Application Patterns Among Participants Randomized to Daily Sunscreen Use in a Skin Cancer Prevention Trial FREE

KERATINOCYTE SKIN cancers are the most commonly occurring cancers in white populations and represent a potentially serious health burden in sun-exposed populations^1- 2 in terms of morbidity, cosmetic disfigurement, and costs to the health care system. For example, of the 10 most expensive cancers in Australia, keratinocye cancers are the most costly, accounting for approximately Austral $232.2 million per year.³ Primary prevention in susceptible people therefore remains a challenge.

It is now generally accepted that the main cause of keratinocyte cancers is exposure to solar UV radiation,⁴ and it is likely that sun exposure during adulthood may contribute to the progression of keratinocytic cells down the pathway to neoplasia. Reducing exposure of adult skin to solar UV radiation is thus a primary aim of public education strategies, with promotion of sunscreen use as a key supplementary measure to sun avoidance.⁵

Most public health campaigns widely promote the use of sunscreen on an ad hoc basis when specific activities in the sun are planned, yet many people are still not consistently using sunscreen as an adjunct measure for sun protection. In Victoria, Australia, where the impact of a skin cancer prevention campaign has been comprehensively evaluated, the number of questionnaire respondents reporting that they almost always or usually used maximum protection sunscreen "when outside in the sun" did not increase between 1991 and 1997, but remained constant at approximately 45%.⁶ When a sample from the same population was asked to specify their sunscreen use when outdoors the previous day,⁷ the proportion of sunscreen users was lower, suggesting that, when asked to generalize their sun protection behavior, people failed to consider sun exposure occurring during routine daily activities. It is likely that in sunny places, such as Australia and much of the United States and Europe in summer, a high proportion of ultraviolet exposure is accumulated "incidentally" during routine outdoor activities rather than as a result of intentional exposure. Strategies to reduce such incidental exposures may represent a promising future for skin cancer control.

This was the rationale underpinning the design of the Nambour Skin Cancer Prevention Trial, which demonstrated the effectiveness of an intervention specifying the daily application of sunscreen to exposed body sites for the reduction of squamous cell carcinoma tumors.⁵ We have reported previously that the effectiveness of this community-based intervention did not appear to be due to measurable changes in sun protection behaviors apart from sunscreen application and was most likely attributable to changes in patterns of sunscreen use.⁸ The aim of the study reported herein was to further explore the implementation patterns of the intervention group, particularly the weight of sunscreen used, the application frequency, and the amount of sunscreen applied per unit of skin surface area (hereafter referred to as quantity), among the people randomized to sunscreen use. Better understanding of the sunscreen application patterns among the intervention group in this successfully implemented trial will help inform future public health skin cancer prevention strategies among similarly exposed populations.

The conduct of the Nambour Trial was approved by the Queensland Institute of Medical Research Ethics Committee, and written informed consent was obtained from all participants. The design and methods have been described in detail previously.⁹ Briefly, the Nambour Trial was a 2 × 2 factorial field trial to test the interventions of daily sunscreen application and beta carotene supplementation for the prevention of skin cancer over 5 years. Participants were residents of Nambour (a subtropical town of around 8000 people lying 100 km north of Queensland's capital, Brisbane, Australia). In 1992, 1621 residents randomly selected from those who had participated in a previous skin cancer survey were randomly assigned to 1 of 4 groups (beta carotene tablets and sunscreen, placebo tablets and sunscreen, beta carotene only, or placebo only). The original skin cancer survey participants had been randomly selected from the electoral roll (enrollment is compulsory). The mean age of participants in the present trial was 49 years, and 44% were male. Nineteen percent had worked in mainly outdoor occupations, and approximately 27% reported having had a previous skin cancer.

The sunscreen intervention required daily application of a supplied standard sun protection factor 15+ sunscreen to the face, rest of the head and neck, hands, and arms. The sunscreen was a water-resistant, oil-in-water, cream formulation containing octyl methoxycinnamate, oxybenzone, and butyl methoxydibenzoylmethane. It was supplied in standard 250-mL bottles with a flip-open cap covering an opening of approximately 5 mm. Participants were instructed to apply sunscreen every day, regardless of planned activities or weather conditions. The weight or quantity of sunscreen to be applied at each application was not specified beyond "adequate." The weight of sunscreen used by each participant in the sunscreen intervention group was calculated, in addition to the self-reported frequency of application.

