|
|
| |||||||||||||||||||||||||||||||
Background The incidence of and mortality from skin cancer are increasing in many countries. In view of the added concern about ozone depletion, many organizations are promoting the regular use of sunscreens to prevent skin cancer, despite the absence of evidence that these products have this effect. Solar (actinic) keratosis is a precursor of squamous-cell carcinoma of the skin.
Methods We conducted a randomized, controlled trial of the effect on solar keratoses of daily use of a broad-spectrum sunscreen cream with a sun-protection factor of 17 in 588 people 40 years of age or older in Australia during one summer (September 1991 to March 1992). The subjects applied either a sunscreen cream or the base cream minus the active ingredients of the sunscreen to the head, neck, forearms, and hands.
Results The mean number of solar keratoses increased by 1.0 per subject in the base-cream group and decreased by 0.6 in the sunscreen group (difference, 1.53; 95 percent confidence interval, 0.81 to 2.25). The sunscreen group had fewer new lesions (rate ratio, 0.62; 95 percent confidence interval, 0.54 to 0.71) and more remissions (odds ratio, 1.53; 95 percent confidence interval, 1.29 to 1.80) than the base-cream group. There was a dose-response relation: the amount of sunscreen cream used was related to both the development of new lesions and the remission of existing ones.
Conclusions Regular use of sunscreens prevents the development of solar keratoses and, by implication, possibly reduces the risk of skin cancer in the long term.
Methods
Persons 40 years of age or older who were living in Maryborough and surrounding districts in the state of Victoria, Australia, were invited by letter to attend a clinic for a free skin-cancer screening in September 1991. One member of a medical team consisting of one dermatologist, five dermatology trainees (registrars), and a local medical officer with an interest in skin cancer examined the head, neck, hands, and forearms of each respondent, all of whom were white. All solar keratoses were recorded on grid maps of these regions. Persons with 1 to 30 solar keratoses were invited to participate in the sunscreen study and, after they had given informed consent, were randomly assigned to the sunscreen group or the base-cream group. The randomization of the 588 subjects who enrolled in the study was stratified according to sex and self-rated skin type (burn only and never tan, burn first and then tan, or tan only and never burn).
The diagnosis of solar keratosis was based on the clinical appearance of the skin lesions, with a randomized subsample of subjects selected for biopsy of a lesion from the hands or forearms. If the solar keratosis chosen for biopsy was no longer present at follow-up, the disappearance was termed a remission; this occurred in some but not all lesions chosen for biopsy. Any lesion treated by a doctor during the course of the study was excluded from analysis.
The sunscreen was a cream containing 8 percent (wt/wt) 2-ethylhexyl p-methoxycinnamate and 2 percent (wt/wt) 4-tert-butyl-4-methoxy-4-dibenzoylmethane, rated according to Australian Standard 2604 (1986) as having a sun-protection factor (SPF) of 17 for ultraviolet B (resulting in a 94 percent reduction of radiation in the wavelengths 290 to 320 nm)11. It was also rated according to the same standard as a broad-spectrum filter for ultraviolet A (resulting in a 90 percent reduction of radiation in the wavelengths 320 to 360 nm). The base cream, minus the two active ingredients, was made up to have the same consistency as the sunscreen by adding 10 percent (wt/wt) mineral oil.
The subjects kept daily diaries in which they recorded the time of application of the cream. They were instructed to apply approximately 1.5 ml to the head and neck and the same amount to each forearm and hand once every morning, and to reapply if necessary during the day. All subjects were told not to rely entirely on the cream, but also to avoid the sun around the middle of the day and to wear hats and clothing where appropriate. They were asked not to use other sunscreen products during the study.
Follow-up examinations were performed on three occasions during the next seven months, the time of maximal daily sunlight in southern Australia (from spring until autumn); the last examination was in March 1992. At each examination, the total number of solar keratoses, remissions, and new lesions were recorded, the diaries were examined, and the bottles of cream were weighed. Any untoward reactions to the creams were recorded. Subjects who withdrew from the study were contacted to determine the reasons.
At the end of the study, the subjects were asked which cream they thought they had been given. They selected one of the following answers: "definitely the sunscreen," "probably the sunscreen," "don't know," "probably the moisturizer," or "definitely the moisturizer." (The term "moisturizer" was used to describe the base cream to the participants.) Of the 431 subjects who completed the study, 413 (96 percent) were examined by the same physician at the beginning and end of the study. The remaining 4 percent were seen by different examiners. The results of the final examination were recorded on a new grid map, so that the examiner was unaware of the results of previous examinations.
Three outcome variables were identified for each subject: the number of solar keratoses at the end of the trial and two components of this number -- the number of new lesions appearing during the study (incident lesions) and the number of remissions, expressed as a proportion of the number of base-line lesions. Each of these variables could be expected, a priori, to depend linearly on the number of base-line lesions.
