at age 18 years, number of nevi, and skin phenotype. A pigmentation-related score16 was created to summarize 6 phenotypic variables associated with melanoma risk: skin color, hair color, eye color, self-reported ability to tan, propensity to sunburn, and childhood freckling; the continuous score ranged from 0.02 for darker pigmentation to 0.56 for lighter pigmentation (higher score).
Sun exposure, sunburn, and sunscreen use at the ages of 10, 15, and 20 years and the subsequent decade years were collected. Data on parent-reported sun exposure and sunscreen use were also collected when possible as a proxy measure of childhood exposures at ages 3, 5, 10, and 15 years. Summary values for childhood and parent-reported childhood exposures (up to age 17 years) and lifetime exposures (up to the age at interview or diagnosis) were derived from questionnaire responses as described previously.3 Questions about sunscreen use (eAppendix in the Supplement) included frequency of use, SPF factor (≥8), and whether sunscreen was used to stay in the sun longer. Summary variables for childhood and lifetime sunscreen use were calculated as weighted mean scores based on Likert scale responses and number of years in each age period. The weighted mean scores were categorized into tertiles, a standard practice when no clear cutoff points exist, based on the distribution of controls for lifetime sunscreen use. The reference tertile was for less sunscreen use.
To examine the influence of sunscreen use on the association of sun exposure with melanoma risk, we created a variable for total sun exposure that was inversely weighted for sunscreen use. This variable was generated by multiplying the hours of sun exposure at each age by a value that corresponded to frequency of sunscreen use reported at the same age. Thus, the number of sun exposure hours was multiplied by 0 if participants always or almost always used sunscreen when outdoors, by 0.25 if sunscreen was used more than half the time, by 0.5 if it was used about half the time, by 0.75 if it was used less than half the time, and by 1 if it was never used. Therefore, higher scores of sun exposure inversely weighted for sunscreen use indicated greater UV radiation exposure and less sunscreen use.
Unconditional logistic regression was used to calculate ORs and 95% CIs for correlates of sunscreen use and for the association between sunscreen use and melanoma risk among unrelated controls. Associations between sunscreen use and melanoma risk among sibling controls were analyzed using conditional logistic regression matched by family; thus, case patients who had no sibling control were excluded from those analyses. Participants recruited via the electoral roll and spouse or friend controls were analyzed together as 1 control group given their similar associations for standard risk factors.14 Multivariable models for correlates of sunscreen use were constructed using forward stepwise selection, with P < .10 for candidate selection and P < .05 for model entry, and covariates of age, sex, and state of recruitment were forced into the models. No adjustments were made for multiple testing.
For models examining the association between sunscreen use and melanoma risk, we adjusted for age, sex, and state of recruitment in the minimally adjusted models, and in fully adjusted models, we further adjusted for ancestry, educational level, confirmed family history of melanoma in a first-degree relative, nevus density, sun bed use, lifetime blistering sunburns, outdoor hours, and pigment score. Because childhood UV radiation exposure is an important determinant of nevus density, we examined nevus density as both a potential confounder and effect modifier.
We examined the interrater reliability of childhood sunscreen use self-reported by participants with that reported by their parents or caregivers, using the weighted κ statistic to compare tertiles of exposure and using the intraclass correlation coefficient to compare continuous exposure.17
Using data from the case participants and unrelated controls, we tested whether the associations between sunscreen use and melanoma risk differed by age, sex, age at diagnosis, time spent outdoors, pigmentation score, SPF factor used, application of sunscreen to stay outdoors longer, history of blistering sunburns, and nevus density. Interactions were tested by adding to the model a product term between the sunscreen tertile variable and covariate fitted as a 1 df ordinal variable to test for interaction in the trend effect. Age and pigmentation score were dichotomized in the interaction analyses. All analyses were conducted from October 2017 to February 2018 using SAS, version 9.4 (SAS Institute Inc). A 2-sided P < .05 was considered statistically significant. We followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines for reporting results.18
The analysis included 603 cases and 1088 controls who were younger than 45 years with questionnaire data on sunscreen use. The controls comprised 478 unrelated controls (229 electoral roll and 249 spouse or friend controls) for the primary analyses and 610 sibling controls for secondary analyses. Parent-reported data were available for 231 of 372 cases (62%) and 77 of 192 unrelated controls (40%) for case-control analyses and for 101 of 246 cases (41%) and 102 of 249 sibling controls (41%) for case-sibling analyses. Table 1 provides the demographic characteristics for the sample. The median (interquartile range) age was 32 (28-36) years for case participants, 35 (30-38) years for unrelated controls, and 34 (29-38) years for sibling controls. There were more women (range, 57%-62%) than men in all groups, approximately 40% (39%-43%) of participants had a university educational level, and most participants (58%-73%) had British/northern European ethnicity.
