Randomized Controlled Trial of a Sun Protection Intervention for Children of Melanoma Survivors

In 2013, about 76,690 new cases of invasive melanoma are expected in the United States, and incidence increases by about 3% per year (1, 2). Sun exposure during childhood increases melanoma risk (3), yet 43% to 76% of children experience sunburns (4–7). Recommended sun protection behaviors include using sunscreen, wearing protective clothing, seeking shade, and limiting time outdoors during midday (8).

Children of melanoma survivors are an important yet understudied population for sun protection intervention. Melanoma risk is approximately doubled in individuals who have a first-degree relative with melanoma, due to possible shared genotypic and/or phenotypic factors (9–11). A pilot study of melanoma survivors identified from the Los Angeles County cancer registry showed that 49% of survivors reported that their children experienced sunburns in the past year (12). To our knowledge, there are no studies of interventions to promote sun protection in the children of melanoma survivors.

Sun protection interventions involving melanoma survivors may have more relevance to and hence greater impact on survivors who are parents than standard sun protection education available to parents in the general public. We developed a melanoma survivor–centered sun protection intervention and hypothesized that it would be more effective than standard education in decreasing children's sunburns, increasing children's sun protection, and improving survivors' psychosocial factors regarding children's sun protection.

Eligible survivors, identified from our hospital's patient registry, were diagnosed between 1990 and 2008 with stage 0 to stage IIIB melanoma (13); were age ≥18 years; were able to speak, read, and write English; and had a child ≤12 years old. Survivors were mailed a study invitation letter and study information. Study personnel followed up by telephone to ascertain eligibility and obtain informed consent (ClinicalTrials.gov #NCT00394134). The study was approved by the Institutional Review Board of The University of Texas MD Anderson Cancer Center (Houston, TX).

We conducted a 2-armed randomized controlled trial. After baseline assessment, we used minimization, a covariate adaptive approach to randomization (14). Compared with techniques such as stratification, minimization results in better study arm balance with respect to participant characteristics (14). Participants were assigned to receive either a targeted sun protection intervention (n = 170) or standard education (n = 170), while balancing study arms on several characteristics: children's and survivors' baseline sun protection, children's age, and survivors' sex, melanoma stage, year of diagnosis, and melanoma family history. We randomly assigned the first participant to one of the study arms. Before assigning each subsequent participant, we totaled the numbers of participants in each study arm who had characteristics that were similar to the participant to be assigned. We then assigned the participant to the study arm that minimized characteristic imbalance between arms.

The baseline assessment was conducted in spring and summer. Participants received 3 mailings (targeted intervention or standard education) at their homes over a 5-month intervention period in fall and winter. Participants completed follow-up assessments in spring and summer at 1 and 4 months postintervention.

The sun protection intervention included: (i) print booklet #1 and 10-minute DVD, (ii) print booklet #2 and magnet, and (iii) print booklet #3 and children's activity booklet. We developed materials based on our previous experience in developing video and print interventions to increase children's sun protection (15–17) and findings from qualitative research with melanoma survivors and their families. Materials development was further guided by social cognitive theory (SCT; ref. 18) and the health belief model (HBM; ref. 19), which have informed effective sun protection interventions for children and parents in the general population (15, 20–22).

SCT suggests mediators of behavior change, including self-efficacy (belief in one's capability to perform a behavior), outcome expectations (beliefs regarding anticipated outcomes of behavior), and proximal goals (behavioral intentions; ref/23). Observational learning is a central tenet of SCT. Observing the actions, reinforcement, and positive self-evaluative reactions (e.g., feeling good about efforts to reduce children's sun exposure) of role models depicted in the intervention was expected to improve survivors' self-efficacy, outcome expectations, and intentions related to children's sun protection. In HBM, health behavior performance is more likely if an individual perceives disease threat (risk and severity), believes that performance will reduce threat, perceives few performance barriers, and experiences cues to action (19). Knowledge also is an important determinant of health behavior (24). Parents' self-efficacy, outcome expectations, intentions, risk perceptions, perceived benefits, perceived barriers, and knowledge have been associated with children's sun protection (25–29).

Role models featured in the DVD and booklets were melanoma survivors who did not participate in the trial, their spouses, and their children of various ages. Families modeled the practice of sun protection in different settings (e.g., at home, in the park, and at the zoo). Through on-camera interviews in the DVD and personal stories and testimonials written by survivors for the booklets, survivors discussed sun protection expectations, why they protect their children (including risk perceptions and benefits), how they protect their children, and how they overcome sun protection barriers.

