Daily self-administered questionnaire (translated from original questionnaire in French). Participants filled in the questionnaire each day, circling by half-hour periods their type of clothing and exposure. From these data, a duration of daily sun exposure was calculated. Clothing types 1, 2, and 4 for women were considered to represent sun exposure with a swimming suit, and exposure was counted as 30 minutes in situation C, 15 minutes in situation B, and 0 otherwise. In the example shown, for the 7 periods mentioned, calculated exposure duration is 45 minutes.
Trial profile. SPF indicates sun protection factor.
Dupuy A, Dunant A, Grob J, for Réseau d’Epidémiologie en Dermatologie. Randomized Controlled Trial Testing the Impact of High-Protection Sunscreens on Sun-Exposure Behavior. Arch Dermatol. 2005;141(8):950–956. doi:10.1001/archderm.141.8.950
High-protection sunscreens have been suspected to prompt people to increase sun exposure, and thus to increase skin cancer risk. We tested the influence of both the actual protection (sun protection factor [SPF]) and the information about protection (label) on sun-exposure behavior.
Randomized controlled trial.
Four French seaside resorts during summer 2001.
A total of 367 healthy subjects during their 1-week holiday. Outcome was assessable in 98% of them.
Subjects were offered free sunscreens, with randomization into the following study arms: (1) SPF 40 labeled as “high protection”; (2) SPF 40 labeled as “basic protection”; and (3) SPF 12 labeled as “basic protection.” Arm 4, ie, SPF 12 labeled as “high protection,” was not implemented for ethical reasons. Subjects were not aware of the real target of the study and were blinded to the SPF value.
Main Outcome Measure
Duration of sunbathing exposure during 1 week. Secondary outcomes were occurrence of sunburns and amount of sunscreen used. Influences of SPF and label were assessed separately.
Compared with the low-SPF group, the high-SPF group did not have longer sunbathing exposure (12.9 ± 7.2 h/wk for high SPF vs 14.6 ± 6.7 h/wk for low SPF; P = .06), experienced fewer sunburns (14% vs 24%; P = .049), and used less sunscreen (median, 30 g vs 109 g; P<.001). The label “high protection” or “basic protection” had no influence on these end points.
In this adult population, higher SPF had no influence on duration of sun exposure and offered better protection against sunburns. Although higher SPF may increase sun exposure duration in specific populations, this effect cannot be viewed as a universal side effect of high-SPF sunscreens.
Sun exposure is the most important environmental factor involved in the development of skin cancer.1 Melanoma is one of the solid tumors with the greatest increase in incidence during the past 3 decades and accounts for the vast majority of skin cancer deaths.2 Although increased detection may possibly account for some of the increased incidence, behaviors that increase sun exposure3,4 have been suggested to be major contributors to the rising incidence. The sustained increase in melanoma incidence during the past few decades highlights the fact that this disease represents a major public health management issue.5 Public health policies stress prevention through simple behavioral measures: minimizing sun exposure, wearing protective clothing, and using sunscreens.6
Daily sunscreen use has been demonstrated to prevent squamous cell skin carcinoma in a community-based randomized controlled trial.7 However, there is a controversy about the relationship between sunscreen use and melanoma.8- 13 Several case-control studies have suggested that sunscreen use could be a risk factor rather than a protective one for melanoma,14- 16 although 2 meta-analyses have been inconclusive.17,18 Various hypotheses have been raised regarding sunscreen use and melanoma risk. First, sunscreens with high but selective anti–UV-B protection may have played a harmful role by inducing relative overexposure in the UV-A spectrum, which may be an important source of carcinogenic radiation. However, this explanation may become less and less relevant, since new sunscreens with more balanced protection against UV-A and UV-B have been developed in recent decades; moreover, the wavelength responsible for melanoma development is still unknown. Second, inadequate use in terms of quantity applied, number of applications, and practical modalities of application may also play a deleterious role.19,20 Finally, one of the most troublesome current criticisms of sunscreen use is that higher protection by more potent sunscreens may induce longer sun exposure by postponing warning signs such as sunburn, or by providing a false impression of safety in the sun. In this regard, 2 recent randomized trials in students during their holidays have shown that the use of sunscreens with high sun protection factor (SPF) prompted them to increase the duration of their sun exposure.21,22
Whether these results in students can be extrapolated to other populations is a crucial issue in terms of public health. Indeed, despite campaigns in favor of protection by clothing and limitation of sunbathing, most people in developed countries use sunscreens as a first-line protection. Furthermore, official guidelines recommend the use of high-SPF sunscreens,23 rather than low-SPF ones. We thus designed a randomized trial to test whether high-SPF sunscreens have an impact on the sun-exposure behavior in people spending family holidays. We jointly addressed the impact of the actual high protection and the impact of the impression of being well protected.
