- This topic has 21 replies, 6 voices, and was last updated 11 years, 9 months ago by
.
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
fyi
http://jco.ascopubs.org/content/29/3/257.full
- © 2010 by American Society of Clinical Oncology
Reduced Melanoma After Regular Sunscreen Use: Randomized Trial Follow-Up
+ Author Affiliations
From the Queensland Institute of Medical Research; University of Queensland; and Princess Alexandra Hospital, Brisbane, Australia.
- Corresponding author: Adèle Green, Queensland Institute of Medical Research, PO Royal Brisbane Hospital, Queensland 4029, Australia; e-mail: [email protected].
Abstract
Purpose Regular sunscreen use prevents cutaneous squamous cell carcinoma long term, but the effect on melanoma is highly controversial. We evaluated whether long-term application of sunscreen decreases risk of cutaneous melanoma.
Participants and Methods In 1992, 1,621 randomly selected residents of Nambour, a township in Queensland, Australia, age 25 to 75 years, were randomly assigned to daily or discretionary sunscreen application to head and arms in combination with 30 mg beta carotene or placebo supplements until 1996. Participants were observed until 2006 with questionnaires and/or through pathology laboratories and the cancer registry to ascertain primary melanoma occurrence.
Results Ten years after trial cessation, 11 new primary melanomas had been identified in the daily sunscreen group, and 22 had been identified in the discretionary group, which represented a reduction of the observed rate in those randomly assigned to daily sunscreen use (hazard ratio [HR], 0.50; 95% CI, 0.24 to 1.02; P = .051). The reduction in invasive melanomas was substantial (n = 3 in active v 11 in control group; HR, 0.27; 95% CI, 0.08 to 0.97) compared with that for preinvasive melanomas (HR, 0.73; 95% CI, 0.29 to 1.81).
Conclusion Melanoma may be preventable by regular sunscreen use in adults.
INTRODUCTION
The need for more effective prevention of melanoma is recognized around the world as climbing incidence and high mortality in white populations persist.1–3 In the United States, approximately 68,700 new melanoma occurrences and more than 8,600 deaths were expected to occur in 2009.4 Exposure to solar ultraviolet (UV) radiation is the only established modifiable cause of melanoma.5,6
Despite the known etiologic role of sun exposure, the question regarding sunscreen use to prevent melanoma remains open7 and controversial.8 Although case-control and cohort studies of sunscreen use and melanoma risk abound, their findings have been uninformative.9–11 Nonrandomized studies of the melanoma-sunscreen association are unable to distinguish the main determinants of sunscreen use from those of melanoma, because they are the same—namely, susceptibility to sunburn, high occupational or recreational sun exposure, and family history.11 The only relevant evidence comes from a randomized trial of sunscreen application in Canadian children conducted from 1993 to 1996 that showed a small reduction in new melanocytic nevi,12 the strongest predictors of melanoma, in children allocated to a sunscreen arm, especially if the children had freckles.
Here, we present new evidence from the follow-up of a community-based, pragmatic trial of sunscreen to prevent skin cancer in Queensland, Australia.13 By primarily aiming to evaluate prevention of basal cell carcinoma (BCC) and squamous cell carcinomas (SCC) of the skin and their precursors,13,14 we have shown prolonged prevention of SCC by the sunscreen intervention.15 Now, this trial and its 10-year aftermath have given us the unique opportunity to examine melanoma as a secondary trial end point by using unconfounded evidence of the long-term effectiveness of sunscreen in reducing melanoma incidence. The hypothesis was that regular sunscreen use by white adults prevents the occurrence of primary cutaneous melanoma,11,16 with a possible latent effect of up to 10 years.
METHODS
Study Design
In 1992, 1,621 residents of the Queensland township of Nambour who were ascertained in 1986 at ages 20 to 69 years for a skin cancer prevalence survey were enlisted in the Nambour Skin Cancer Prevention Trial. Original survey participants had been randomly sampled from the Nambour electoral roll (enrollment is compulsory by law),14 and those who participated in the trial were representative of the original sample.17 Trial participants were randomly assigned individually by using a computer-generated, randomized list (without stratification or blocking). The 812 participants randomly assigned to sunscreen intervention were given a free, unlimited supply of broad-spectrum sunscreen containing 8% (by weight) 2-ethyl hexyl-p-methoxycinnamate and 2% (by weight) 4-tert-butyl-4′ methoxy-4-dibenzoylmethane and with a sun protection factor (SPF) of 16. They were asked to apply it to head, neck, arms, and hands every morning (and reapplication was advised after heavy sweating, bathing, or long sun exposure). The 809 participants randomly assigned to the comparison group continued using sunscreen of any SPF at their usual, discretionary frequency, which included no use.13 Allocation of a placebo sunscreen to the control group was unethical, given the subtropical location. According to a 2 × 2 factorial design, 820 participants (n = 404 and n = 416 in daily and discretionary sunscreen groups, respectively) were also independently randomly assigned to 30 mg beta carotene, and 801 participants (n = 408 and n = 393 in daily and discretionary sunscreen groups, respectively) were randomly assigned to placebo supplements, because beta carotene potentially could counteract the oxidative damage to DNA involved in solar UV carcinogenesis.13
Compliance with sunscreen treatment was assessed by measured weights of returned sunscreen bottles and by questionnaires asking average frequency of use in a normal week. Intake of supplements was assessed by remaining tablet counts.13 Dermatologists unaware of treatment allocations conducted skin examinations of participants at baseline (March 1992), including assessment of number of nevi on the back; midway (1994); and at trial end (August 1996). Participants diagnosed with suspected melanomas were referred to their physicians for immediate management. All skin cancers, including melanomas diagnosed between surveys, were ascertained quarterly with histologic confirmation of any reported. Information about risk factors for skin cancers, such as skin color, outdoor behavior, and sunburn history, was obtained at baseline, and information on sun exposure and protection was updated throughout the trial. Ethical approval was obtained from institutional ethics committees, and study participants gave their written informed consent.
Follow-Up
After the scheduled trial completion in 1996, 1,339 participants (82%, including 14 participants who moved outside Queensland) agreed to take part in the follow-up study actively (Fig 1); they completed biannual or annual questionnaires about all new skin cancers, including melanomas. In addition, they reported average time outdoors on weekdays and weekends in the previous 6 or 12 months and average sunscreen use (although no sunscreen was supplied after 1996). Participants who withdrew from active trial participation or active follow-up were asked to continue with ongoing passive monitoring of skin cancers through their medical records.14 Investigators thus obtained notification of all melanomas diagnosed by regional pathology laboratories in active and passive participants. Finally, we cross-checked for any melanomas diagnosed between 1992 and 2006 through a search of the Queensland Cancer Registry (because melanoma registration is compulsory, is a particularly high priority, and is considered virtually complete); however, no new melanomas were uncovered in the Cancer Registry checks that had not already been ascertained.
Fig 1.Nambour Skin Cancer Prevention Trial follow-up profile.
Review of each diagnosed melanoma was undertaken by two expert dermatopathologists who were unaware of sunscreen allocation, and reviews were based on available pathology slides. The histologic diagnosis of melanomas of any type, both in situ and invasive, was based on a constellation of features developed during several decades.18 Melanoma in situ, lentigo maligna type, was diagnosed by using defined criteria.19 Invasion was defined and classified according to Clark's level of invasion: Level 1 (in situ); Level 2, tumor in the papillary dermis; Level 3, tumor filling the papillary dermis and extending to the papillary dermis/reticular dermis interface; Level 4, tumor in reticular dermis; and Level 5, tumor in fat.20 Classification by histologic type was not undertaken, although no melanoma types were excluded.
Statistical Analysis
When we formally assessed the long-term trial results for BCC and SCC to the end of 2004, one of us (G.W.) also carried out a preliminary intention-to-treat analysis on the basis of accumulated but unreviewed melanoma reports for all body sites. The decision then was made to evaluate melanoma occurrence classified according to invasiveness up to December 2006, because 15 years of follow-up (calendar time) was deemed sufficient to detect an effect of sunscreen, if present. On the basis of the observed rate of melanoma in the control group to 2004, the power was estimated to be 66%, and 50%, for detecting hazard ratios of 0.3 and 0.4, respectively, with a two-sided α of .05. As for the other skin cancer end points,13 melanomas diagnosed in the first year of intervention were excluded a priori, because their development was unlikely to have been affected by the introduced sunscreen treatment. Cox proportional hazards regression, with the sunscreen and beta carotene interventions as two main effects, was used to examine treatment effects in relation to primary melanoma occurrence with incorporation of lead time. Individual effects of sunscreen and beta carotene were tested by using likelihood ratio tests. Subgroup analyses were performed to assess consistency of effect according to age, sex, phenotype, sun exposure, and history of skin cancer by using Cox regression and by incorporating a subgroup interaction term to detect heterogeneity of effects.
During trial follow-up, mean hours each day spent outdoors on weekdays and on weekends were calculated, and sun protection habits were assessed from questionnaire responses. Sun exposure and protection durations were compared across the two sunscreen treatment groups by using a two-sample t test. All reported P values were two sided.
RESULTS
Balance was achieved with respect to established risk factors for skin cancer and melanoma among the Nambour trial participants randomly assigned to sunscreen or control in 1992 (Table 1). Compliance with sunscreen treatment, assessed by average of reported frequencies of application, measured weights of returned sunscreen bottles, and diaries, was approximately 75%,13 and 25% of the intervention group applied sunscreen to trunk and/or lower limbs as well as to the intervention sites.21 The majority of participants in the control group either did not apply sunscreen (38%) or applied it once or twice a week at most (35%), and 8% applied it to nonintervention sites.21 Compliance was approximately 70% for beta carotene and placebo supplementation.13
View this table:Table 1.Demographic and Clinical Characteristics of Participants at Baseline in 1992 According to Sunscreen Allocation
By the end of 2006, 846 people (52%) were actively completing questionnaires, 600 (37%) were passive participants, and 173 (11%) had died (n = 87, sunscreen group; n = 86, controls; n = 71, beta carotene group; n = 102, placebo controls), including one person who died as a result of melanoma diagnosed in 1978. One person from each sunscreen treatment group had withdrawn their permission for passive follow-up, in 2002 and 2001, respectively, and data for both were censored accordingly. There was no significant difference in mode of follow-up in relation to sunscreen allocation: duration of active follow-up was 14.3 person-years (95% CI, 14.1 to 14.4 person-years) versus 14.2 person-years (95% CI, 14.0 to 14.3 person-years), and active response rates were 94.2% and 94.3% in each group, respectively. On the basis of reports of active participants, 25% of those randomly assigned to daily sunscreen continued to use sunscreen on a regular basis after the trial22 compared with 18% of the nonintervention group (P = .004).15 By 2001, less than 3% of either supplement group took beta carotene supplements.