The children's activity booklet contained educational puzzles, songs, and quizzes for children to educate them about sun protection, increase their cooperation with survivors' sun protection efforts, and create opportunities for survivor–child interactions about sun protection. The magnet was designed to be a cue to action.

Standard education comprised 3 health-related brochures available to the general public: (i) sun protection (30), (ii) physical activity (31), and (iii) nutrition (32). These brochures were mailed on the same schedule as the sun protection intervention. The standard education group received all intervention materials after the study.

Survivors completed assessments by telephone at baseline, 1 month postintervention, and 4 months postintervention. If the survivor had 2 or more children ≤12 years old, one child was randomly selected and the survivor was asked to respond with this child in mind. Survivors were compensated $20 per assessment.

We assessed the number of sunburns children experienced during the past 6 months (baseline and 1 month postintervention) and during the past 3 months (4 months postintervention).

We measured individual behaviors and composite sun protection (a standard outcome that combines measures of each recommended behavior; ref. 33). Questions were developed for this study and on the basis of our previous research (15, 34). Questions assessed behavior regarding sunscreen (7 questions), clothing (5 questions), shade (1 question), and limiting time outdoors between 10 am and 4 pm (1 question) on 5-point response scales [1 (never) to 5 (always)]. As we assessed multiple aspects of sunscreen (e.g., application, reapplication, and coverage) and clothing (e.g., hats, sunglasses, sleeved shirts, and pants) behavior, we averaged each survivor's responses to these questions to create composite scores for children's sunscreen behavior and for children's clothing behavior. A composite sun protection score was computed by averaging the composite sunscreen score, the composite clothing score, the score for shade, and the score for limiting time outdoors.

We developed questions on the basis of our prior research (27, 34). Internal consistency reliability of multi-item scales was estimated by Cronbach's alpha (α), which was averaged over the 3 assessments. Self-efficacy for children's sun protection was assessed with behavior-specific measures: sunscreen (3 questions; α = 0.82), hats/clothing (2 questions; α = 0.60), shade (2 questions; α = 0.78), and limiting time outdoors (3 questions; α = 0.89). Questions (e.g., “How confident are you that you can make sure that your child keeps a hat on each time he/she is outdoors, even on a windy day?”) had 5-point response scales [1 (not confident at all) to 5 (extremely confident)].

Outcome expectations were assessed for tanning (3 questions; α = 0.79) and children's sunscreen (2 questions; α = 0.40), hats (3 questions; α = 0.72), clothing (3 questions; α = 0.76), shade (3 questions; α = 0.66), and limiting time outdoors (2 questions; α = 0.69). Questions (e.g., “Wearing clothing that covers most of his/her body would make my child feel too hot”) had 5-point response scales [1 (strongly disagree) to 5 (strongly agree)]. Higher scores indicated more positive tanning outcome expectations and more negative sun protection outcome expectations.

Intentions were assessed for children's sunscreen (2 questions; α = 0.79), hats/clothing (2 questions; α = 0.74), shade (1 question), and limiting time outdoors (1 question) behaviors. Survivors were asked about their intentions to protect their children each time they would be outdoors during the next 3 months with responses on 5-point scales [1 (strongly disagree) to 5 (strongly agree)].

Two perceived risk questions assessed survivors' perceptions of the likelihood that their children would develop melanoma/skin cancer or sunburn in the future. Questions had 4-point response scales [1 (very unlikely) to 4 (very likely)].

Perceived benefits of children's sun protection (5 questions; α = 0.73) was assessed by asking survivors whether sun protection would reduce their children's risk of developing skin cancer [1 (strongly disagree) to 5 (strongly agree)].

Sun protection knowledge (6 questions; α = 0.39) was assessed. Questions (e.g., “Does a sunscreen's SPF number tell you the sunscreen's level of protection against UVA and UVB rays from the sun?”) had yes/no/don't know response options that were coded as correct/incorrect. Missing and don't know responses were treated as incorrect.

At baseline, demographics, sun sensitivity, and clinical characteristics were collected (Table 1).

We asked survivors in the intervention group if they viewed/read none, some, most or all of the DVD and each print booklet, and whether their families viewed/read (yes/no) these materials. We asked survivors whether they used the magnet (yes/no), their children liked the activity booklet (yes/no), and the activity booklet helped them teach their children about sun protection (yes/no). On a scale of 1 (strongly disagree) to 5 (strongly agree), survivors rated the attractiveness of materials and whether the information presented was credible, useful, new or interesting. Survivors in the standard education group were asked if they read none, some, most or all of each brochure. All survivors were asked if they received sun protection information from other sources during the study.