Self-reported childhood sunscreen use and lifetime sunscreen use were both associated with a lower risk of melanoma (for highest tertile, fully adjusted OR, 0.60; 95% CI, 0.42-0.87; P = .02 for trend vs for lowest tertile, fully adjusted OR, 0.65; 95% CI, 0.45-0.93; P = .07 for trend) (Table 2). The risk estimates were stronger for the fully adjusted than for the minimally adjusted models; removing nevus density from the model had minimal influence (≤5% change). No protective association was apparent using parent-reported (child proxy) data.
The analysis using sibling controls showed a significant inverse association of parent-reported childhood sunscreen use with melanoma (OR, 0.08; 95% CI, 0.02-0.41; P = .001 for trend) but not for self-reported sunscreen use (Table 2). Weighted κ statistics and intraclass correlation coefficients comparing self-reported and parent-reported sunscreen use indicated moderate interrater agreement (Table 3).
Total lifetime sun exposure was unrelated to melanoma risk (for highest vs lowest tertile of exposure: OR, 0.97; 95% CI, 0.66-1.43; P = .94 for trend). However, when sun exposure was inversely weighted by sunscreen use (ie, always, 0; more than half the time, 0.25; half the time, 0.5; less than half the time, 0.75; never, 1) as a measure of UV radiation exposure “unprotected” by sunscreen, there was a significant association with melanoma risk (for highest vs lowest tertile of weighted exposure: OR, 1.80; 95% CI, 1.22-2.65; P = .007 for trend) (Table 4). A similar pattern was observed for childhood exposures. When sibling controls were analyzed, total lifetime sun exposure was inversely associated with melanoma risk, and this association was attenuated by weighting with sunscreen use (Table 4).
Regular users of sunscreen (ie, used sunscreen more than half the time when outdoors) during childhood were more likely to be female and younger and to have a higher educational level, lighter skin pigmentation, and stronger history of blistering sunburns (Table 5). Correlates of lifetime sunscreen use included the same factors with the addition of British/northern European ancestry and the exclusion of sunburn (Table 5).
Self-reported lifetime sunscreen use appeared more strongly associated with lower risk of melanoma among people reporting at least 1 blistering sunburn than among people with no blistering sunburn (P = .06 for interaction) (eTable 1 in the Supplement). When based on parent-reported sunscreen use, childhood sunscreen use appeared more strongly associated with reduced risk of melanoma for people who received a diagnosis of melanoma at a younger age (≤33 years) than at older ages (34-39 years) (P = .06 for interaction) and for people with some or many nevi than those with fewer nevi (P = .01 for interaction) (eTable 1 in the Supplement). Although we did not show the results, the association of sunscreen use with melanoma risk did not vary by other factors.
Lifetime sun exposure inversely weighted for sunscreen use was more strongly associated with melanoma risk among people who reported using sunscreen to stay outside in the sun longer (P = .07 for interaction), among people with lighter pigmentation (P = .07 for interaction), and among people with some or many nevi (P = .03 for interaction) (eTable 2 in the Supplement).
To our knowledge, the present study is the first to examine the association of sunscreen use with melanoma among adults younger than 45 years, and the results suggest that both childhood and lifetime use of sunscreen may be protective against melanoma. Evidence for this association is strongest in our analysis showing that total sun exposure, otherwise apparently not associated with melanoma risk, was associated with melanoma risk when inversely weighted by sunscreen use. Subgroup analyses suggested that the protective association of sunscreen with melanoma was stronger for people reporting blistering sunburn, receiving a diagnosis of melanoma at a younger age, or having some or many nevi and that sun exposure “unprotected” by sunscreen was more strongly associated with melanoma risk for people with lighter pigmentation, people with some or many nevi, or people who used sunscreen to stay longer in the sun. Regular users of sunscreen were more likely to be female, younger, and of British or northern European ancestry and have higher educational levels, lighter skin pigmentation, and a stronger history of blistering sunburn.