Our objective was to test the influence on sun-exposure behavior in a usual family holiday week at the seaside of (1) the actual ability of a sunscreen to protect against UV-B, which is linked to its SPF, and (2) the information on the protection given by the label on the tube.
We initially designed a theoretical 2 × 2 factorial randomized study testing both SPF (SPF 40 vs SPF 12) and label on the tube (“high protection” vs “basic protection”) on the duration of sun exposure in an adult population in seaside resorts. The 4 groups were designed as follows: (1) SPF 40 sunscreen labeled as “high protection”; (2) SPF 40 sunscreen labeled as “basic protection”; (3) SPF 12 sunscreen labeled as “basic protection”; and (4) SPF 12 sunscreen labeled as “high protection.” The ethical committee did not allow the implementation of group 4. We therefore implemented 3 randomization groups (1, 2, and 3) according to the actual SPF and sunscreen label.
The study took place in 4 French V V F Vacances resorts situated on the French Mediterranean (Balaruc-les-Bains, La Grande Motte, Le Grau-du-Roi) and Atlantic (Seignosse) coasts, during 5 consecutive weeks in July and August 2001.
Adults on their day of arrival at the resort were eligible for the study, if they considered themselves sunscreen users and volunteered to participate in a 1-week study on “skin and sun,” ie, they agreed to use exclusively the free sunscreen given by the on-site investigator during 1 week, to keep this sunscreen for their own use, and to complete a questionnaire every evening. Exclusion criteria were a history of skin cancer, recent history of severe sunburn, contraindication to sun exposure, known contact dermatitis to a sunscreening agent, pregnancy or breastfeeding, and participation of another member of the family in the study.
With the agreement of the ethical committee, participants were not informed (1) of the existence of a randomization process, (2) that different SPFs and labels were being tested, and (3) of the actual objectives and outcomes, since this information was expected to influence their behavior. An information sheet was given containing pseudo-objectives not related to the actual ones. After the study had been completed, participants were fully informed by mail of the true objectives and outcomes, and the reasons for masking were explained.
Sunscreens were especially made for this study by one of the biggest manufacturers in France; SPF 40 and SPF 12 sunscreens had the same components except for amount of filters and screens. No mention of the actual SPF was made on the label. Sunscreens had slightly different cosmetic properties on direct comparison, SPF 12 being easier to spread.
A computer-generated randomization list, stratified by center, with a fixed 8-length block design was drawn up by the statistician (A. Dupuy). According to this list, the manufacturer prepared sealed boxes labeled with the randomization number, each containing 4 identical 150-mL sunscreen tubes. The randomization number was also mentioned on each of the 4 tubes. The on-site investigator received the boxes ordered by randomization number. He was unaware of the allocation until the participant opened the box and was then able to acknowledge only the type of label, not the actual SPF.
Participants were randomly assigned on an individual basis to 1 of the following 3 groups: (1) SPF 40 labeled as “high protection” (referred to herein as the “high/40” group); (2) SPF 40 labeled as “basic protection” (the “basic/40” group); and (3) SPF 12 labeled as “basic protection” (the “basic/12” group). No mention of the actual SPF was made on the label. Each participant received 4 identical 150-mL tubes of the assigned sunscreen. Brand-name sunscreens of intermediate SPF were given to the rest of the family to avoid use of the tested sunscreen by the participant’s relatives.
Participants had to complete an initial questionnaire that included questions about their general sun-exposure behavior. Every evening, they had to complete a self-administered questionnaire describing, for every half-hour period during that day, the type of sun exposure and body surface exposed to sun (Figure 1). A final interview was collected at the end of the last day.