In the almost 15 years from commencement of the trial in March 1992 until the end of follow-up in December 2006, 36 of the 1,621 trial participants developed first primary melanomas (n = 22, in situ; n = 14, invasive; none metastatic), and one person in each trial arm developed two primary melanomas. Three people (n = 1, intervention group; n = 2, control) who had melanomas diagnosed in 1992 were excluded a priori. In the remainder of the trial from 1993 to 1996, two participants in the daily sunscreen group and seven in the discretionary group were diagnosed with melanoma (Fig 2). From trial cessation until the end of 2006, nine more participants allocated to daily sunscreen and 15 allocated to discretionary use were diagnosed with incident melanomas (Fig 2). In all, 11 trial participants in the sunscreen intervention and 22 in the control group (Table 2) were newly diagnosed with primary melanoma between 1993 and 2006. Risk of melanoma overall was reduced in those randomly assigned to daily sunscreen compared with discretionary use (hazard ratio [HR], 0.50; 95% CI, 0.24 to 1.02; P = .051), although the result was of borderline statistical significance. Invasive melanoma was reduced by 73% in the daily sunscreen group (HR, 0.27; 95% CI, 0.08 to 0.97; P = .045; Table 2); average thickness was 0.53 mm in the sunscreen group and 1.2 mm in controls (P = .08) on the basis of the original pathology reports (except for one level 2 melanoma in the control group for which thickness measurement was unavailable). There were no significant differences between intervention and control arms with respect to either in situ melanomas (8 v 11, respectively; HR, 0.73; 95% CI, 0.29 to 1.81) or melanomas on prescribed application sites (HR, 0.46; 95% CI, 0.17 to 1.20). When a multivariate proportional hazards regression was carried out that included sex, skin type, numbers of nevi, previous history of skin cancer, and sun exposure along with the two treatment categories, the overall effect estimate for sunscreen varied little (HR, 0.49; 95% CI, 0.24 to 1.02). Regarding outcome according to beta carotene randomization, 16 and 17 melanomas occurred in those taking active and placebo supplements, respectively (HR, 0.89; 95% CI, 0.45 to 1.76).
Fig 2.Occurrence of first primary melanoma by level of invasion and anatomic site in the two sunscreen treatment groups.
View this table:Table 2.First Primary Melanomas During 1993-2006 According to Randomized Sunscreen Intervention During 1992-1996 and Risk of Melanoma
No significant interactions with any baseline characteristics were found (Fig 3). Sun exposure was similar between the daily and discretionary sunscreen groups during the trial (79% and 77%, respectively, spent less than 50% of weekend time outdoors23) or after the trial (3.8 and 3.9 hours each day, respectively, spent outdoors on weekdays [P = .46]; 4.6 and 4.7 hours each day, respectively, spent outdoors on weekends [P = .79]). Use of sun protection measures other than sunscreen was similar during the trial (approximately 60% of both groups usually sought shade; around 75% usually wore a hat) and after the trial (at midpoint of follow-up, 40% and 67% usually sought shade and wore a hat in the sun, respectively).
Fig 3.Effect of sunscreen intervention on melanoma according to baseline characteristics. Hazard ratios are for melanoma in a comparison of the sunscreen intervention and control groups. Hazard ratios, 95% CIs, and P values were calculated by using Cox regression that incorporated a subgroup interaction term to detect heterogeneity of effects.
DISCUSSION
Long-term follow-up of this randomized trial showed that, among adults age 25 to 75 years, regular application of SPF 15+ sunscreen in a 5-year period appeared to reduce the incidence of new primary melanomas for up to 10 years after trial cessation. A protective effect was also evident for invasive melanoma, which showed a 73% decrease in those randomly assigned to daily sunscreen after approximately 15 years of follow-up. This result for the invasive subgroup was an exploratory finding, however, and should be interpreted cautiously. The apparent decrease in melanoma across all body sites, not only prescribed application sites, likely reflected the ongoing tendency of more participants in the daily than the discretionary sunscreen group to apply sunscreen regularly to the trunk and lower limbs.21,22 Intensity of application also tended to be higher in the intervention group.21
There was no evidence to suggest that the observed difference could be explained by a difference in sun exposure in the intervention group relative to the control group, because outdoor behavior was similar during23 and after the trial. Similarly, there were no differences in active completion of follow-up questionnaires with self-reports (later validated) of all new skin cancers, including melanomas. Without reference to treatment groups, we attempted to capture all melanomas diagnosed between 1993 and 2006 among passive trial participants (except for two who withdrew consent for follow-up) from the Queensland Cancer Registry as well as from the pathology laboratories. Ultimately, however, the final diagnosis of each melanoma reported during or after the trial was determined by the two reviewing dermatopathologists (one who was highly experienced in melanoma diagnosis) who were unaware of allocated treatment groups. There were technical limitations of the histopathologic review, because it was based on the slides available and because much of the material was archival. Even when tissue blocks were available, the remaining tissue might not have been representative of the tumor as a whole. We had no information on presence of ulceration or mitotic figures; definitive assessment of thickness of these study melanomas required reference to original diagnostic pathology reports, because original slides or tissue blocks no longer existed in many cases. Level of invasion also may have been subject to some uncertainty, though notdifferentially according to sunscreen, because the dermatopathologists had no knowledge of treatment allocation.
With regard to the beta carotene intervention, we found no evidence of harmful effect on melanoma or other health outcomes (mortality overall or as a result of other cancers). Although we observed fewer deaths in the intervention than in the control group on follow-up, we observed no specific beneficial effect on cancer outcomes, which is in agreement with two other randomized trials of beta carotene supplementation with follow-up24,25 and with a recent meta-analysis.26
Though this community-based skin cancer prevention trial is the only one of its kind, its findings that suggest the general preventability of melanoma after the regular application of broad-spectrum sunscreen have been predicted.11,16 They accord with those of a trial of sunscreen to prevent melanocytic nevus development in children,12 despite the limited number of children involved and the brief follow-up.27 The results are consistent with the knowledge that excessive sun exposure causes melanoma,5 notwithstanding the evidence that solar UV carcinogenesis may operate by different pathways to cause melanomas on different body sites.28,29 Results also accord with background experimental data in mice,9 including prevention of p53 mutations in UV-irradiated skin by sunscreen.30 In addition, in people with sun-sensitive skin types, daily-care sunscreen can inhibit clinical, cellular, and molecular damage caused by daily suberythemal solar-simulating radiation.31
Given the importance of early-life sun exposure in the genesis of melanoma,32 a long-term sunscreen intervention among children and adolescents may yield even greater benefits in cancer prevention33 than did this intervention in adults. The adult participants in Nambour would have experienced relatively high ambient sun exposure for years, so skin carcinogenesis may already have been initiated in many of them; only the promoting effects of ongoing adult sun exposure would have been targeted by this intervention.
In conclusion, our findings provide reassurance in view of the widespread uncertainty to date about sunscreen's ability to prevent melanoma.7,34 Although the results are directly relevant to people who live in sunny climates like Australia's and who receive relatively high levels of ambient sun exposure as a matter of course, they also have implications for white people living in temperate climates in North America and Europe35 who are at increased risk of melanoma because of increased solar UV exposure caused by the predilection for holidays in sunny places.5,36
———————————————————————————————————————-
http://jco.ascopubs.org/content/29/3/249
- © 2010 by American Society of Clinical Oncology
Sunscreen and Melanoma: What Is the Evidence?
– Author Affiliations
In this issue of Journal of Clinical Oncology, Green et al1 report on an extension study of a randomized controlled trial of daily sunscreen application and beta carotene supplementation2 that examined the melanoma experience of 1,621 patients who were enrolled onto that trial between 1992 and 1996 and who were subsequently observed for 10 years (1996 to 2006). The two interventions, sunscreen application and beta carotene supplementation, were evaluated both alone and in combination, relative to a placebo-control group. The primary end points used in the evaluation of these interventions were the incidence of basal cell carcinoma and the incidence of squamous cell carcinoma after 4.5 years of follow-up. In their current report, the authors focus on the incidence of in situ and invasive melanoma during the 15 years of trial participation as a secondary end point, and use it to test the hypothesis that “regular sunscreen use by white adults prevents the occurrence of primary cutaneous melanoma, with a possible latent effect of up to 10 years.”1 The P value associated with the Cox regression analysis used to test the hypothesis that the risk of in situ and invasive melanoma together differed between those who used the sunscreen and those who did not was .051, whereas the P value for the risk of invasive melanoma alone (an end point less confounded by variable interpretations of pathologists) was .045. Both P values could be considered of borderline significance. The authors conclude that “our findings provide reassurance… about sunscreen's ability to prevent melanoma.”
There seems to be some contradiction between the authors' conclusions and the P values that were reported. This brings to mind an article written by Goodman,3 wherein he reminds us that P values have historically been considered in two different conceptual frameworks. The first framework, as originally proposed by Fisher, was that the P value is an index or measure of evidence that can be used to measure disagreement between what is observed and what would be expected under a null hypothesis. One interpretational weakness of this approach is that it does not consider the effect size; the P value for a large effect in a small sample could be the same as the P value for a small effect in a large sample. Consequently, evidence provided by the P value alone is sometimes limited. It is important to consider additional evidence from the data beyond the P value, such as within-group estimates relevant to the clinical end point and their difference, as well as their 95% CIs.
The second framework described by Goodman3 is the one proposed by Neyman and Pearson, wherein a significance level is defined within the context of the assessment of the errors associated with a statistical hypothesis test. Here, two hypotheses are defined, the null and the alternative, and the statistical test defines one of two behaviors, to reject the null hypothesis or to accept the alternative hypothesis. The use of the statistical hypothesis test involves two types of possible errors: rejecting the null hypothesis when it is true, or accepting the alternative hypothesis when it is false, because the first error has been fixed to be .05 (the significance level generally assumed). Observed P values are then compared with the significance level. What this second framework considers, that the first does not, is the power of the statistical test. The confidence in a statistical test can be increased by using a power calculation to determine the sample size necessary to ensure that a meaningful effect can be detected; this effect is specified by the alternative hypothesis when the alternative hypothesis is true.