Every 3 to 4 months during the 5 years of the trial, beginning in February 1992 and ending in August 1996, participants were required to attend designated "distribution" weekends at a central location, when used sunscreen bottles were collected and new bottles supplied as required. Research assistants, using carefully standardized scales, weighed returned bottles. Any participants who failed to attend distribution weekends were contacted, provided with a fresh bottle, and asked to estimate the proportion of sunscreen remaining in each bottle (the used bottle was to be weighed later, when possible). Some used bottles were not subsequently weighed, and in such cases the estimated proportions remaining were converted to weights of sunscreen and marked as proxy measurements. A total of 1621 participants were initially enrolled in the trial, 812 of whom were in the sunscreen intervention group.⁵ Information regarding the weight of sunscreen used was available for at least one distribution period for 764 participants in the intervention group (94%).

To calculate the average weight of sunscreen used per day for the intervals between sunscreen distributions, the weight of sunscreen remaining in the bottle was subtracted from the total weight of sunscreen supplied, and the difference was divided by the number of days in the interval between supply and return. Because the dates of contact with each participant were not the same, the average daily rates of use were calculated for 18 standard 3-month periods, corresponding to seasons, to enable analysis of changes in the weight of sunscreen used by season and over time. The average daily sunscreen used over the entire trial period was calculated as the total weight of sunscreen supplied minus the total returned, divided by the total number of days over which this sunscreen was used.

A questionnaire that asked a broad range of questions regarding attitudes toward sun protection and sun-protective behaviors was self-administered in the final year of the trial. With regard to sunscreen, participants specifically reported their usual frequency of sunscreen application (days per week and times per day) separately for the face, rest of the head and neck, hands and arms, and other body sites.

Statistical analysis was mainly performed with SAS software.¹⁰ Measurements calculated and compared were the frequency and weight of sunscreen used during the course of the trial. The analysis was predominantly descriptive; weighted κ (using default SAS weights) was used to estimate the differences in sunscreen application frequency to different body sites. To assess associations between personal characteristics and the weekly frequency of sunscreen use, we calculated adjusted prevalence ratios (PRs) by using Poisson regression. Possible predictors were divided into groups relating to exposure, complexion, and sun-related skin damage (eg, nuchal elastosis: degeneration of the dermal elastic tissue of the neck, clinically assessed by dermatologists). Variables with the largest effect size within each group were selected first, and the remaining variables were then assessed for their contribution to the fit of the model.

The body surface area of participants was calculated from the following formula: (height×weight)/3600.¹¹ If height or weight was missing, the sex-specific mean was assigned for the calculation of body surface area. The exposed parts of the head and the neck together were considered to constitute 4%, and the hands and arms, 8.5%, of the total body surface area.¹² The quantity of sunscreen applied was estimated by first calculating the body surface area in square centimeters to which sunscreen was applied in a typical week, based on the number of days and times per day sunscreen was recorded in the questionnaire as being applied to the required sites. The total sunscreen applied in a typical week was then divided by the estimated surface area of application.

To assess associations with the average quantity of sunscreen used, the 18 people who did not use sunscreen were excluded and the distribution among users was normalized by a logarithmic transformation. Multivariate linear modeling was used to determine independent associations with sunscreen quantity. The coefficients from the final model were exponentiated to calculate the ratio of the quantity of sunscreen used in the relevant category relative to the reference category.

Among the 764 participants in the sunscreen intervention group for whom sunscreen usage during the course of the trial was available, the median daily weight of sunscreen applied on average during the duration of the trial was 1.5 g/d (range, 0-7.4 g/d). The rate of sunscreen application varied across the 18 distribution periods of the trial according to season, with generally less sunscreen being used in the 3 winter months than in the 3 summer months. Median sunscreen use in winter was 1.39 g/d (interquartile range, 0.84-1.99 g/d), and in summer, 1.59 g/d (interquartile range, 1.01-2.36 g/d) (P<.001 for winter vs summer difference). In addition, sunscreen use decreased as the trial progressed; the median daily weight of sunscreen application each year from 1992 to 1996 was 1.67, 1.61, 1.46, 1.41, and 1.22 g, respectively.