The main outcome measure was the duration of “sunbathing,” defined as sun exposure while wearing a swimming suit or equivalent. The method of calculating duration of exposure from the questionnaire data is demonstrated in Figure 1. The mean cumulative exposure by subject for the week was compared across the randomization groups. The other outcome measures were the occurrence of sunburns (defined as “sunburn or painful skin reddening” assessed during a structured interview at the end of the last day of participation) and sunscreen consumption, measured by weighing all the sunscreen tubes at the end of the study. Tube weighing was done with the investigator blinded to allocation group.
To have a 90% chance of detecting as significant (at the 2-sided 5% level) a 2-hour difference during the week between 2 groups with an assumed standard deviation of 4.5 hours and a mean duration of exposure during the week of 14 hours, 107 subjects in each group were required.
As the 2 × 2 factorial analysis could not be used, 2 separate analyses were conducted: (1) the effect of SPF was analyzed by comparing both groups with 2 different SPFs but featuring the same label (the basic/40 and basic/12 groups); and (2) the effect of label was analyzed by comparing both groups with 2 different labels but with the same SPF (the high/40 and basic/40 groups). Proportions of subjects with sunburn during the week were compared with the χ2 test and by logistic regression. Sunscreen consumptions were compared by nonparametric analysis of variance (Kruskal-Wallis test). All analyses were adjusted by center and by week period (all people having the same weather within center and week period). Because phototype (skin color and sensitivity to sun) is known to influence sunscreen consumption and sun exposure, it was also included in the regression models. Unless otherwise specified, the analyses were carried out on an intention-to-treat basis. Two-sided significance tests (5% level) were used throughout. Statistical analysis was conducted with SAS software, version 8.02 (SAS Institute Inc, Cary, NC).
Three hundred sixty-seven subjects were randomized. Eighty percent were female; their mean ± SD age was 39 ± 10 years. Eight subjects resigned just after randomization and could not be analyzed for any outcome. Three hundred fifty-nine subjects were followed up for at least 1 day and could be analyzed for duration of sun exposure. Among them, 342 (95%) completed the 1-week study. A flow diagram of the trial is presented in Figure 2. Baseline characteristics in the 3 groups are presented in Table 1.
Mean ± SD duration of sunbathing exposure during the week was 14.2 ± 7.6 hours in the high/40 group, 12.9 ± 7.2 hours in the basic/40 group, and 14.6 ± 6.7 hours in the basic/12 group. The SPF was not associated with significantly different durations of sunbathing during the week (P = .06), nor was the label (P = .13). Of note, the duration of exposure was higher in the basic/12 group than in the basic/40 group. No significant overall difference between the groups in duration of sun exposure regardless of clothing could be demonstrated either (Table 2). Even when the analysis was performed in a “per-protocol” population (n = 276; 77% of the intention-to-treat population), ie, after exclusion of the participants who declared that they had used, at least once during the week, another sunscreen than the one provided, neither the SPF nor the label significantly influenced the duration of sunbathing (P = .13 and P = .25, respectively).
The proportion of subjects who experienced sunburn during the week was higher in the low-SPF group than the high-SPF group with the same label (29 subjects [24%] in the basic/12 group vs 16 [14%]; P = .049 by χ2 test). The adjusted risk for sunburn associated with a low SPF was close to significance: odds ratio, 1.96; 95% confidence interval, 0.98-3.92; P = .06. Conversely, in the 2 groups with the same SPF, the type of label was not related to the risk of sunburn: 16 subjects (14%) in the “basic protection” label group vs 18 subjects (15%) in the “high protection” label group (odds ratio, 0.91; 95% confidence interval, 0.43-1.91; P = .80).
Overall, 63 participants experienced a sunburn, and it was severe (pain during more than 2 days or blistering) in 6 (10%) of them. The occurrence of sunburns did not stop participation in the study for most of the participants (58 [92%] of subjects who had a sunburn completed the study).
Although it was not the main reason for their holidays (see Table 1), 344 (96%) of the participants declared that they intended to get a tan, and 325 (91%) declared they had gotten a tan at the end of the week.