Next, we look at the data presented in Green et al1 from these two perspectives. What level of evidence was presented from this trial about the relationship between the regular use of sunscreen and the incidence of melanoma? The authors opted to present hazard ratios to describe the incidence of melanoma in the two groups. They reported a 50% reduction in the hazard related to in situ and invasive melanoma and a 73% reduction in the hazard related to invasive melanoma compared with the hazards of patients who were not assigned to the intervention. This information is less useful for weighing the evidence for regular sunscreen use and its prevention of melanoma in the absence of information about the relevant hazards in the nonintervention group. Ideally, we would have been given estimates of those hazards as well. For this discussion, we used the data in the authors' Table 2 to estimate rates for the observed period and assumed an intention-to-treat analysis (Table 1). The P values for Pearson's χ2 statistic under the assumption of equal rates in the two groups were .0517 and .0312, on the basis of rates defined by in situ and invasive melanomas together as well as rates for invasive melanomas only, respectively, which provided evidence that both χ2 statistics had a much higher value than one would expect according to the null hypothesis. However, more convincing evidence would involve translation of these estimates to what might be observed in a similar population. For example, in 100,000 adults in Queensland with baseline characteristics similar to those of the patients in the study by Green et al, during a period similar in length to that of the study by Green et al, all other characteristics being equal, one would expect a reduction in the number of in situ and invasive melanomas to be 1,370 (95% CI, 1,030 to 1,850) and the reduction in the number of invasive melanomas to be 990 (95% CI, 600 to 1,340) for those using sunscreen compared with those not using sunscreen.
View this table:Table 1.Group-Specific Rates, Differences, and Estimated Reduction During a Period of 15 Years for In Situ and Invasive Melanoma and Invasive Melanoma (according to intention-to-treat analysis)
What evidence do we have that the statistical hypothesis tests that were performed had sufficient power to detect a meaningful effect? On the basis of data collected up to 2004, the authors report that the power of their statistical test would be 50% to detect a hazard ratio of 0.4 on the basis of an intention-to-treat analysis for the primary end point. Because the power was lower than one would like and the observed effect size in this study (0.5) was smaller than the one assumed in 2004, an updated power calculation performed at the end of the extension study would have been informative for assessing the quality of the statistical test. For our example above, we calculated the power for the two χ2 tests using the trial sample sizes and the observed rates in the two groups to define the effect size. The power for the χ2 test performed to investigate the difference in the rates of both in situ and invasive melanomas in the two groups was 43%, whereas the power of this test to investigate differences in the invasive melanoma rates was 98%. A statistical test can be thought of as a diagnostic test, and its power is similar to the sensitivity (or specificity) of a diagnostic test. Just as with diagnostic tests, in which higher sensitivity and specificity provide greater confidence in the test, a statistical test with a power of 98% provides greater confidence than a statistical test with a power of 43% that the null hypothesis will be rejected when the alternative hypothesis is true.
What, then, is our bottom line? To our knowledge, the trial's findings are the first to provide strong evidence for a reduction in the incidence of invasive melanoma after regular application of broad-spectrum sunscreen in adults. The trial was an ambitious and unique study: it was conducted in a region with the highest rate of skin cancer in the world, had a follow-up period of 10 years after the trial, and achieved relatively high rates of compliance among the participants assigned to the group using sunscreen.4 It is unlikely that another trial of comparable scope and rigor will be conducted in the foreseeable future. Therefore, the question as to how these findings should be translated into clinical practice is important.
Do these findings end the debate about whether sunscreen use can prevent melanoma? Should clinicians and public health educators recommend daily application of sunscreen to the heads, necks, arms, and hands of many of the general population? In fact, sunscreen is the form of sun protection most often used, and is used by people of all ages,5 although its use is suboptimal. Two community surveys found that 76% to 93% of pediatricians routinely recommend that sunscreen be used when children are outside,6,7 although they usually do not give advice about optimal use and reapplication. Given current knowledge, we propose that the following are the most important clinical applications of the new findings from the trial by Green et al.1 First, clinicians should advise patients at high risk for skin cancer because of phenotypic characteristics (fair skin, freckling, tendency to sunburn, and so on), who live in or visit sunny climates, and/or who have a family history of melanoma, to routinely and thoroughly apply sunscreen before going outside. Second, public health and cancer prevention agencies should provide clear instructions regarding the use and reapplication of sunscreen in the manner described in the study by Green et al. Third, highly exposed and at-risk individuals should consider making regular sunscreen use a habit, much like other health routines; in addition, parents should apply sunscreen to their children's skin and should model the practice of sunscreen use.
Sunscreen use alone will likely not reduce the incidence of skin cancer, which is an increasingly widespread and serious public health problem in the United States and globally. In addition to sunscreen use, excess exposure to ultraviolet rays should be avoided, clothing should be used to shield skin from the sun, and sun-safe environments should be used for outdoor recreation. In addition, sunscreen use should be paired with regular self-examination of the skin. The question of its efficacy with respect to melanoma prevention should no longer deter scientists or clinicians from recommending sunscreen use.
AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST
The author(s) indicated no potential conflicts of interest.
AUTHOR CONTRIBUTIONS
Conception and design: Phyllis A. Gimotty
Data analysis and interpretation: Phyllis A. Gimotty
Manuscript writing: Phyllis A. Gimotty, Karen Glanz
Final approval of manuscript: Phyllis A. Gimotty, Karen Glanz
REFERENCES
CiteULike
Connotea
Delicious
Digg
Facebook
Reddit
Technorati
Twitter
Articles citing this article
- © 2006 by American Society of Clinical Oncology
Anatomic Site, Sun Exposure, and Risk of Cutaneous Melanoma
- David C. Whiteman,
- Mark Stickley,
- Peter Watt,
- Maria Celia Hughes,
- Marcia B. Davis and
- Adèle C. Green
+ Author Affiliations
- Address reprint requests to David C. Whiteman, MBBS, PhD, Division of Population Studies and Human Genetics, Queensland Institute of Medical Research, PO Royal Brisbane Hospital, Queensland 4029 Australia; e-mail: [email protected]
Abstract
Purpose Sunlight is the principal environmental risk factor for cutaneous melanoma. A current hypothesis postulates that the role of sunlight in causing melanoma differs according to anatomic site. We tested this hypothesis in a population-based case-case comparative study of melanoma patients.
Methods Patients were sampled from the Queensland cancer registry in three groups: superficial spreading or nodular melanomas of the trunk (n = 154), of the head and neck (HN; n = 76), and lentigo maligna (LM) and lentigo maligna melanoma (LMM; for both LM and LMM, n = 76). Data were collected on school-age sun exposure and occupational and recreational sun exposure in adulthood. Odds ratios (OR) and 95% CIs were calculated using polytomous logistic regression.
Results HN melanoma patients were substantially more likely than trunk patients to have higher levels of sun exposure in adulthood (OR, 2.43; 95% CI, 0.98 to 5.99) and specifically, higher levels of occupational exposure (OR, 3.25; 95% CI, 1.32 to 8.00), but lower levels of recreational sun exposure (OR, 0.50; 95% CI, 0.21 to 1.19). LM and LMM patients reported higher occupational exposure and lower recreational sun exposure than trunk melanoma patients, although this was not significant. We found no significant differences between the groups for school-age sun exposures.
Conclusion Melanomas developing at different body sites are associated with distinct patterns of sun exposure. Melanomas of the head and neck are associated with chronic patterns of sun exposure whereas trunk melanomas are associated with intermittent patterns of sun exposure, supporting the hypothesis that melanomas may arise through divergent causal pathways.
Footnotes
-
Supported by grants from the Queensland Cancer Fund and the National Cancer Institute (CA 88363-01A1). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute.
Presented in part during an invited plenary session at the 6th World Congress on Melanoma, Vancouver, British Columbia, Canada, September 8, 2005 (D.C.W.).
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
- Received February 9, 2006.
- Accepted April 21, 2006.
-
fyi
http://jco.ascopubs.org/content/29/3/257.full
- © 2010 by American Society of Clinical Oncology
Reduced Melanoma After Regular Sunscreen Use: Randomized Trial Follow-Up
+ Author Affiliations
From the Queensland Institute of Medical Research; University of Queensland; and Princess Alexandra Hospital, Brisbane, Australia.
- Corresponding author: Adèle Green, Queensland Institute of Medical Research, PO Royal Brisbane Hospital, Queensland 4029, Australia; e-mail: [email protected].
Abstract
Purpose Regular sunscreen use prevents cutaneous squamous cell carcinoma long term, but the effect on melanoma is highly controversial. We evaluated whether long-term application of sunscreen decreases risk of cutaneous melanoma.
Participants and Methods In 1992, 1,621 randomly selected residents of Nambour, a township in Queensland, Australia, age 25 to 75 years, were randomly assigned to daily or discretionary sunscreen application to head and arms in combination with 30 mg beta carotene or placebo supplements until 1996. Participants were observed until 2006 with questionnaires and/or through pathology laboratories and the cancer registry to ascertain primary melanoma occurrence.
Results Ten years after trial cessation, 11 new primary melanomas had been identified in the daily sunscreen group, and 22 had been identified in the discretionary group, which represented a reduction of the observed rate in those randomly assigned to daily sunscreen use (hazard ratio [HR], 0.50; 95% CI, 0.24 to 1.02; P = .051). The reduction in invasive melanomas was substantial (n = 3 in active v 11 in control group; HR, 0.27; 95% CI, 0.08 to 0.97) compared with that for preinvasive melanomas (HR, 0.73; 95% CI, 0.29 to 1.81).
Conclusion Melanoma may be preventable by regular sunscreen use in adults.
INTRODUCTION
The need for more effective prevention of melanoma is recognized around the world as climbing incidence and high mortality in white populations persist.1–3 In the United States, approximately 68,700 new melanoma occurrences and more than 8,600 deaths were expected to occur in 2009.4 Exposure to solar ultraviolet (UV) radiation is the only established modifiable cause of melanoma.5,6
Despite the known etiologic role of sun exposure, the question regarding sunscreen use to prevent melanoma remains open7 and controversial.8 Although case-control and cohort studies of sunscreen use and melanoma risk abound, their findings have been uninformative.9–11 Nonrandomized studies of the melanoma-sunscreen association are unable to distinguish the main determinants of sunscreen use from those of melanoma, because they are the same—namely, susceptibility to sunburn, high occupational or recreational sun exposure, and family history.11 The only relevant evidence comes from a randomized trial of sunscreen application in Canadian children conducted from 1993 to 1996 that showed a small reduction in new melanocytic nevi,12 the strongest predictors of melanoma, in children allocated to a sunscreen arm, especially if the children had freckles.
Here, we present new evidence from the follow-up of a community-based, pragmatic trial of sunscreen to prevent skin cancer in Queensland, Australia.13 By primarily aiming to evaluate prevention of basal cell carcinoma (BCC) and squamous cell carcinomas (SCC) of the skin and their precursors,13,14 we have shown prolonged prevention of SCC by the sunscreen intervention.15 Now, this trial and its 10-year aftermath have given us the unique opportunity to examine melanoma as a secondary trial end point by using unconfounded evidence of the long-term effectiveness of sunscreen in reducing melanoma incidence. The hypothesis was that regular sunscreen use by white adults prevents the occurrence of primary cutaneous melanoma,11,16 with a possible latent effect of up to 10 years.