Linear mixed models (PROC MIXED of SAS 9.1.3; ref. 35) were used to examine main effects on continuous sun protection and psychosocial outcomes. Generalized linear mixed models (PROC GLIMMIX of SAS 9.1.3; ref. 35) were constructed using a logit link function to examine effects on binomial sunburn outcomes. Because we assessed sunburns for different recall periods, we divided the number of sunburns by the number of months in the recall period to calculate a sunburn rate for each assessment. Binary scores were defined as a success if either the sunburn rate decreased from baseline to postintervention assessments or no sunburns were reported at baseline and postintervention assessments. Covariates (demographics, sun sensitivity, and clinical characteristics) that were statistically significant (P < 0.05) in univariate models were retained in final models that included terms for intervention effect and assessment. The probability of a type I error was 5%. No adjustments were made for multiple testing. For measures with low estimates of internal consistency reliability (sun protection knowledge and sunscreen outcome expectations), we also examined the individual items as intervention outcomes. Findings were consistent with analyses based on the complete measures so we do not report item–level outcomes.

We explored whether the intervention was more effective in survivor subgroups by examining potential moderators (e.g., children's age and baseline level of composite sun protection and survivors' sex, number of melanomas, time since diagnosis, and melanoma family history). Mixed models included covariates and a 3-way interaction term (study arm by time by moderator). Mediation analyses will be reported in a separate paper.

We used linear mixed models to compare children's composite sun protection outcomes for different levels of intervention use (none, some/most, and all) with outcomes for standard education. We also examined effects within the intervention group by assuming a linear dose response. To understand whether survivor subgroups were more likely to use the intervention, we used Wilcoxon rank-sum tests, Kruskal–Wallis tests, or Spearman correlation coefficients to examine the association between intervention use and survivor/child demographics, survivor/child sun sensitivity, survivor clinical characteristics, survivor baseline sun protection behaviors and psychosocial variables, and child baseline sun protection behaviors.

Of 2,014 screened melanoma survivors, 372 were eligible, 362 provided informed consent, and 340 completed the baseline assessment (Fig. 1). Most resided in Texas (82%) or other southern states (10%). Survivors randomized to standard education were more likely to be married (P = 0.03; Table 1); otherwise, study arms did not differ in baseline demographics, sun sensitivity, or clinical characteristics. Postintervention response rates were 87% (1 month, n = 295) and 83% (4 months, n = 281). Survivors who completed both follow-up assessments and those who did not complete one or both did not differ in baseline variables.

We evaluated intervention effects on individual sun protection behaviors and a composite sun protection behavior score. We observed positive effects on children's sunscreen reapplication after each hour outdoors at 1 month postintervention (Cohen's effect size, d = 0.37) and on children's wearing of wide-brimmed hats at 4 months postintervention (d = 0.24; Table 2). However, we did not observe intervention effects on other sun protection outcomes including children's sunscreen composite score, clothing composite score, shade behavior, limiting time outdoors behavior, or composite sun protection score. Children's sunburn rate also did not decrease following the intervention (1 month: OR = 0.95, P = 0.90; 4 months: OR = 1.01, P = 0.98).

Child age significantly moderated effects on children's sunscreen (1 month: B = −0.03, P = 0.005; 4 months: B = −0.04, P = 0.001) and clothing (1 month, B = −0.02, P = 0.043) behaviors: as child age increased, the intervention was less effective. For example, greatest effects on sunscreen behavior were in survivors who had children younger than 8 years old (1 month: B = 0.22, P = 0.013; 4 months: B = 0.31, P = 0.001). Survivors' sex moderated intervention effects on children's clothing (1 month: B = −0.17, P = 0.026; 4 months: B = −0.16, P = 0.037) and limiting time outdoors (4 months: B = −0.31, P = 0.049) behaviors: the intervention was less effective in female survivors. Family history of melanoma moderated effects on children's limiting time outdoors behaviors (1 month: B = −0.42, P = 0.028): the intervention was less effective in survivors who had a family history of melanoma. Moderators of the intervention effect on children's sunburns or shade behavior were not identified.