Three hundred forty-three subjects (96%) declared that they had used the sunscreen given by the investigator at least once, and 276 (77%) reported having used that sunscreen exclusively. Sunscreen consumption could be assessed by weighing tubes for 347 subjects (97%). From a total amount of 600 g freely given to each participant, the consumption per participant was significantly higher in the low-SPF group (median, 109 g; range, 0-470 g) than the high-SPF group with the same label (median, 30 g; range, 0-420 g) (P<.001). Across both groups with the same SPF 40, the label had no influence on the amount of sunscreen used during the week: 26 g (range, 0-330 g) in the “high protection” label group vs 30 g (range, 0-420 g) in the “basic protection” label group (P = .17). The results were similar in the population who reported having used exclusively the study sunscreens (data not shown).
This randomized intervention study compared the effect of different sunscreens on sun-exposure behavior in resorts for family summer holidays. In this population, our findings do not support the hypothesis that a higher SPF induces a higher exposure by delaying the alarm signs, nor the hypothesis that mentioning “high protection” on the label may induce longer exposure by giving an impression of safety. In addition, this study logically confirms that the use of higher-SPF sunscreens does reduce the number of sunburns in real life. Finally, our results suggest that people tend to self-regulate their sun protection with sunscreens, by inversely adapting the amount of sunscreen to the SPF, at least when sunscreens are freely delivered.
This trial is the first one, to our knowledge, to assess at the same time the influence of both the actual protection (the SPF) and the information about the quality of protection (the label) on sun-exposure behavior. Although, for ethical reasons, we could not use a complete factorial design that would allow a joint analysis of these 2 factors, the power was sufficient to analyze the differences by pair groups.
Several strengths in the design should be stressed. First, the study took place in real-life conditions, and we paid close attention to not modifying the usual behavior of the participants except for the intervention. They were not only blinded to SPF but also not informed of the actual design and objectives of the study, as is allowed by French law for studies dedicated to behavior analyses. Second, stratifying by center and adjusting the analyses for the week period led to ideal conditions for comparability of the weather conditions, which are an obvious determinant of sun-exposure behavior. Third, we avoided cross-contamination between participants by including a small number of participants each week and allowing only 1 participant per family; in addition, participants’ family members were given free open-label sunscreens of intermediate SPF. Finally, on-site investigators collecting the exposure questionnaire every evening ensured high-quality follow-up of participants.
Some limitations should also be discussed. First, the sex disequilibrium in the participants had not been anticipated. As we allowed only 1 adult participant per family, the female spouse or the mother was more likely to participate than the male. Whether this bias may have selected or rejected people with more risky sun-exposure behavior is difficult to assess. Second, some cosmetic differences could be noted between the SPF 40 and the SPF 12 sunscreens, the latter being easier to spread. Therefore, although participants were blinded to SPF value and knowledge of the study objective (ie, that different SPFs were being tested), it cannot be excluded that differences observed in the amount of sunscreen used may have been a consequence of cosmetic properties, rather than a consequence of a sense of suboptimal protection with lower-SPF sunscreens. However, a participant could not compare the 2 different sunscreens directly. Moreover, SPF is intrinsically linked to cosmetic properties and, therefore, these 2 factors cannot be studied separately.
Because of a low tumor incidence and a very long latency period, the preventive effect of sunscreen use on melanoma cannot be directly demonstrated as it has been for squamous cell skin carcinomas.7 Some epidemiologic studies have raised suspicion about a positive association between sunscreen use and melanoma.14- 16 Several questions have thus been raised about the ability of sunscreens to play a role in skin cancer prevention. Their uneven protection against the sunlight spectrum and their incorrect use in terms of quantity and modalities of application may be improved with technical advances and information. However, if more potent sunscreens systematically induce longer exposure, they may increase skin cancer risk.