Multiplicative Poisson models were adopted to analyze the total number of solar keratoses and the number of new solar keratoses. A binomial logistic model was assumed for remission probabilities. The logarithm of the number of each subject's lesions at base line was included in each of the three models, with a fixed coefficient of 1.0 (as an offset) in the case of the total number of lesions12. For all models, treatment effects were assessed for significance by comparing changes in model deviance with the chi-square distribution. Effect estimates were obtained from model coefficients, and two-sided confidence intervals were calculated from the unscaled matrix of coefficient covariances.
Dose-response relations were potentially confounded by other variables, since the subjects determined their own level of cream application. Therefore, differences between the dose-responses for sunscreen and base cream were compared after adjustment for sex and the number of lesions.
Results
A total of 588 white subjects with 1 to 30 solar keratoses were enrolled in the study. One hundred fifty-seven withdrew during the study or failed to attend the final examination, so that 431 subjects completed the study successfully. The reasons for withdrawal are shown in Table 1. There was no significant difference in the base-line demographic characteristics or study variables (including side effects of treatment) between those who completed the study and those who did not. Of the 431 subjects who completed the study, 180 were men and 251 were women (age range, 40 to 93 years; mean [±SD], 63 ±11). There were 221 subjects (89 men and 132 women) in the base-cream group and 210 subjects (91 men and 119 women) in the sunscreen group. The subjects in the sunscreen group initially had a total of 1856 solar keratoses (mean, 9 ±7 lesions per subject), and the subjects in the base-cream group had a total of 1642 solar keratoses (mean, 8 ±6 lesions per subject).
|
At the end of the study there was a mean (±SE) increase of 1.0 ±0.3 in the total number of solar keratoses per subject in the base-cream group as compared with a mean decrease of 0.6 ±0.3 lesions per subject in the sunscreen group (difference, 1.53; 95 percent confidence interval, 0.81 to 2.25). Poisson modeling showed that the relative change in the total number of solar keratoses in the sunscreen group, with the relative change in the base-cream group used as a reference, was 0.83 (95 percent confidence interval, 0.78 to 0.89). Figure 1 shows the estimated relative changes in the sunscreen and base-cream groups, according to sex. Although there was a sex-based difference in the change in the number of lesions during the study (the change was smaller in women than in men) (rate ratio, 0.87; 95 percent confidence interval, 0.80 to 0.94), the effect of treatment was not modified by age or skin type. The results are summarized in Table 2 according to the development of new lesions and the number of remissions.
|
|
At the completion of the study, 25 percent of the lesions present at base line had remitted in the sunscreen group, as compared with 18 percent in the base-cream group (odds ratio, 1.45; 95 percent confidence interval, 1.10 to 1.88). The mean percentage of lesions remitting throughout the study was 28 percent in the sunscreen group, as compared with 20 percent in the base-cream group (difference, 8 percent; 95 percent confidence interval, 2 to 13 percent). Logistic-regression analysis incorporating sex and skin type showed that the likelihood of remission was greater in the sunscreen group than in the base-cream group (odds ratio, 1.53; 95 percent confidence interval, 1.29 to 1.80).
There was no difference in the amount of cream used by the base-cream group and the sunscreen group (mean [±SD] weight, 666 ±284 vs. 663 ±299 g), although men used more cream than women (721 ±298 vs. 624 ±279 g). There was considerable variation between subjects in the amount of cream used, with many reporting that they required less than the recommended daily amount of 4.5 ml. Nevertheless, 81 percent of the participants reported applying the cream daily for at least 80 percent of the study period.
The sunscreen and base-cream groups differed in their dose-response relation; both the number of new lesions and the probability of remission were affected by the amount of cream used (chi-square = 6.3, P = 0.04 for new lesions; chi-square = 13.3, P = 0.001 for remissions). The number of new lesions was 23 percent of the number present at base line in the subjects who used less than 500 g of sunscreen, and it was 12 percent of base line among those who used more than 1000 g; there was no corresponding relation in the group using base cream (Figure 2). There was no difference in the probability of remission between treatment groups if the subjects used less than 500 g, but there was a clear difference when use exceeded 500 g.
|
Discussion
Our study shows that the regular use of sunscreens prevents the development of solar keratoses and hastens the remission of existing ones. In the long term the lesion of concern is skin cancer; however, solar keratoses are a major risk factor for skin cancer, and they are widely accepted as precursors of squamous-cell carcinoma. The use of similar precursors or risk factors for epithelial cancer (for example, the use of adenomatous polyposis of the colon to predict the effect of dietary interventions on the risk of carcinoma of the colon) has been accepted as a method for assessing the effect of an intervention on the prevention of cancer13.