METHODS
Study Design
In 1992, 1,621 residents of the Queensland township of Nambour who were ascertained in 1986 at ages 20 to 69 years for a skin cancer prevalence survey were enlisted in the Nambour Skin Cancer Prevention Trial. Original survey participants had been randomly sampled from the Nambour electoral roll (enrollment is compulsory by law),14 and those who participated in the trial were representative of the original sample.17 Trial participants were randomly assigned individually by using a computer-generated, randomized list (without stratification or blocking). The 812 participants randomly assigned to sunscreen intervention were given a free, unlimited supply of broad-spectrum sunscreen containing 8% (by weight) 2-ethyl hexyl-p-methoxycinnamate and 2% (by weight) 4-tert-butyl-4′ methoxy-4-dibenzoylmethane and with a sun protection factor (SPF) of 16. They were asked to apply it to head, neck, arms, and hands every morning (and reapplication was advised after heavy sweating, bathing, or long sun exposure). The 809 participants randomly assigned to the comparison group continued using sunscreen of any SPF at their usual, discretionary frequency, which included no use.13 Allocation of a placebo sunscreen to the control group was unethical, given the subtropical location. According to a 2 × 2 factorial design, 820 participants (n = 404 and n = 416 in daily and discretionary sunscreen groups, respectively) were also independently randomly assigned to 30 mg beta carotene, and 801 participants (n = 408 and n = 393 in daily and discretionary sunscreen groups, respectively) were randomly assigned to placebo supplements, because beta carotene potentially could counteract the oxidative damage to DNA involved in solar UV carcinogenesis.13
Compliance with sunscreen treatment was assessed by measured weights of returned sunscreen bottles and by questionnaires asking average frequency of use in a normal week. Intake of supplements was assessed by remaining tablet counts.13 Dermatologists unaware of treatment allocations conducted skin examinations of participants at baseline (March 1992), including assessment of number of nevi on the back; midway (1994); and at trial end (August 1996). Participants diagnosed with suspected melanomas were referred to their physicians for immediate management. All skin cancers, including melanomas diagnosed between surveys, were ascertained quarterly with histologic confirmation of any reported. Information about risk factors for skin cancers, such as skin color, outdoor behavior, and sunburn history, was obtained at baseline, and information on sun exposure and protection was updated throughout the trial. Ethical approval was obtained from institutional ethics committees, and study participants gave their written informed consent.
Follow-Up
After the scheduled trial completion in 1996, 1,339 participants (82%, including 14 participants who moved outside Queensland) agreed to take part in the follow-up study actively (Fig 1); they completed biannual or annual questionnaires about all new skin cancers, including melanomas. In addition, they reported average time outdoors on weekdays and weekends in the previous 6 or 12 months and average sunscreen use (although no sunscreen was supplied after 1996). Participants who withdrew from active trial participation or active follow-up were asked to continue with ongoing passive monitoring of skin cancers through their medical records.14 Investigators thus obtained notification of all melanomas diagnosed by regional pathology laboratories in active and passive participants. Finally, we cross-checked for any melanomas diagnosed between 1992 and 2006 through a search of the Queensland Cancer Registry (because melanoma registration is compulsory, is a particularly high priority, and is considered virtually complete); however, no new melanomas were uncovered in the Cancer Registry checks that had not already been ascertained.
Fig 1.Nambour Skin Cancer Prevention Trial follow-up profile.
Review of each diagnosed melanoma was undertaken by two expert dermatopathologists who were unaware of sunscreen allocation, and reviews were based on available pathology slides. The histologic diagnosis of melanomas of any type, both in situ and invasive, was based on a constellation of features developed during several decades.18 Melanoma in situ, lentigo maligna type, was diagnosed by using defined criteria.19 Invasion was defined and classified according to Clark's level of invasion: Level 1 (in situ); Level 2, tumor in the papillary dermis; Level 3, tumor filling the papillary dermis and extending to the papillary dermis/reticular dermis interface; Level 4, tumor in reticular dermis; and Level 5, tumor in fat.20 Classification by histologic type was not undertaken, although no melanoma types were excluded.
Statistical Analysis
When we formally assessed the long-term trial results for BCC and SCC to the end of 2004, one of us (G.W.) also carried out a preliminary intention-to-treat analysis on the basis of accumulated but unreviewed melanoma reports for all body sites. The decision then was made to evaluate melanoma occurrence classified according to invasiveness up to December 2006, because 15 years of follow-up (calendar time) was deemed sufficient to detect an effect of sunscreen, if present. On the basis of the observed rate of melanoma in the control group to 2004, the power was estimated to be 66%, and 50%, for detecting hazard ratios of 0.3 and 0.4, respectively, with a two-sided α of .05. As for the other skin cancer end points,13 melanomas diagnosed in the first year of intervention were excluded a priori, because their development was unlikely to have been affected by the introduced sunscreen treatment. Cox proportional hazards regression, with the sunscreen and beta carotene interventions as two main effects, was used to examine treatment effects in relation to primary melanoma occurrence with incorporation of lead time. Individual effects of sunscreen and beta carotene were tested by using likelihood ratio tests. Subgroup analyses were performed to assess consistency of effect according to age, sex, phenotype, sun exposure, and history of skin cancer by using Cox regression and by incorporating a subgroup interaction term to detect heterogeneity of effects.
During trial follow-up, mean hours each day spent outdoors on weekdays and on weekends were calculated, and sun protection habits were assessed from questionnaire responses. Sun exposure and protection durations were compared across the two sunscreen treatment groups by using a two-sample t test. All reported P values were two sided.
RESULTS
Balance was achieved with respect to established risk factors for skin cancer and melanoma among the Nambour trial participants randomly assigned to sunscreen or control in 1992 (Table 1). Compliance with sunscreen treatment, assessed by average of reported frequencies of application, measured weights of returned sunscreen bottles, and diaries, was approximately 75%,13 and 25% of the intervention group applied sunscreen to trunk and/or lower limbs as well as to the intervention sites.21 The majority of participants in the control group either did not apply sunscreen (38%) or applied it once or twice a week at most (35%), and 8% applied it to nonintervention sites.21 Compliance was approximately 70% for beta carotene and placebo supplementation.13
View this table:Table 1.Demographic and Clinical Characteristics of Participants at Baseline in 1992 According to Sunscreen Allocation
By the end of 2006, 846 people (52%) were actively completing questionnaires, 600 (37%) were passive participants, and 173 (11%) had died (n = 87, sunscreen group; n = 86, controls; n = 71, beta carotene group; n = 102, placebo controls), including one person who died as a result of melanoma diagnosed in 1978. One person from each sunscreen treatment group had withdrawn their permission for passive follow-up, in 2002 and 2001, respectively, and data for both were censored accordingly. There was no significant difference in mode of follow-up in relation to sunscreen allocation: duration of active follow-up was 14.3 person-years (95% CI, 14.1 to 14.4 person-years) versus 14.2 person-years (95% CI, 14.0 to 14.3 person-years), and active response rates were 94.2% and 94.3% in each group, respectively. On the basis of reports of active participants, 25% of those randomly assigned to daily sunscreen continued to use sunscreen on a regular basis after the trial22 compared with 18% of the nonintervention group (P = .004).15 By 2001, less than 3% of either supplement group took beta carotene supplements.
In the almost 15 years from commencement of the trial in March 1992 until the end of follow-up in December 2006, 36 of the 1,621 trial participants developed first primary melanomas (n = 22, in situ; n = 14, invasive; none metastatic), and one person in each trial arm developed two primary melanomas. Three people (n = 1, intervention group; n = 2, control) who had melanomas diagnosed in 1992 were excluded a priori. In the remainder of the trial from 1993 to 1996, two participants in the daily sunscreen group and seven in the discretionary group were diagnosed with melanoma (Fig 2). From trial cessation until the end of 2006, nine more participants allocated to daily sunscreen and 15 allocated to discretionary use were diagnosed with incident melanomas (Fig 2). In all, 11 trial participants in the sunscreen intervention and 22 in the control group (Table 2) were newly diagnosed with primary melanoma between 1993 and 2006. Risk of melanoma overall was reduced in those randomly assigned to daily sunscreen compared with discretionary use (hazard ratio [HR], 0.50; 95% CI, 0.24 to 1.02; P = .051), although the result was of borderline statistical significance. Invasive melanoma was reduced by 73% in the daily sunscreen group (HR, 0.27; 95% CI, 0.08 to 0.97; P = .045; Table 2); average thickness was 0.53 mm in the sunscreen group and 1.2 mm in controls (P = .08) on the basis of the original pathology reports (except for one level 2 melanoma in the control group for which thickness measurement was unavailable). There were no significant differences between intervention and control arms with respect to either in situ melanomas (8 v 11, respectively; HR, 0.73; 95% CI, 0.29 to 1.81) or melanomas on prescribed application sites (HR, 0.46; 95% CI, 0.17 to 1.20). When a multivariate proportional hazards regression was carried out that included sex, skin type, numbers of nevi, previous history of skin cancer, and sun exposure along with the two treatment categories, the overall effect estimate for sunscreen varied little (HR, 0.49; 95% CI, 0.24 to 1.02). Regarding outcome according to beta carotene randomization, 16 and 17 melanomas occurred in those taking active and placebo supplements, respectively (HR, 0.89; 95% CI, 0.45 to 1.76).
Fig 2.Occurrence of first primary melanoma by level of invasion and anatomic site in the two sunscreen treatment groups.
View this table:Table 2.First Primary Melanomas During 1993-2006 According to Randomized Sunscreen Intervention During 1992-1996 and Risk of Melanoma
No significant interactions with any baseline characteristics were found (Fig 3). Sun exposure was similar between the daily and discretionary sunscreen groups during the trial (79% and 77%, respectively, spent less than 50% of weekend time outdoors23) or after the trial (3.8 and 3.9 hours each day, respectively, spent outdoors on weekdays [P = .46]; 4.6 and 4.7 hours each day, respectively, spent outdoors on weekends [P = .79]). Use of sun protection measures other than sunscreen was similar during the trial (approximately 60% of both groups usually sought shade; around 75% usually wore a hat) and after the trial (at midpoint of follow-up, 40% and 67% usually sought shade and wore a hat in the sun, respectively).