Our findings are consistent with and add to the evidence from the 1 relevant randomized clinical trial12 and 2 more recent observational studies,11,13 which are briefly summarized in the Introduction. None of these studies predominantly investigated adults younger than 45 years nor reported results separately for this group. In the study by Green et al,12 participants were 25 to 75 years of age at entry: 55% were younger than 50 years, as were 48% of those who received a diagnosis of melanoma on follow-up. In the study by Lazovich et al,13 participants were 25 to 59 years of age, and about 31% were younger than 40 years old. In the study by Ghiasvand et al,11 women were 40 to 75 years of age at entry to the cohort.
Some subgroup associations have been examined in previous observational studies. Lazovich et al13 observed an apparently protective effect of sunscreen only among people with lower sun exposure. They hypothesized that sunscreen was being used to compensate for longer time in the sun among people with higher sun exposure and that the lack of any evident benefit from sunscreen was due to inadequate sunscreen application. We did not observe modification of the effect of sunscreen by total sun exposure. Ghiasvand et al11 found that, relative to nonusers, sunscreen users with a history of sunburn had a higher risk of melanoma, whereas sunscreen users with no history of sunburn had a lower risk of melanoma. The authors also hypothesized that sunscreen was being used to compensate for longer time in the sun and that overall UV radiation exposure was increased owing to inadequate use of sunscreen. Variations in the way sunscreens are applied and differences in the responses of the skin to UV radiation exposure have been well documented.19,20 Our results indicated the opposite of those reported by Ghiasvand et al11; that is, the inverse association between sunscreen use and melanoma risk was slightly stronger among those who had reported a previous blistering sunburn. We observed that the association of sun exposure “unprotected” by sunscreen with increased melanoma risk was particularly strong for people with some or many nevi or with lighter pigmentation and for those who reported using sunscreen to stay outside in the sun longer; these results strongly support a protective role for sunscreen use among these subgroups.
To our knowledge, inverse weighting of sun exposure by sunscreen use is a novel approach to exposure measurement in epidemiologic studies of the effects or the association of sunscreen use on or with the risk of melanoma. In our analysis, inverse weighting of sun exposure by sunscreen use meant that sun exposure in a person who always used sunscreen when outdoors was represented as zero, whatever the actual value. For a person who never used sunscreen, the full value of sun exposure was used, and for others with partial sunscreen use, a ratio (0.25, 0.5, 0.75) was used. This approach is justified because it treats sunscreen as a product working as it was intended (ie, to reduce the amount of sun exposure incident on otherwise unprotected skin that reaches the epidermis). The inverse weighting by sunscreen use is an approximation, and because sunscreen is not always properly applied or reapplied, this method may overestimate the level of UV radiation protection at each level of exposure.
Our findings of the characteristics independently associated with more regular sunscreen use are consistent with previous studies11,21,22 and highlight the difficulty of influencing the sun protection behaviors of men, people with lower levels of education, and people who perceive themselves to be at lower risk of melanoma, such as those with darker pigmentation or apparent resistance to sunburn.23 The randomized clinical trial of sunscreen use12 observed similar relative effect sizes across different subgroups of skin color and tanning ability, history of sunburn, and history of skin cancer, among others, in a largely European origin participant population. It is important, therefore, that sun protection messages reach the whole population.
That participants in our study were relatively young is likely to have led to better recall of early-life sun exposure compared with studies investigating older participants. Our findings are also more likely to reflect contemporary patterns of sunscreen use, which may be, speculatively, more a consequence of planning for sun exposure than of responding to sun exposure. Such a change could explain our observation of an apparent protective association with melanoma, whereas earlier studies generally showed no association.8,9 Participants in our study would have been born in the 1960s, 1970s, and early 1980s. In Australia during the 1980s (corresponding to participants’ childhood and early adulthood), mass media sun protection campaigns, such as Slip-Slop-Slap (launched in 1981),22 were effective in improving sun protection behaviors.24 Our finding that educational level correlated with sunscreen use suggests that educational level probably influenced adoption of media and clinician recommendations for improved sun protection behaviors. Other strengths of our study include the detailed and comprehensive data on sunscreen use and other melanoma risk factors collected at different time points, including childhood.