Questionnaires were sent to 1329 participants, including 96 participants who had withdrawn but were able to be located. Seven hundred thirty-four participants in the sunscreen intervention group received the questionnaire, 659 (90%) of whom completed the questions regarding the frequency of sunscreen use. Most reported applying sunscreen to the face, arms, and hands the same number of days per week (weighted κ = 0.85). Applications to the face have therefore been used as indicative of the frequency of sunscreen application to the sites specified by the intervention regimen. A total of 356 participants (56%) who completed the relevant question reported applying sunscreen to the face at least 5 days per week, while 27% used sunscreen infrequently at 2 or fewer days per week (Table 1). Sunscreen application to other sites not specified in the study protocol was less frequent, with almost half of participants reporting no regular application. Women (P = .04) and participants older than 60 years (P<.001) were more likely to report application to these sites than men and those younger than 60 years. These results are in stark contrast to the control group, 74% of whom applied sunscreen to the specified sites and 92% to unspecified sites on 2 or fewer days per week.⁵ Most people who regularly applied sunscreen in both the intervention and control groups applied it only once per day (Table 2).

There was a strong association between the self-reported frequency with which sunscreen was applied to the face or hands and the weight of sunscreen used during the distribution period during which the questionnaire was completed (Spearman rank ρ = 0.52, P<.001). Nevertheless, within each of the self-reported categories there was a considerable range in the weight of sunscreen used per day (Table 3), and this was not altered when the analysis was stratified by the number of applications per day or applications to other parts of the body. We therefore considered that it reflected true variation in the thickness of sunscreen application.

For the 595 participants for whom this information was available, the quantity of sunscreen applied ranged from 0 mg/cm² (indicating that no sunscreen was used) to 6.3 mg/cm², with a mean of 0.99 mg/cm² (95% confidence interval [CI], 0.92-1.06 mg/cm²) and median of 0.79 mg/cm² (interquartile range, 0.46-1.22 mg/cm²). Excluding the 7 people who did not use sunscreen did not materially alter these estimates. A total of 324 participants reportedly applied sunscreen to sites other than the face and hands at least once per day, possibly resulting in an overestimation of the amount of sunscreen applied to the required sites. Excluding these people from the analysis resulted in a mean of 1.03 mg/cm² (95% CI, 0.92-1.14 mg/cm²) and median of 0.79 mg/cm² (interquartile range, 0.46-1.25 mg/cm²).

Given the relatively high adherence with the requested regimen of daily sunscreen application, we explored the predictors of nonadherence, ie, using sunscreen on 2 or fewer days per week. In a model containing age, sex, time outside, and skin cancer diagnosed during the trial, the strongest predictor of sunscreen use frequency was the usual amount of time spent outside. People who reported being outside 4 or more hours per day were two thirds less likely to be nonadherent than those who spent less than an hour per day outside (PR, 0.33; 95% CI, 0.20-0.59), with those spending between 1 and 4 hours per day outside being approximately half as likely to be poor adherers (PR, 0.57; 95% CI, 0.41-0.80). Men were approximately twice as likely as women to be nonadherent (PR, 1.77; 95% CI, 1.29-2.44), but age was not associated with the weekly frequency of use (P = .69). Diagnosis of a skin cancer before the trial and during the trial were highly correlated and were added separately to the model containing sex, age, and time outside. In both cases, those who had been diagnosed as having skin cancer were approximately 50% less likely than others to be low sunscreen users (skin cancer during trial PR, 0.48; 95% CI, 0.29-0.80). People with high levels of actinic skin damage such as a high degree of facial telangiectasia and the presence of more than 10 solar keratoses on the skin also used sunscreen more frequently than those with less solar-damaged skin. Skin type, assessed by skin color (P = .99), tanning ability (P = .77), and hair (P = .65) or eye (P = .90) color were not associated with the weekly frequency of sunscreen application in either univariate or adjusted models. Similarly, the number of sunburns during the trial (P = .70) or reported lifetime burns (P = .79) did not contribute to the fit of the model.

The usual amount of time spent outdoors was also associated with the number of sunscreen applications per day, with those spending more than 4 hours outside being 50% more likely to apply sunscreen more than once (Table 4). People who reported having 2 or more sunburns during the 5 years of the trial were approximately twice as likely to report more frequent daily sunscreen applications than those who had not burned, even after adjustment for skin type and the amount of time spent outdoors (PR, 1.85; 95% CI, 1.21-2.83). There was a somewhat paradoxical association between skin type and daily application frequency, with people who rarely burned reporting more frequent sunscreen applications than those who were prone to sunburn.