Fig 3.Effect of sunscreen intervention on melanoma according to baseline characteristics. Hazard ratios are for melanoma in a comparison of the sunscreen intervention and control groups. Hazard ratios, 95% CIs, and P values were calculated by using Cox regression that incorporated a subgroup interaction term to detect heterogeneity of effects.
DISCUSSION
Long-term follow-up of this randomized trial showed that, among adults age 25 to 75 years, regular application of SPF 15+ sunscreen in a 5-year period appeared to reduce the incidence of new primary melanomas for up to 10 years after trial cessation. A protective effect was also evident for invasive melanoma, which showed a 73% decrease in those randomly assigned to daily sunscreen after approximately 15 years of follow-up. This result for the invasive subgroup was an exploratory finding, however, and should be interpreted cautiously. The apparent decrease in melanoma across all body sites, not only prescribed application sites, likely reflected the ongoing tendency of more participants in the daily than the discretionary sunscreen group to apply sunscreen regularly to the trunk and lower limbs.21,22 Intensity of application also tended to be higher in the intervention group.21
There was no evidence to suggest that the observed difference could be explained by a difference in sun exposure in the intervention group relative to the control group, because outdoor behavior was similar during23 and after the trial. Similarly, there were no differences in active completion of follow-up questionnaires with self-reports (later validated) of all new skin cancers, including melanomas. Without reference to treatment groups, we attempted to capture all melanomas diagnosed between 1993 and 2006 among passive trial participants (except for two who withdrew consent for follow-up) from the Queensland Cancer Registry as well as from the pathology laboratories. Ultimately, however, the final diagnosis of each melanoma reported during or after the trial was determined by the two reviewing dermatopathologists (one who was highly experienced in melanoma diagnosis) who were unaware of allocated treatment groups. There were technical limitations of the histopathologic review, because it was based on the slides available and because much of the material was archival. Even when tissue blocks were available, the remaining tissue might not have been representative of the tumor as a whole. We had no information on presence of ulceration or mitotic figures; definitive assessment of thickness of these study melanomas required reference to original diagnostic pathology reports, because original slides or tissue blocks no longer existed in many cases. Level of invasion also may have been subject to some uncertainty, though notdifferentially according to sunscreen, because the dermatopathologists had no knowledge of treatment allocation.
With regard to the beta carotene intervention, we found no evidence of harmful effect on melanoma or other health outcomes (mortality overall or as a result of other cancers). Although we observed fewer deaths in the intervention than in the control group on follow-up, we observed no specific beneficial effect on cancer outcomes, which is in agreement with two other randomized trials of beta carotene supplementation with follow-up24,25 and with a recent meta-analysis.26
Though this community-based skin cancer prevention trial is the only one of its kind, its findings that suggest the general preventability of melanoma after the regular application of broad-spectrum sunscreen have been predicted.11,16 They accord with those of a trial of sunscreen to prevent melanocytic nevus development in children,12 despite the limited number of children involved and the brief follow-up.27 The results are consistent with the knowledge that excessive sun exposure causes melanoma,5 notwithstanding the evidence that solar UV carcinogenesis may operate by different pathways to cause melanomas on different body sites.28,29 Results also accord with background experimental data in mice,9 including prevention of p53 mutations in UV-irradiated skin by sunscreen.30 In addition, in people with sun-sensitive skin types, daily-care sunscreen can inhibit clinical, cellular, and molecular damage caused by daily suberythemal solar-simulating radiation.31
Given the importance of early-life sun exposure in the genesis of melanoma,32 a long-term sunscreen intervention among children and adolescents may yield even greater benefits in cancer prevention33 than did this intervention in adults. The adult participants in Nambour would have experienced relatively high ambient sun exposure for years, so skin carcinogenesis may already have been initiated in many of them; only the promoting effects of ongoing adult sun exposure would have been targeted by this intervention.
In conclusion, our findings provide reassurance in view of the widespread uncertainty to date about sunscreen's ability to prevent melanoma.7,34 Although the results are directly relevant to people who live in sunny climates like Australia's and who receive relatively high levels of ambient sun exposure as a matter of course, they also have implications for white people living in temperate climates in North America and Europe35 who are at increased risk of melanoma because of increased solar UV exposure caused by the predilection for holidays in sunny places.5,36
———————————————————————————————————————-
http://jco.ascopubs.org/content/29/3/249
- © 2010 by American Society of Clinical Oncology
Sunscreen and Melanoma: What Is the Evidence?
– Author Affiliations
In this issue of Journal of Clinical Oncology, Green et al1 report on an extension study of a randomized controlled trial of daily sunscreen application and beta carotene supplementation2 that examined the melanoma experience of 1,621 patients who were enrolled onto that trial between 1992 and 1996 and who were subsequently observed for 10 years (1996 to 2006). The two interventions, sunscreen application and beta carotene supplementation, were evaluated both alone and in combination, relative to a placebo-control group. The primary end points used in the evaluation of these interventions were the incidence of basal cell carcinoma and the incidence of squamous cell carcinoma after 4.5 years of follow-up. In their current report, the authors focus on the incidence of in situ and invasive melanoma during the 15 years of trial participation as a secondary end point, and use it to test the hypothesis that “regular sunscreen use by white adults prevents the occurrence of primary cutaneous melanoma, with a possible latent effect of up to 10 years.”1 The P value associated with the Cox regression analysis used to test the hypothesis that the risk of in situ and invasive melanoma together differed between those who used the sunscreen and those who did not was .051, whereas the P value for the risk of invasive melanoma alone (an end point less confounded by variable interpretations of pathologists) was .045. Both P values could be considered of borderline significance. The authors conclude that “our findings provide reassurance… about sunscreen's ability to prevent melanoma.”
There seems to be some contradiction between the authors' conclusions and the P values that were reported. This brings to mind an article written by Goodman,3 wherein he reminds us that P values have historically been considered in two different conceptual frameworks. The first framework, as originally proposed by Fisher, was that the P value is an index or measure of evidence that can be used to measure disagreement between what is observed and what would be expected under a null hypothesis. One interpretational weakness of this approach is that it does not consider the effect size; the P value for a large effect in a small sample could be the same as the P value for a small effect in a large sample. Consequently, evidence provided by the P value alone is sometimes limited. It is important to consider additional evidence from the data beyond the P value, such as within-group estimates relevant to the clinical end point and their difference, as well as their 95% CIs.
The second framework described by Goodman3 is the one proposed by Neyman and Pearson, wherein a significance level is defined within the context of the assessment of the errors associated with a statistical hypothesis test. Here, two hypotheses are defined, the null and the alternative, and the statistical test defines one of two behaviors, to reject the null hypothesis or to accept the alternative hypothesis. The use of the statistical hypothesis test involves two types of possible errors: rejecting the null hypothesis when it is true, or accepting the alternative hypothesis when it is false, because the first error has been fixed to be .05 (the significance level generally assumed). Observed P values are then compared with the significance level. What this second framework considers, that the first does not, is the power of the statistical test. The confidence in a statistical test can be increased by using a power calculation to determine the sample size necessary to ensure that a meaningful effect can be detected; this effect is specified by the alternative hypothesis when the alternative hypothesis is true.
Next, we look at the data presented in Green et al1 from these two perspectives. What level of evidence was presented from this trial about the relationship between the regular use of sunscreen and the incidence of melanoma? The authors opted to present hazard ratios to describe the incidence of melanoma in the two groups. They reported a 50% reduction in the hazard related to in situ and invasive melanoma and a 73% reduction in the hazard related to invasive melanoma compared with the hazards of patients who were not assigned to the intervention. This information is less useful for weighing the evidence for regular sunscreen use and its prevention of melanoma in the absence of information about the relevant hazards in the nonintervention group. Ideally, we would have been given estimates of those hazards as well. For this discussion, we used the data in the authors' Table 2 to estimate rates for the observed period and assumed an intention-to-treat analysis (Table 1). The P values for Pearson's χ2 statistic under the assumption of equal rates in the two groups were .0517 and .0312, on the basis of rates defined by in situ and invasive melanomas together as well as rates for invasive melanomas only, respectively, which provided evidence that both χ2 statistics had a much higher value than one would expect according to the null hypothesis. However, more convincing evidence would involve translation of these estimates to what might be observed in a similar population. For example, in 100,000 adults in Queensland with baseline characteristics similar to those of the patients in the study by Green et al, during a period similar in length to that of the study by Green et al, all other characteristics being equal, one would expect a reduction in the number of in situ and invasive melanomas to be 1,370 (95% CI, 1,030 to 1,850) and the reduction in the number of invasive melanomas to be 990 (95% CI, 600 to 1,340) for those using sunscreen compared with those not using sunscreen.
View this table:Table 1.Group-Specific Rates, Differences, and Estimated Reduction During a Period of 15 Years for In Situ and Invasive Melanoma and Invasive Melanoma (according to intention-to-treat analysis)
What evidence do we have that the statistical hypothesis tests that were performed had sufficient power to detect a meaningful effect? On the basis of data collected up to 2004, the authors report that the power of their statistical test would be 50% to detect a hazard ratio of 0.4 on the basis of an intention-to-treat analysis for the primary end point. Because the power was lower than one would like and the observed effect size in this study (0.5) was smaller than the one assumed in 2004, an updated power calculation performed at the end of the extension study would have been informative for assessing the quality of the statistical test. For our example above, we calculated the power for the two χ2 tests using the trial sample sizes and the observed rates in the two groups to define the effect size. The power for the χ2 test performed to investigate the difference in the rates of both in situ and invasive melanomas in the two groups was 43%, whereas the power of this test to investigate differences in the invasive melanoma rates was 98%. A statistical test can be thought of as a diagnostic test, and its power is similar to the sensitivity (or specificity) of a diagnostic test. Just as with diagnostic tests, in which higher sensitivity and specificity provide greater confidence in the test, a statistical test with a power of 98% provides greater confidence than a statistical test with a power of 43% that the null hypothesis will be rejected when the alternative hypothesis is true.
What, then, is our bottom line? To our knowledge, the trial's findings are the first to provide strong evidence for a reduction in the incidence of invasive melanoma after regular application of broad-spectrum sunscreen in adults. The trial was an ambitious and unique study: it was conducted in a region with the highest rate of skin cancer in the world, had a follow-up period of 10 years after the trial, and achieved relatively high rates of compliance among the participants assigned to the group using sunscreen.4 It is unlikely that another trial of comparable scope and rigor will be conducted in the foreseeable future. Therefore, the question as to how these findings should be translated into clinical practice is important.