Baseline composite sun protection scores in children moderated intervention effects on children's clothing (1 month: B = −0.16, P = 0.029; 4 months: B = −0.17, P = 0.026) and limiting time outdoors (1 month: B = −0.77, P < 0.0001; 4 months: B = −0.47, P = 0.001) behaviors: as baseline scores increased, the intervention was less effective. The strongest intervention effect on children's clothing behavior was in survivors whose children had baseline sun protection scores lower than 3.4 (1 month: B = 0.32, P = 0.001; 4 months: B = 0.35, P < 0.0001). The strongest effect on limiting time behavior was in survivors whose children had scores lower than 3.0 (1 month: B = 0.45, P = 0.002; 4 months: B = 0.42, P = 0.004).

At 1 month postintervention, the sun protection intervention increased survivors' hats/clothing self-efficacy (d = 0.26; Table 3). At 4 months postintervention, the intervention improved survivors' hats/clothing self-efficacy (d = 0.33), clothing intentions (d = 0.27), sun protection knowledge (d = 0.32), and outcome expectations for children's hat (d = -0.40) and clothing (d = −0.24) behaviors.

Of survivors in the intervention group, 71% used at least some of the DVD and booklets (Table 4). Relatively few reported that their family members used the materials (Table 5). A greater proportion of survivors in the standard education group than in the intervention group reported receiving sun protection information from sources other than the study materials (36% vs. 22% at 1 month, P < 0.05).

Survivors who used all of the intervention DVD and booklets reported higher children's composite sun protection scores than did survivors in the standard education group (B = 0.27, P = 0.002). There were no effects for survivors who used none or some/most of the intervention materials. Dose–response analyses showed that while controlling for survivors' baseline sun protection intentions, children's composite sun protection increased as survivors' use of the DVD and booklets increased (B = 0.11, P < 0.0001).

Intervention use was higher in survivors who had higher baseline levels of sunscreen (P < 0.05), clothing (P < 0.05), limiting time outdoors (P < 0.05), and composite sun protection (P < 0.05) behaviors. Intervention use also was higher in survivors who reported more positive outcome expectations about limiting their children's time outdoors (P < 0.01) and more frequent children's limiting time outdoors behavior (P < 0.05) at baseline. Intervention use was not associated with other survivor or child characteristics.

In general, the intervention increased sunscreen reapplication and use of wide-brimmed hats in children. Other effects on children's sun protection or sunburns were not observed. The sunscreen reapplication findings were encouraging, especially considering that reapplication rates are low in children (36), children use less sunscreen per application than is recommended (37), and reapplication enhances protection (38). The effects on children's hat behavior and survivors' psychosocial factors that facilitate hat and clothing behaviors also were encouraging, as few interventions directed to parents have increased children's hat outcomes (15, 39, 40) and, in general, 8% or fewer of children wear wide-brimmed hats (7, 36).

At baseline, children were protected by sunscreen and clothing only “sometimes” or less frequently. Thus, this sample of melanoma survivors provided an important opportunity to test the effect of this new intervention on these sun protection behavioral and related psychosocial outcomes. The limited intervention effect on other outcomes may be partly due to the relatively high (i.e., desirable) baseline scores on measures and the relatively low baseline prevalence of children's sunburn (28%), which may have limited our ability to evaluate intervention effects. Shade use and limiting time outdoors may be influenced by external factors not addressed by this intervention such as shade availability in the environment and scheduling barriers. We cannot exclude the possibility that survivors in the standard education group engaged in compensatory behaviors, which would likely result in an underestimation of the intervention effect.

Effects may be strengthened by tailoring the intervention. Tailored interventions have increased sun protection (41, 42) and skin examination (41–43) in adult first-degree relatives of melanoma survivors and other adults at higher melanoma risk, suggesting that tailoring has the potential to be effective in melanoma survivors and their children. Personalized interventions may also draw from research on communication and melanoma risk attributions in families with melanoma (44–46).

Incomplete use of the intervention by survivors also may have contributed to the modest intervention effects. The DVD, in particular, had low rates of use, possibly because of inconvenience; substantially more survivors reported viewing none of the DVD compared with reading none of the booklets which could be easily skimmed upon receipt. We do not know to what extent perceived need influenced intervention use; however, survivors reported that they found the intervention to be useful and interesting.

We developed the intervention to be low cost and easily disseminated through regular mail. Future studies should identify strategies that increase intervention use by survivors and their families. For example, very little is known about how tailored materials may enhance intervention use by melanoma survivors. There is some evidence that tailored, compared with generic, print materials are read for a longer period of time by adult first-degree relatives of melanoma survivors (41). Mobile health and other e-health approaches [e.g., texts (ref. 47), mobile phone applications (ref. 48), web-based videos] are promising vehicles for sun protection interventions and need to be studied to determine whether they enhance the use of interventions designed to increase children's sun protection. Communication technology is rapidly evolving and parents may consider it more convenient to access video content through mobile devices than a DVD, as was used in this study.