Associations with the quantity of sunscreen that people applied were modified by their frequency of sunscreen use. Among those who used sunscreen frequently, men applied 20% more sunscreen per square centimeter than women, but the same association was not found among less frequent users (Table 5). In both groups, those aged between 40 and 59 years used the least sunscreen. The quantity of sunscreen frequent users applied was inversely related to duration outdoors, whereas people who used sunscreen less frequently applied more the more time they spent outdoors. A number of indicators of sun-induced damage to the skin were examined, including solar keratoses, solar elastosis, telangiectasia, and a history of skin cancer, with solar keratoses being the only indicator of skin damage that was positively associated with the quantity of sunscreen applied. People with a history of sunburn during the trial applied more sunscreen to their skin than those with no sunburn history.

Of the 393 participants in the intervention group who completed the questionnaire and reported less than daily sunscreen use, 353 (90%) completed the question concerning the reason for not using sunscreen on regular occasions. Almost 50% of participants stated that they did not think sunscreen application was necessary given the weather conditions or their planned activities. Of these respondents, 45% reported that they generally spent almost no time outdoors during the day, while 14% reported that they spent greater than 50% of their time outdoors on average. All 5 participants who reported that they did not use sunscreen because they wanted to obtain a tan were aged between 20 and 39 years, and 4 of the 5 were men.

Studies examining sunscreen application behavior have been numerous, but most research has been conducted in selected people or in restricted settings such as at the beach,^13- 14 in schools,^15- 16 or in outdoor work.¹⁷ Most studies among randomly selected adult populations ask specifically about the use of sunscreen "when outside in the sun" (eg, Robinson et al¹⁸) despite the fact that this may overestimate the proportion of time outside that the skin is protected by sunscreen.^6- 7 The sunscreen intervention of the Nambour Trial specifically required participants to apply sunscreen every day, and more than half of the community participants were able to adhere to this request or to apply it at least 5 days a week. However, almost a quarter of the respondents to this question reported that they applied sunscreen on 2 or fewer days per week. Further investigation showed that of these "low-frequency" sunscreen users, 48% reported that they spent almost no time outside on a typical day, but a further 40% spent at least some time out of doors and 11% spent more than 50% of their time outside. When asked the reason for their less frequent sunscreen application, many participants reported that they "did not believe it to be necessary given their planned activities"—presumably their outdoor activities entailed incidental rather than intentional sun exposure. Similarly, among a group of students holidaying in Europe, intentional sunbathing was associated with the total sunscreen used while on holidays, but the total duration of outdoor exposure was not.¹⁹ Addressing the apparent misconception that no protection is required during frequent incidental exposures to the sun may be one way for future primary prevention campaigns among highly exposed adults to achieve better control of skin cancer.

Although the intention-to-treat analysis of the Nambour Trial⁵ showed that randomization to daily sunscreen use did not increase the risk of skin cancer and was in fact protective for squamous cell carcinoma, those within the intervention group who reported a high frequency of sunscreen application were more likely to have had a skin cancer diagnosed during the trial. This paradoxical association between sunscreen use and nonmelanoma skin cancer or actinic skin damage has been reported previously^20- 21 and is almost certainly due to uncontrollable "confounding by indication" resulting from the highly complex interrelationship between phenotype, exposure factors, and sunscreen use.²¹

The positive relationship found between sunscreen use and sunburns in our study is consistent with results of previous reports. The SunSmart evaluation in Victoria, Australia, reported that 46% of people who almost always used sunscreen had been sunburned in the previous summer compared with only 31% of those who rarely used sunscreen.⁶ Similarly, there was a nonsignificant increase in risk of burning among children who reported frequent sunscreen use.²² A Queensland study found that, despite increased sunscreen use between 1988 and 1991, the number of people reporting a sunburn the previous weekend had not decreased, although the severity of the sunburns was lower.²³ In most cases the site and severity of the sunburn and the type of exposure during which the burn occurred have not been investigated. Unavoidable confounding by risk of burning is likely to partially explain the association.