Do these findings end the debate about whether sunscreen use can prevent melanoma? Should clinicians and public health educators recommend daily application of sunscreen to the heads, necks, arms, and hands of many of the general population? In fact, sunscreen is the form of sun protection most often used, and is used by people of all ages,5 although its use is suboptimal. Two community surveys found that 76% to 93% of pediatricians routinely recommend that sunscreen be used when children are outside,6,7 although they usually do not give advice about optimal use and reapplication. Given current knowledge, we propose that the following are the most important clinical applications of the new findings from the trial by Green et al.1 First, clinicians should advise patients at high risk for skin cancer because of phenotypic characteristics (fair skin, freckling, tendency to sunburn, and so on), who live in or visit sunny climates, and/or who have a family history of melanoma, to routinely and thoroughly apply sunscreen before going outside. Second, public health and cancer prevention agencies should provide clear instructions regarding the use and reapplication of sunscreen in the manner described in the study by Green et al. Third, highly exposed and at-risk individuals should consider making regular sunscreen use a habit, much like other health routines; in addition, parents should apply sunscreen to their children's skin and should model the practice of sunscreen use.
Sunscreen use alone will likely not reduce the incidence of skin cancer, which is an increasingly widespread and serious public health problem in the United States and globally. In addition to sunscreen use, excess exposure to ultraviolet rays should be avoided, clothing should be used to shield skin from the sun, and sun-safe environments should be used for outdoor recreation. In addition, sunscreen use should be paired with regular self-examination of the skin. The question of its efficacy with respect to melanoma prevention should no longer deter scientists or clinicians from recommending sunscreen use.
AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST
The author(s) indicated no potential conflicts of interest.
AUTHOR CONTRIBUTIONS
Conception and design: Phyllis A. Gimotty
Data analysis and interpretation: Phyllis A. Gimotty
Manuscript writing: Phyllis A. Gimotty, Karen Glanz
Final approval of manuscript: Phyllis A. Gimotty, Karen Glanz
REFERENCES
- CiteULike
- Connotea
- Delicious
- Digg
- Facebook
- Reddit
- Technorati
- Twitter
Articles citing this article
- © 2006 by American Society of Clinical Oncology
Anatomic Site, Sun Exposure, and Risk of Cutaneous Melanoma
- David C. Whiteman,
- Mark Stickley,
- Peter Watt,
- Maria Celia Hughes,
- Marcia B. Davis and
- Adèle C. Green
+ Author Affiliations
- Address reprint requests to David C. Whiteman, MBBS, PhD, Division of Population Studies and Human Genetics, Queensland Institute of Medical Research, PO Royal Brisbane Hospital, Queensland 4029 Australia; e-mail: [email protected]
Abstract
Purpose Sunlight is the principal environmental risk factor for cutaneous melanoma. A current hypothesis postulates that the role of sunlight in causing melanoma differs according to anatomic site. We tested this hypothesis in a population-based case-case comparative study of melanoma patients.
Methods Patients were sampled from the Queensland cancer registry in three groups: superficial spreading or nodular melanomas of the trunk (n = 154), of the head and neck (HN; n = 76), and lentigo maligna (LM) and lentigo maligna melanoma (LMM; for both LM and LMM, n = 76). Data were collected on school-age sun exposure and occupational and recreational sun exposure in adulthood. Odds ratios (OR) and 95% CIs were calculated using polytomous logistic regression.
Results HN melanoma patients were substantially more likely than trunk patients to have higher levels of sun exposure in adulthood (OR, 2.43; 95% CI, 0.98 to 5.99) and specifically, higher levels of occupational exposure (OR, 3.25; 95% CI, 1.32 to 8.00), but lower levels of recreational sun exposure (OR, 0.50; 95% CI, 0.21 to 1.19). LM and LMM patients reported higher occupational exposure and lower recreational sun exposure than trunk melanoma patients, although this was not significant. We found no significant differences between the groups for school-age sun exposures.
Conclusion Melanomas developing at different body sites are associated with distinct patterns of sun exposure. Melanomas of the head and neck are associated with chronic patterns of sun exposure whereas trunk melanomas are associated with intermittent patterns of sun exposure, supporting the hypothesis that melanomas may arise through divergent causal pathways.
Footnotes
-
Supported by grants from the Queensland Cancer Fund and the National Cancer Institute (CA 88363-01A1). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute.
Presented in part during an invited plenary session at the 6th World Congress on Melanoma, Vancouver, British Columbia, Canada, September 8, 2005 (D.C.W.).
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
- Received February 9, 2006.
- Accepted April 21, 2006.
-
fyi
http://jco.ascopubs.org/content/29/3/257.full
- © 2010 by American Society of Clinical Oncology
Reduced Melanoma After Regular Sunscreen Use: Randomized Trial Follow-Up
+ Author Affiliations
From the Queensland Institute of Medical Research; University of Queensland; and Princess Alexandra Hospital, Brisbane, Australia.
- Corresponding author: Adèle Green, Queensland Institute of Medical Research, PO Royal Brisbane Hospital, Queensland 4029, Australia; e-mail: [email protected].
Abstract
Purpose Regular sunscreen use prevents cutaneous squamous cell carcinoma long term, but the effect on melanoma is highly controversial. We evaluated whether long-term application of sunscreen decreases risk of cutaneous melanoma.
Participants and Methods In 1992, 1,621 randomly selected residents of Nambour, a township in Queensland, Australia, age 25 to 75 years, were randomly assigned to daily or discretionary sunscreen application to head and arms in combination with 30 mg beta carotene or placebo supplements until 1996. Participants were observed until 2006 with questionnaires and/or through pathology laboratories and the cancer registry to ascertain primary melanoma occurrence.
Results Ten years after trial cessation, 11 new primary melanomas had been identified in the daily sunscreen group, and 22 had been identified in the discretionary group, which represented a reduction of the observed rate in those randomly assigned to daily sunscreen use (hazard ratio [HR], 0.50; 95% CI, 0.24 to 1.02; P = .051). The reduction in invasive melanomas was substantial (n = 3 in active v 11 in control group; HR, 0.27; 95% CI, 0.08 to 0.97) compared with that for preinvasive melanomas (HR, 0.73; 95% CI, 0.29 to 1.81).
Conclusion Melanoma may be preventable by regular sunscreen use in adults.
INTRODUCTION
The need for more effective prevention of melanoma is recognized around the world as climbing incidence and high mortality in white populations persist.1–3 In the United States, approximately 68,700 new melanoma occurrences and more than 8,600 deaths were expected to occur in 2009.4 Exposure to solar ultraviolet (UV) radiation is the only established modifiable cause of melanoma.5,6
Despite the known etiologic role of sun exposure, the question regarding sunscreen use to prevent melanoma remains open7 and controversial.8 Although case-control and cohort studies of sunscreen use and melanoma risk abound, their findings have been uninformative.9–11 Nonrandomized studies of the melanoma-sunscreen association are unable to distinguish the main determinants of sunscreen use from those of melanoma, because they are the same—namely, susceptibility to sunburn, high occupational or recreational sun exposure, and family history.11 The only relevant evidence comes from a randomized trial of sunscreen application in Canadian children conducted from 1993 to 1996 that showed a small reduction in new melanocytic nevi,12 the strongest predictors of melanoma, in children allocated to a sunscreen arm, especially if the children had freckles.
Here, we present new evidence from the follow-up of a community-based, pragmatic trial of sunscreen to prevent skin cancer in Queensland, Australia.13 By primarily aiming to evaluate prevention of basal cell carcinoma (BCC) and squamous cell carcinomas (SCC) of the skin and their precursors,13,14 we have shown prolonged prevention of SCC by the sunscreen intervention.15 Now, this trial and its 10-year aftermath have given us the unique opportunity to examine melanoma as a secondary trial end point by using unconfounded evidence of the long-term effectiveness of sunscreen in reducing melanoma incidence. The hypothesis was that regular sunscreen use by white adults prevents the occurrence of primary cutaneous melanoma,11,16 with a possible latent effect of up to 10 years.
METHODS
Study Design
In 1992, 1,621 residents of the Queensland township of Nambour who were ascertained in 1986 at ages 20 to 69 years for a skin cancer prevalence survey were enlisted in the Nambour Skin Cancer Prevention Trial. Original survey participants had been randomly sampled from the Nambour electoral roll (enrollment is compulsory by law),14 and those who participated in the trial were representative of the original sample.17 Trial participants were randomly assigned individually by using a computer-generated, randomized list (without stratification or blocking). The 812 participants randomly assigned to sunscreen intervention were given a free, unlimited supply of broad-spectrum sunscreen containing 8% (by weight) 2-ethyl hexyl-p-methoxycinnamate and 2% (by weight) 4-tert-butyl-4′ methoxy-4-dibenzoylmethane and with a sun protection factor (SPF) of 16. They were asked to apply it to head, neck, arms, and hands every morning (and reapplication was advised after heavy sweating, bathing, or long sun exposure). The 809 participants randomly assigned to the comparison group continued using sunscreen of any SPF at their usual, discretionary frequency, which included no use.13 Allocation of a placebo sunscreen to the control group was unethical, given the subtropical location. According to a 2 × 2 factorial design, 820 participants (n = 404 and n = 416 in daily and discretionary sunscreen groups, respectively) were also independently randomly assigned to 30 mg beta carotene, and 801 participants (n = 408 and n = 393 in daily and discretionary sunscreen groups, respectively) were randomly assigned to placebo supplements, because beta carotene potentially could counteract the oxidative damage to DNA involved in solar UV carcinogenesis.13
Compliance with sunscreen treatment was assessed by measured weights of returned sunscreen bottles and by questionnaires asking average frequency of use in a normal week. Intake of supplements was assessed by remaining tablet counts.13 Dermatologists unaware of treatment allocations conducted skin examinations of participants at baseline (March 1992), including assessment of number of nevi on the back; midway (1994); and at trial end (August 1996). Participants diagnosed with suspected melanomas were referred to their physicians for immediate management. All skin cancers, including melanomas diagnosed between surveys, were ascertained quarterly with histologic confirmation of any reported. Information about risk factors for skin cancers, such as skin color, outdoor behavior, and sunburn history, was obtained at baseline, and information on sun exposure and protection was updated throughout the trial. Ethical approval was obtained from institutional ethics committees, and study participants gave their written informed consent.
Follow-Up
After the scheduled trial completion in 1996, 1,339 participants (82%, including 14 participants who moved outside Queensland) agreed to take part in the follow-up study actively (Fig 1); they completed biannual or annual questionnaires about all new skin cancers, including melanomas. In addition, they reported average time outdoors on weekdays and weekends in the previous 6 or 12 months and average sunscreen use (although no sunscreen was supplied after 1996). Participants who withdrew from active trial participation or active follow-up were asked to continue with ongoing passive monitoring of skin cancers through their medical records.14 Investigators thus obtained notification of all melanomas diagnosed by regional pathology laboratories in active and passive participants. Finally, we cross-checked for any melanomas diagnosed between 1992 and 2006 through a search of the Queensland Cancer Registry (because melanoma registration is compulsory, is a particularly high priority, and is considered virtually complete); however, no new melanomas were uncovered in the Cancer Registry checks that had not already been ascertained.