Results from moderation analyses suggest that the intervention may be more effective in survivors who have less experience with sun protection, including males, survivors without a melanoma family history, and survivors whose children had lower levels of sun protection before the intervention. We speculate that the intervention may be more effective in this group of survivors because the booklets and DVD could be revisited as sun protection skills were developed, the information addressed frequently asked questions about sun protection, and the role modeling stories, video scenes, and booklet photographs facilitated engagement with, and absorption of, the content. Our intervention may be an important first step in increasing children's sun protection in this group of survivors. Although the intervention was developed for widespread dissemination to survivors regardless of child age, survivors with younger children benefited most from this intervention.

In community-based research, interventions guided by SCT and HBM have increased sun protection in children and parents (15, 20–22). Mediation analyses, currently underway, will help us determine whether a similar conceptual framework is relevant for interventions for melanoma survivors and their children. Interventions should be guided by theory to increase effectiveness in target audiences (33) and may need to combine multiple theories (24). Longitudinal studies are lacking in the literature on children's skin cancer prevention (49) and are needed to better understand the theoretical predictors of sun protection in melanoma survivors and their children. Our findings suggest that our theoretical approach may have been more useful in subgroups of survivors. Further study is needed to understand whether and how to adapt and apply theory to different subgroups in this population. The timing of intervention relative to the survivor's melanoma diagnosis also may suggest the use of theories such as Protection Motivation Theory (50) which address threat appraisals and coping processes.

Social desirability and recall biases are potential measurement limitations of this study. Validated psychosocial measures regarding children's sun protection are lacking (49). We developed measures for the current study to assess psychosocial constructs specific to individual sun protection behaviors. A few of the measures, namely sunscreen outcome expectations and sun protection knowledge, had low Cronbach's alpha estimates of internal consistency reliability, and this may limit the interpretation of findings based on these measures. Continual testing and refinement of measures will enhance their validity and reliability for future research.

We were unable to compare respondents with nonrespondents because the registry had no records on the age of survivors' children. A large proportion of survivors screened for the study did not have a child in the eligible age range. The mean age of melanoma survivors in the patient registry (51 years) was older compared to population-based estimates for parents who have children younger than 12 years (37 years; ref. 51). Most of the sample resided in Texas or another southern state, but we acknowledge that the potential for seasonal or geographic variation in intervention effect is a challenge in sun protection research and a possible limitation of this study. Our sample had a relatively high education level, which has been associated with increased melanoma incidence (52). The sample's distributions of education, race, and ethnicity were similar to those reported in other studies of melanoma survivors recruited from population-based cancer registries and comprehensive cancer centers (53–55), although distributions may vary nationally (12).

This is the first study to examine the effects of an intervention to promote sun protection in children of melanoma survivors. Our findings provide a valuable starting point for developing effective sun protection interventions in this group of children at higher risk of melanoma. Our melanoma survivor–centered sun protection intervention can improve some outcomes related to children's sun protection and the intervention may be more effective in survivors who have younger children or less experience with sun protection. Furthermore, intervention delivery must be enhanced to maximize use. Future research also should focus on evaluating this intervention in different samples of melanoma survivors to test the external validity of our findings and better describe the populations that may receive the most benefit.

No potential conflicts of interest were disclosed.

Conception and design: E.R. Gritz, M.K. Tripp, S.K. Peterson, A.V. Prokhorov

Development of methodology: E.R. Gritz, M.K. Tripp, S.K. Peterson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.R. Gritz, M.K. Tripp

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.R. Gritz, S.S. Shete, D.L. Urbauer, B.M. Fellman, J.E. Gershenwald

Writing, review, and/or revision of the manuscript: E.R. Gritz, M.K. Tripp, S.K. Peterson, A.V. Prokhorov, S.S. Shete, D.L. Urbauer, B.M. Fellman, J.E. Lee, J.E. Gershenwald

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.K. Tripp

Study supervision: E.R. Gritz

The authors thank Valen Johnson, PhD, and Ying Yuan, PhD, for their statistical contributions during this study.

This research was funded by a grant from the American Cancer Society (RSGPB 04-010-01-CPPB; to E.R. Gritz, PhD), and supported by the NIH/NCI under award number P30CA016672 and used the Patient-Reported Outcomes, Survey & Population Research Shared Resource and the Biostatistics Resource Group.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.