Data on both self- and parent-reported early-life sunscreen use were collected in our study. We found moderate agreement between self-reports and parent reports, but the association of sunscreen use with melanoma risk was stronger when based on parent reports. The differences might reflect the different ages captured in the questionnaires because parent-reported questionnaires included additional questions about sunscreen use at ages 3 and 5 years. Differences in the strength of the association may also reflect recall bias of parental reports because a strong protective association between childhood sunscreen use and melanoma risk based on case-sibling control analyses suggested that children in the same family had quite different uses of sunscreen. It seems unlikely that parents would apply sunscreen differently to their children, although once children are more independent, parents have reduced influence on their children’s sunscreen use.25 Studies have found moderate agreement when comparing proxy and self-reports of risk factors26 and minimal evidence for biased recall of UV radiation-associated exposures,14 although, in a previous analysis, we did observe possible parental recall bias of childhood sunburn.3
Our study was designed when sunscreens of SPF 8 were widely available and considered to provide high-level protection. Because products with SPFs higher than 30 are now commonly used,27 it is possible that these higher SPF sunscreens have a stronger protective association with melanoma risk than that observed in our study, assuming the same frequency of application as with lower SPF sunscreens. Greater choice of sunscreen products and increasing awareness from social media and health care professional sources may have also influenced sunscreen use. Nevertheless, factors associated with less sunscreen use in our study, such as lower educational level, male sex, and sunburn, are consistent with more recent studies.28 Changes in the prevalence of sunscreen use would be unlikely to influence our finding that sunscreen use was associated with reduced melanoma risk because our analyses focused on dose-response associations. We were unable to define an exact amount of sunscreen use associated with lower risk owing to the complexity of the sunscreen exposure variables that were summarized as weighted mean scores from exposure at different ages. We do not know whether sunscreen was applied effectively nor whether other types of sun protection were used. It is also possible that participants overreported sunscreen use owing to social desirability bias. We did not adjust for multiple testing to avoid being too conservative in examining our prespecified hypotheses.29 Finally, we were unable to assess the association of sunscreen use with stage of melanoma.
Our results support the regular use of sunscreen to reduce the risk of melanoma in early adulthood and emphasize the need to reach all population subgroups with sun protection messages.
Accepted for Publication: April 30, 2018.
Corresponding Author: Caroline G. Watts, MPH, PhD, Cancer Epidemiology and Prevention Research, Sydney School of Public Health, The University of Sydney, Level 6-North, The Chris O’Brien Lifehouse, 119-143 Missenden Rd, Camperdown, New South Wales, Australia 2050 (firstname.lastname@example.org).
Published Online: July 18, 2018. doi:10.1001/jamadermatol.2018.1774
Author Contributions: Drs Watts and Cust had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Watts, Armstrong, Giles, Hopper, Mann, Cust.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Hopper, Cust.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Drummond, Goumas.
Obtained funding: Armstrong, Aitken, Giles, Hopper, Mann, Cust.
Administrative, technical, or material support: Watts, Schmid, Hopper, Mann, Cust.
Supervision: Hopper, Mann, Cust.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported in part by the Australian Melanoma Family Study and received funding from the National Health and Medical Research Council (NHMRC) of Australia (grants 566946, 107359, 211172, and 402761); the Cancer Council New South Wales (grants 77/00 and 06/10); the Cancer Council Victoria; the Cancer Council Queensland (grant 371); and R01 grant CA-83115-01A2 from the US National Institutes of Health to the Melanoma Genetics Consortium.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We thank the participants as well as the research coordinators, interviewers, examiners, and data management staff for their work and dedication.
Additional Information: Dr Cust is the recipient of a Career Development Fellowship (grant 1147843) from the NHMRC and a Career Development Fellowship (grant 15/CDF/1-14) from the Cancer Institute NSW.