Fig 1.Nambour Skin Cancer Prevention Trial follow-up profile.
Review of each diagnosed melanoma was undertaken by two expert dermatopathologists who were unaware of sunscreen allocation, and reviews were based on available pathology slides. The histologic diagnosis of melanomas of any type, both in situ and invasive, was based on a constellation of features developed during several decades.18 Melanoma in situ, lentigo maligna type, was diagnosed by using defined criteria.19 Invasion was defined and classified according to Clark's level of invasion: Level 1 (in situ); Level 2, tumor in the papillary dermis; Level 3, tumor filling the papillary dermis and extending to the papillary dermis/reticular dermis interface; Level 4, tumor in reticular dermis; and Level 5, tumor in fat.20 Classification by histologic type was not undertaken, although no melanoma types were excluded.
Statistical Analysis
When we formally assessed the long-term trial results for BCC and SCC to the end of 2004, one of us (G.W.) also carried out a preliminary intention-to-treat analysis on the basis of accumulated but unreviewed melanoma reports for all body sites. The decision then was made to evaluate melanoma occurrence classified according to invasiveness up to December 2006, because 15 years of follow-up (calendar time) was deemed sufficient to detect an effect of sunscreen, if present. On the basis of the observed rate of melanoma in the control group to 2004, the power was estimated to be 66%, and 50%, for detecting hazard ratios of 0.3 and 0.4, respectively, with a two-sided α of .05. As for the other skin cancer end points,13 melanomas diagnosed in the first year of intervention were excluded a priori, because their development was unlikely to have been affected by the introduced sunscreen treatment. Cox proportional hazards regression, with the sunscreen and beta carotene interventions as two main effects, was used to examine treatment effects in relation to primary melanoma occurrence with incorporation of lead time. Individual effects of sunscreen and beta carotene were tested by using likelihood ratio tests. Subgroup analyses were performed to assess consistency of effect according to age, sex, phenotype, sun exposure, and history of skin cancer by using Cox regression and by incorporating a subgroup interaction term to detect heterogeneity of effects.
During trial follow-up, mean hours each day spent outdoors on weekdays and on weekends were calculated, and sun protection habits were assessed from questionnaire responses. Sun exposure and protection durations were compared across the two sunscreen treatment groups by using a two-sample t test. All reported P values were two sided.
RESULTS
Balance was achieved with respect to established risk factors for skin cancer and melanoma among the Nambour trial participants randomly assigned to sunscreen or control in 1992 (Table 1). Compliance with sunscreen treatment, assessed by average of reported frequencies of application, measured weights of returned sunscreen bottles, and diaries, was approximately 75%,13 and 25% of the intervention group applied sunscreen to trunk and/or lower limbs as well as to the intervention sites.21 The majority of participants in the control group either did not apply sunscreen (38%) or applied it once or twice a week at most (35%), and 8% applied it to nonintervention sites.21 Compliance was approximately 70% for beta carotene and placebo supplementation.13
View this table:Table 1.Demographic and Clinical Characteristics of Participants at Baseline in 1992 According to Sunscreen Allocation
By the end of 2006, 846 people (52%) were actively completing questionnaires, 600 (37%) were passive participants, and 173 (11%) had died (n = 87, sunscreen group; n = 86, controls; n = 71, beta carotene group; n = 102, placebo controls), including one person who died as a result of melanoma diagnosed in 1978. One person from each sunscreen treatment group had withdrawn their permission for passive follow-up, in 2002 and 2001, respectively, and data for both were censored accordingly. There was no significant difference in mode of follow-up in relation to sunscreen allocation: duration of active follow-up was 14.3 person-years (95% CI, 14.1 to 14.4 person-years) versus 14.2 person-years (95% CI, 14.0 to 14.3 person-years), and active response rates were 94.2% and 94.3% in each group, respectively. On the basis of reports of active participants, 25% of those randomly assigned to daily sunscreen continued to use sunscreen on a regular basis after the trial22 compared with 18% of the nonintervention group (P = .004).15 By 2001, less than 3% of either supplement group took beta carotene supplements.
In the almost 15 years from commencement of the trial in March 1992 until the end of follow-up in December 2006, 36 of the 1,621 trial participants developed first primary melanomas (n = 22, in situ; n = 14, invasive; none metastatic), and one person in each trial arm developed two primary melanomas. Three people (n = 1, intervention group; n = 2, control) who had melanomas diagnosed in 1992 were excluded a priori. In the remainder of the trial from 1993 to 1996, two participants in the daily sunscreen group and seven in the discretionary group were diagnosed with melanoma (Fig 2). From trial cessation until the end of 2006, nine more participants allocated to daily sunscreen and 15 allocated to discretionary use were diagnosed with incident melanomas (Fig 2). In all, 11 trial participants in the sunscreen intervention and 22 in the control group (Table 2) were newly diagnosed with primary melanoma between 1993 and 2006. Risk of melanoma overall was reduced in those randomly assigned to daily sunscreen compared with discretionary use (hazard ratio [HR], 0.50; 95% CI, 0.24 to 1.02; P = .051), although the result was of borderline statistical significance. Invasive melanoma was reduced by 73% in the daily sunscreen group (HR, 0.27; 95% CI, 0.08 to 0.97; P = .045; Table 2); average thickness was 0.53 mm in the sunscreen group and 1.2 mm in controls (P = .08) on the basis of the original pathology reports (except for one level 2 melanoma in the control group for which thickness measurement was unavailable). There were no significant differences between intervention and control arms with respect to either in situ melanomas (8 v 11, respectively; HR, 0.73; 95% CI, 0.29 to 1.81) or melanomas on prescribed application sites (HR, 0.46; 95% CI, 0.17 to 1.20). When a multivariate proportional hazards regression was carried out that included sex, skin type, numbers of nevi, previous history of skin cancer, and sun exposure along with the two treatment categories, the overall effect estimate for sunscreen varied little (HR, 0.49; 95% CI, 0.24 to 1.02). Regarding outcome according to beta carotene randomization, 16 and 17 melanomas occurred in those taking active and placebo supplements, respectively (HR, 0.89; 95% CI, 0.45 to 1.76).
Fig 2.Occurrence of first primary melanoma by level of invasion and anatomic site in the two sunscreen treatment groups.
View this table:Table 2.First Primary Melanomas During 1993-2006 According to Randomized Sunscreen Intervention During 1992-1996 and Risk of Melanoma
No significant interactions with any baseline characteristics were found (Fig 3). Sun exposure was similar between the daily and discretionary sunscreen groups during the trial (79% and 77%, respectively, spent less than 50% of weekend time outdoors23) or after the trial (3.8 and 3.9 hours each day, respectively, spent outdoors on weekdays [P = .46]; 4.6 and 4.7 hours each day, respectively, spent outdoors on weekends [P = .79]). Use of sun protection measures other than sunscreen was similar during the trial (approximately 60% of both groups usually sought shade; around 75% usually wore a hat) and after the trial (at midpoint of follow-up, 40% and 67% usually sought shade and wore a hat in the sun, respectively).
Fig 3.Effect of sunscreen intervention on melanoma according to baseline characteristics. Hazard ratios are for melanoma in a comparison of the sunscreen intervention and control groups. Hazard ratios, 95% CIs, and P values were calculated by using Cox regression that incorporated a subgroup interaction term to detect heterogeneity of effects.
DISCUSSION
Long-term follow-up of this randomized trial showed that, among adults age 25 to 75 years, regular application of SPF 15+ sunscreen in a 5-year period appeared to reduce the incidence of new primary melanomas for up to 10 years after trial cessation. A protective effect was also evident for invasive melanoma, which showed a 73% decrease in those randomly assigned to daily sunscreen after approximately 15 years of follow-up. This result for the invasive subgroup was an exploratory finding, however, and should be interpreted cautiously. The apparent decrease in melanoma across all body sites, not only prescribed application sites, likely reflected the ongoing tendency of more participants in the daily than the discretionary sunscreen group to apply sunscreen regularly to the trunk and lower limbs.21,22 Intensity of application also tended to be higher in the intervention group.21
There was no evidence to suggest that the observed difference could be explained by a difference in sun exposure in the intervention group relative to the control group, because outdoor behavior was similar during23 and after the trial. Similarly, there were no differences in active completion of follow-up questionnaires with self-reports (later validated) of all new skin cancers, including melanomas. Without reference to treatment groups, we attempted to capture all melanomas diagnosed between 1993 and 2006 among passive trial participants (except for two who withdrew consent for follow-up) from the Queensland Cancer Registry as well as from the pathology laboratories. Ultimately, however, the final diagnosis of each melanoma reported during or after the trial was determined by the two reviewing dermatopathologists (one who was highly experienced in melanoma diagnosis) who were unaware of allocated treatment groups. There were technical limitations of the histopathologic review, because it was based on the slides available and because much of the material was archival. Even when tissue blocks were available, the remaining tissue might not have been representative of the tumor as a whole. We had no information on presence of ulceration or mitotic figures; definitive assessment of thickness of these study melanomas required reference to original diagnostic pathology reports, because original slides or tissue blocks no longer existed in many cases. Level of invasion also may have been subject to some uncertainty, though notdifferentially according to sunscreen, because the dermatopathologists had no knowledge of treatment allocation.
With regard to the beta carotene intervention, we found no evidence of harmful effect on melanoma or other health outcomes (mortality overall or as a result of other cancers). Although we observed fewer deaths in the intervention than in the control group on follow-up, we observed no specific beneficial effect on cancer outcomes, which is in agreement with two other randomized trials of beta carotene supplementation with follow-up24,25 and with a recent meta-analysis.26
Though this community-based skin cancer prevention trial is the only one of its kind, its findings that suggest the general preventability of melanoma after the regular application of broad-spectrum sunscreen have been predicted.11,16 They accord with those of a trial of sunscreen to prevent melanocytic nevus development in children,12 despite the limited number of children involved and the brief follow-up.27 The results are consistent with the knowledge that excessive sun exposure causes melanoma,5 notwithstanding the evidence that solar UV carcinogenesis may operate by different pathways to cause melanomas on different body sites.28,29 Results also accord with background experimental data in mice,9 including prevention of p53 mutations in UV-irradiated skin by sunscreen.30 In addition, in people with sun-sensitive skin types, daily-care sunscreen can inhibit clinical, cellular, and molecular damage caused by daily suberythemal solar-simulating radiation.31
Given the importance of early-life sun exposure in the genesis of melanoma,32 a long-term sunscreen intervention among children and adolescents may yield even greater benefits in cancer prevention33 than did this intervention in adults. The adult participants in Nambour would have experienced relatively high ambient sun exposure for years, so skin carcinogenesis may already have been initiated in many of them; only the promoting effects of ongoing adult sun exposure would have been targeted by this intervention.