Given the effectiveness of sunscreen preparations with high SPFs in preventing sunburn, the study subjects may have guessed which cream they were using. Although more subjects correctly identified their treatment than incorrectly identified it, 63 percent either chose the wrong cream or were unable or unprepared to offer an opinion. That both groups used very similar amounts of cream makes it likely that the proportion of subjects who knew which cream they were using was similar in both groups. Subjects who knew they were receiving the base cream might be expected to have used less cream, to have used a sunscreen in addition to the cream given, or to have increased other sun-protective behavior, all of which would tend to diminish the difference between the groups. A positive bias may have arisen if the subjects who knew they were using a sunscreen actually increased their sun-protective behavior -- an unlikely event.
The clinical diagnosis of solar keratosis in this study had a positive predictive value of 81 percent, at minimum, and it could have been as high as 94 percent if the seven lesions with doubtful histologic findings were all solar keratoses. There was no evidence that biopsy results differed between the two treatment groups. Misclassification of lesions does not affect the estimates of rate differences, but tends to reduce the power of the study by leading to a bias toward the null hypothesis for estimates of the rate ratio14. Thus, the effect measured in this study may be, if anything, an underestimate.
In a previous study, up to 25 percent of the solar keratoses remitted during a 12-month period10. We used these results to estimate sample sizes necessary to detect an effect during a 6-month intervention period, assuming linearity of exposure over a 12-month period. However, the study was conducted during the time of maximal sun exposure in Maryborough, Australia, so that at least 80 percent of the total ultraviolet radiation to be received in the 12-month period would have been received during that period15.
Reversal of epithelial dysplasia after cessation of exposure to a carcinogen has been observed in humans. For example, bronchial mucosal dysplasia in smokers has been reported to remit after cessation of smoking16. However, longitudinal, controlled use of sunscreens has not been proved to reduce the risk of skin cancer in humans, despite the fact that sunscreens have been available and recommended for many years. The rationale for the use of sunscreens in the absence of data in humans relates to a number of factors. Skin cancer is clearly related to exposure to ultraviolet radiation, and sunscreens absorb ultraviolet radiation17,18,19,20. Sunscreens reduce the number of nonmelanoma skin cancers induced by ultraviolet radiation in animals5,6. They also reduce erythema induced by ultraviolet radiation (sunburn) in humans, and this reduction is indeed the basis of the SPF efficacy rating21. Nevertheless, some people doubt the ability of sunscreens to prevent carcinogenesis in humans, particularly those sunscreens that selectively block ultraviolet-B radiation22. The effect demonstrated in this study occurred with the use of broad-spectrum ultraviolet-radiation filters. In these circumstances, it is not possible to say whether it was a reduction of ultraviolet A or B or a combination of both that was responsible for the effect.
This study should not be used by physicians to provide unrealistic expectations for sunscreen use by patients with solar keratoses. The study subjects were highly motivated, with two thirds having taken part in previous studies of skin cancer and more than a quarter having a history of skin cancer. Thus, the rates of compliance and response are unlikely to be as high in the whole population after the institution of a public-education program on the dangers of exposure to the sun.
Nevertheless, solar keratoses and nonmelanoma skin cancers are the commonest dysplasias and cancers in white persons in many countries throughout the world. The incidence of these lesions has been increasing recently, and concern about ozone depletion has led to predictions of even further increases in the future. The results of this study confirm the appropriateness of the approach to this problem being taken by an increasing number of public health organizations in many countries; that is, they should continue to recommend the use of sunscreens in addition to other methods of protection from sunlight (the use of shade, clothing, and hats) to reduce the risk of skin cancer.
Supported by grants from the Victorian Health Promotion Foundation, Melbourne; the Skin and Cancer Foundation, Sydney; the Skin and Psoriasis Foundation, Melbourne; the Lloyd Williams Trust, Maryborough; the Sydney Melanoma Foundation; and the Australasian College of Dermatologists.
We are indebted to Drs. P. Foley, C. Hadwen, D. McColl, F. Bruce, F. Watkins, S. Gilmore, and J. Horton, who examined the participants; to Dr. R. O'Keefe, Department of Anatomical Pathology, St. Vincent's Hospital, who examined the biopsies; to members of the Rotary, Mr. R. Were, the City of Maryborough, and the Maryborough Community Health Centre for assistance and use of their facilities; and to Jenny Greed, our nurse, for her dedicated effort throughout the study.
Source Information
From the Anti-Cancer Council of Victoria, Carlton, Victoria (S.C.T., D.J., R.M.); the Health and Community Services Department, Melbourne, Victoria (S.C.T.); and the University of Melbourne, Department of Medicine, St. Vincent's Hospital, Fitzroy, Victoria (R.M.) -- all in Australia.
Address reprint requests to Dr. Marks at the University of Melbourne, Department of Medicine, St. Vincent's Hospital, Fitzroy, Victoria 3065, Australia.
References
| |||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