The hypothesis of longer exposures being induced by potent sunscreens has been substantiated by the results of 2 randomized controlled trials in students.21,22 We did not reproduce such results in our study. Although the difference in duration of sunbathing according to SPF may be viewed, in our study, as close to significance rather than nonsignificant (P = .06), it must be noted that the duration of sun exposure was lower in the high-SPF group, ie, inverse to the effect previously demonstrated. Compared with previous studies, methodologic differences in the design across these trials can be considered minor. Consecutive days were considered in this study vs scattered days during the whole summer in the previous trials; the duration of sun exposure was indirectly calculated in this study, whereas it was directly monitored by the participants in the previous trials. It is unlikely that these differences fully account for these discrepant results. Because this study included 4 times more participants than previous studies, it cannot be argued that a difference was left undetected because of a lack of power. The matter of population selection may be the most relevant explanation for the discrepancies in results. A group composed of young students, analyzed in previous studies, may have different behavior regarding sunscreen use than a group mainly composed of women in a family holiday setting. Young people are known to have riskier behavior and may tend to expose themselves up to the limit, even when the limit is pushed forward by high-SPF sunscreens. Conversely, in other populations, exposure duration may be more a matter of personal decision than a consequence of sunscreen characteristics. Whatever the interpretation of the discrepant results across studies, the global conclusion is that high SPF does not automatically induce longer exposure, although it may be true in specific subgroups.
In a skin cancer prevention strategy, behavioral measures (clothes, hat, and avoiding sun) must be preferred to sunscreens. However, for many societal reasons that we are not able to change rapidly, sunscreens will be used by most people as the predominant sun protection. From this pragmatic point of view, we must advise the population on how to make the best choice and the best use of sunscreens. This study highlights that the information about the quality of protection mentioned on the tube label does not seem to have, per se, important consequences in terms of quantity of sunscreen used and duration of sun exposure. However, the actual protection, represented by the SPF, does influence behavior. The SPF value, even masked, was a determinant of the amount of sunscreen applied. This link between SPF and amount of sunscreen used may be a consequence of an adaptive process: suboptimal protection given by lower-SPF sunscreen may be compensated for by using a greater amount of low-SPF sunscreen; alternatively, difficulties in getting an expected tan may be compensated for by using a smaller amount of high-SPF sunscreen. However, as the lower-SPF group did experience more sunburns than the higher-SPF group, this process should be considered nonoptimal for low-SPF sunscreens. Although better protection against UV-B provided by high SPF is not at all a direct proof of better protection against skin cancer, the overall data show that high-SPF sunscreens should be preferred to low-SPF ones.
Acknowledging that, although they represent suboptimal protection, sunscreens will be used by the population, future areas of research should focus on the different influences that sunscreen characteristics may have on sun behavior in different sociological groups, and on the best ways to improve intrinsic protective properties of sunscreens against the carcinogenic effect of UV light. While further data are awaited, public health recommendations should stress wearing clothing and hats and limiting midday exposures, but should not be reluctant to promote high-SPF rather than low-SPF sunscreens.
Correspondence: Alain Dupuy, MD, MPH, Service de Dermatologie (Pr Morel), Hôpital Saint Louis, AP-HP, 1 avenue Claude Vellefaux, 75010 Paris, France (email@example.com).
Accepted for Publication: February 1, 2005.
Funding/Support: This study was supported by a grant from Société Française de Dermatologie, Paris, and by Pierre Fabre Dermo-Cosmétique, Castanet-Tolosan, France, which provided sunscreens.
Acknowledgment: We thank Laurent Bouvet and Carolina Paz (students at the Laboratoire d’Investigation des Maladies de la Peau, Marseille, France), Leslie Noyelles, and Yang Yang for collecting and managing data; and the V V F Vacances, Paris, for welcoming the study in their resorts.
Financial Disclosure: None.
The Réseau d’Epidémiologie en Dermatologie (RED) includes Drs Dupuy and Grob and the following investigators: Sylvie Bastuji-Garin, MD, PhD, Hôpital Henri Mondor, Créteil, France; Philippe Bernard, MD, PhD, Hôpital Robert Debré, Reims, France; Eva Bourdon-Lanoy, MD, MPH, Hôpital Necker, Paris, France; Olivier Chosidow, MD, PhD, Hôpital Pitié-Salpétrière, Paris; Christian Derancourt, MD, Hôpital Robert Debré; Jean-Claude Guillaume, MD, Hôpital Pasteur, Colmar, France; Dominique Penso, MD, Issy-les-Moulineaux, France; Jean-Claude Roujeau, MD, Hôpital Henri Mondor; and Bruno Sassolas, Hôpital Morvan, Brest, France.