The average weight of sunscreen used during the entire 5 years of the trial varied widely in this population, with a median of 1.5 g/d. Among self-reported higher users, however, this increased to a median of 2 g/d. There have been a number of previous longitudinal studies of sunscreen use,^19,24- 26 only one of which reported comparable data.²⁶ Participants in the Australian trial evaluating sunscreen in the prevention of solar keratoses²⁶ used a mean of 666 g of sunscreen during a 7-month period, approximating a mean of 3.2 g/d, higher than our mean of 1.65 g/d. This may have occurred because their study was of shorter duration and/or because their participants were volunteers older than 40 years with at least 1 solar keratosis at baseline. The decreasing weight of sunscreen used during the 5 years of our trial may have resulted from decreased frequency or thickness of application. More frequent administration of questionnaires through the trial would, in theory, have detected variation in sunscreen application frequency. However, the accuracy of the reports might have been compromised by continued involvement in the trial.

This is only the second study to examine the quantity of sunscreen used per unit of skin surface area in a relatively unselected population of adults. Evidence from the previous study¹⁹ is more difficult to generalize to adults in community settings because the data pertained to students aged 18 to 24 years who were on holidays. Other studies have been among volunteers,^27- 28 nude sunbathers,²⁹ and people with photosensitivity disorders.³⁰ Two studies provided the sunscreens in smaller containers than usual,³⁰ and this has been shown to affect the amount of sunscreen applied.³¹ In all but 3 studies,^19,26,29 participants were observed by investigators during their sunscreen application. The estimated amount of sunscreen applied has ranged from approximately 0.55 mg/cm^219,29- 30 to 1.0 mg/cm², although this has varied with both the type of sunscreen being applied²⁷ and the site of application.³⁰ We recognize the limitations in our study that result from estimating the amount of sunscreen applied by community participants. These include disregarding variations in the weight of sunscreen used in favor of average weight per day and the assumption that estimated weight and frequency were correct and that the reported frequency of application did not fluctuate throughout the trial. We also lacked knowledge about the precise surface area of application and about variations in amount used on different body sites. Nevertheless, our estimation falls in the range of usage estimates from similar studies. It is possible that our estimates of amount used are higher than the recently published longitudinal study among students¹⁹ because of the greater awareness of skin cancer risk among the Australian population. Our study confirms previous reports that most sunscreen users are not achieving the level of protection of the sun protection factor stated on the bottle.^19,28- 29 A greater educational focus on the adequacy of each application may partly resolve this problem, although it is unlikely that most people will double the amount of sunscreen that they naturally apply to their skin. Given the technical difficulties and implications for education that would arise if sunscreen testing standards required a more realistic testing thickness, we agree with Autier et al¹⁹ and Diffey³² that more informative labeling of sunscreen bottles is long overdue.

This is the most complete set of prospective data on actual sunscreen consumption and use patterns during a 5-year period among a community of adults whose demographic characteristics and sun exposure are broadly generalizable. Clearly the impetus to adopt sun-safe practices was higher in this trial population than would be expected in the wider community. Nevertheless, these results indicate that educational campaigns that specifically recommend daily sunscreen application to target incidental exposures might significantly contribute to the improvement of sun protection behavior, with continual advice needed to promote maintenance of the behavior.

In April 2000, a group of experts convened by the International Agency for Research on Cancer concluded that "sunscreens probably prevent squamous cell carcinoma of the skin when used mainly during unintentional sun exposure."³³ On the basis of current recommendations to use sunscreen on an ad hoc basis, a national skin cancer primary prevention campaign in a high-risk population such as in Australia is thought to be economically worthwhile.¹ Encouraging a daily sunscreen application strategy is a highly effective and achievable method of ensuring protection during unintentional sun exposure; reducing intentional exposure will require a different approach.

Accepted for publication February 12, 2002.

This study was supported by the Public Health Research and Development Committee of the National Health and Medical Research Council of Australia, Canberra, and the Department of Health and Aging, Canberra. Sunscreen was supplied by Ross Cosmetics Aust Pty Ltd, Melbourne, Australia, and Woolworths Ltd, Sydney, Australia.

We thank Anny Fourtanier and Romano Mascotto, PhD, for their valuable comments on this article.

Corresponding author and reprints: Rachel Neale, PhD, Queensland Institute of Medical Research, PO Royal Brisbane Hospital, Brisbane 4029, Queensland, Australia (e-mail: rachelN@qimr.edu.au).