In conclusion, our findings provide reassurance in view of the widespread uncertainty to date about sunscreen's ability to prevent melanoma.7,34 Although the results are directly relevant to people who live in sunny climates like Australia's and who receive relatively high levels of ambient sun exposure as a matter of course, they also have implications for white people living in temperate climates in North America and Europe35 who are at increased risk of melanoma because of increased solar UV exposure caused by the predilection for holidays in sunny places.5,36
———————————————————————————————————————-
http://jco.ascopubs.org/content/29/3/249
- © 2010 by American Society of Clinical Oncology
Sunscreen and Melanoma: What Is the Evidence?
– Author Affiliations
In this issue of Journal of Clinical Oncology, Green et al1 report on an extension study of a randomized controlled trial of daily sunscreen application and beta carotene supplementation2 that examined the melanoma experience of 1,621 patients who were enrolled onto that trial between 1992 and 1996 and who were subsequently observed for 10 years (1996 to 2006). The two interventions, sunscreen application and beta carotene supplementation, were evaluated both alone and in combination, relative to a placebo-control group. The primary end points used in the evaluation of these interventions were the incidence of basal cell carcinoma and the incidence of squamous cell carcinoma after 4.5 years of follow-up. In their current report, the authors focus on the incidence of in situ and invasive melanoma during the 15 years of trial participation as a secondary end point, and use it to test the hypothesis that “regular sunscreen use by white adults prevents the occurrence of primary cutaneous melanoma, with a possible latent effect of up to 10 years.”1 The P value associated with the Cox regression analysis used to test the hypothesis that the risk of in situ and invasive melanoma together differed between those who used the sunscreen and those who did not was .051, whereas the P value for the risk of invasive melanoma alone (an end point less confounded by variable interpretations of pathologists) was .045. Both P values could be considered of borderline significance. The authors conclude that “our findings provide reassurance… about sunscreen's ability to prevent melanoma.”
There seems to be some contradiction between the authors' conclusions and the P values that were reported. This brings to mind an article written by Goodman,3 wherein he reminds us that P values have historically been considered in two different conceptual frameworks. The first framework, as originally proposed by Fisher, was that the P value is an index or measure of evidence that can be used to measure disagreement between what is observed and what would be expected under a null hypothesis. One interpretational weakness of this approach is that it does not consider the effect size; the P value for a large effect in a small sample could be the same as the P value for a small effect in a large sample. Consequently, evidence provided by the P value alone is sometimes limited. It is important to consider additional evidence from the data beyond the P value, such as within-group estimates relevant to the clinical end point and their difference, as well as their 95% CIs.
The second framework described by Goodman3 is the one proposed by Neyman and Pearson, wherein a significance level is defined within the context of the assessment of the errors associated with a statistical hypothesis test. Here, two hypotheses are defined, the null and the alternative, and the statistical test defines one of two behaviors, to reject the null hypothesis or to accept the alternative hypothesis. The use of the statistical hypothesis test involves two types of possible errors: rejecting the null hypothesis when it is true, or accepting the alternative hypothesis when it is false, because the first error has been fixed to be .05 (the significance level generally assumed). Observed P values are then compared with the significance level. What this second framework considers, that the first does not, is the power of the statistical test. The confidence in a statistical test can be increased by using a power calculation to determine the sample size necessary to ensure that a meaningful effect can be detected; this effect is specified by the alternative hypothesis when the alternative hypothesis is true.
Next, we look at the data presented in Green et al1 from these two perspectives. What level of evidence was presented from this trial about the relationship between the regular use of sunscreen and the incidence of melanoma? The authors opted to present hazard ratios to describe the incidence of melanoma in the two groups. They reported a 50% reduction in the hazard related to in situ and invasive melanoma and a 73% reduction in the hazard related to invasive melanoma compared with the hazards of patients who were not assigned to the intervention. This information is less useful for weighing the evidence for regular sunscreen use and its prevention of melanoma in the absence of information about the relevant hazards in the nonintervention group. Ideally, we would have been given estimates of those hazards as well. For this discussion, we used the data in the authors' Table 2 to estimate rates for the observed period and assumed an intention-to-treat analysis (Table 1). The P values for Pearson's χ2 statistic under the assumption of equal rates in the two groups were .0517 and .0312, on the basis of rates defined by in situ and invasive melanomas together as well as rates for invasive melanomas only, respectively, which provided evidence that both χ2 statistics had a much higher value than one would expect according to the null hypothesis. However, more convincing evidence would involve translation of these estimates to what might be observed in a similar population. For example, in 100,000 adults in Queensland with baseline characteristics similar to those of the patients in the study by Green et al, during a period similar in length to that of the study by Green et al, all other characteristics being equal, one would expect a reduction in the number of in situ and invasive melanomas to be 1,370 (95% CI, 1,030 to 1,850) and the reduction in the number of invasive melanomas to be 990 (95% CI, 600 to 1,340) for those using sunscreen compared with those not using sunscreen.
View this table:Table 1.Group-Specific Rates, Differences, and Estimated Reduction During a Period of 15 Years for In Situ and Invasive Melanoma and Invasive Melanoma (according to intention-to-treat analysis)
What evidence do we have that the statistical hypothesis tests that were performed had sufficient power to detect a meaningful effect? On the basis of data collected up to 2004, the authors report that the power of their statistical test would be 50% to detect a hazard ratio of 0.4 on the basis of an intention-to-treat analysis for the primary end point. Because the power was lower than one would like and the observed effect size in this study (0.5) was smaller than the one assumed in 2004, an updated power calculation performed at the end of the extension study would have been informative for assessing the quality of the statistical test. For our example above, we calculated the power for the two χ2 tests using the trial sample sizes and the observed rates in the two groups to define the effect size. The power for the χ2 test performed to investigate the difference in the rates of both in situ and invasive melanomas in the two groups was 43%, whereas the power of this test to investigate differences in the invasive melanoma rates was 98%. A statistical test can be thought of as a diagnostic test, and its power is similar to the sensitivity (or specificity) of a diagnostic test. Just as with diagnostic tests, in which higher sensitivity and specificity provide greater confidence in the test, a statistical test with a power of 98% provides greater confidence than a statistical test with a power of 43% that the null hypothesis will be rejected when the alternative hypothesis is true.
What, then, is our bottom line? To our knowledge, the trial's findings are the first to provide strong evidence for a reduction in the incidence of invasive melanoma after regular application of broad-spectrum sunscreen in adults. The trial was an ambitious and unique study: it was conducted in a region with the highest rate of skin cancer in the world, had a follow-up period of 10 years after the trial, and achieved relatively high rates of compliance among the participants assigned to the group using sunscreen.4 It is unlikely that another trial of comparable scope and rigor will be conducted in the foreseeable future. Therefore, the question as to how these findings should be translated into clinical practice is important.
Do these findings end the debate about whether sunscreen use can prevent melanoma? Should clinicians and public health educators recommend daily application of sunscreen to the heads, necks, arms, and hands of many of the general population? In fact, sunscreen is the form of sun protection most often used, and is used by people of all ages,5 although its use is suboptimal. Two community surveys found that 76% to 93% of pediatricians routinely recommend that sunscreen be used when children are outside,6,7 although they usually do not give advice about optimal use and reapplication. Given current knowledge, we propose that the following are the most important clinical applications of the new findings from the trial by Green et al.1 First, clinicians should advise patients at high risk for skin cancer because of phenotypic characteristics (fair skin, freckling, tendency to sunburn, and so on), who live in or visit sunny climates, and/or who have a family history of melanoma, to routinely and thoroughly apply sunscreen before going outside. Second, public health and cancer prevention agencies should provide clear instructions regarding the use and reapplication of sunscreen in the manner described in the study by Green et al. Third, highly exposed and at-risk individuals should consider making regular sunscreen use a habit, much like other health routines; in addition, parents should apply sunscreen to their children's skin and should model the practice of sunscreen use.
Sunscreen use alone will likely not reduce the incidence of skin cancer, which is an increasingly widespread and serious public health problem in the United States and globally. In addition to sunscreen use, excess exposure to ultraviolet rays should be avoided, clothing should be used to shield skin from the sun, and sun-safe environments should be used for outdoor recreation. In addition, sunscreen use should be paired with regular self-examination of the skin. The question of its efficacy with respect to melanoma prevention should no longer deter scientists or clinicians from recommending sunscreen use.
AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST
The author(s) indicated no potential conflicts of interest.
AUTHOR CONTRIBUTIONS
Conception and design: Phyllis A. Gimotty
Data analysis and interpretation: Phyllis A. Gimotty
Manuscript writing: Phyllis A. Gimotty, Karen Glanz
Final approval of manuscript: Phyllis A. Gimotty, Karen Glanz
REFERENCES
- CiteULike
- Connotea
- Delicious
- Digg
- Facebook
- Reddit
- Technorati
- Twitter
Articles citing this article
- © 2006 by American Society of Clinical Oncology
Anatomic Site, Sun Exposure, and Risk of Cutaneous Melanoma
- David C. Whiteman,
- Mark Stickley,
- Peter Watt,
- Maria Celia Hughes,
- Marcia B. Davis and
- Adèle C. Green
+ Author Affiliations
- Address reprint requests to David C. Whiteman, MBBS, PhD, Division of Population Studies and Human Genetics, Queensland Institute of Medical Research, PO Royal Brisbane Hospital, Queensland 4029 Australia; e-mail: [email protected]
Abstract
Purpose Sunlight is the principal environmental risk factor for cutaneous melanoma. A current hypothesis postulates that the role of sunlight in causing melanoma differs according to anatomic site. We tested this hypothesis in a population-based case-case comparative study of melanoma patients.
Methods Patients were sampled from the Queensland cancer registry in three groups: superficial spreading or nodular melanomas of the trunk (n = 154), of the head and neck (HN; n = 76), and lentigo maligna (LM) and lentigo maligna melanoma (LMM; for both LM and LMM, n = 76). Data were collected on school-age sun exposure and occupational and recreational sun exposure in adulthood. Odds ratios (OR) and 95% CIs were calculated using polytomous logistic regression.
Results HN melanoma patients were substantially more likely than trunk patients to have higher levels of sun exposure in adulthood (OR, 2.43; 95% CI, 0.98 to 5.99) and specifically, higher levels of occupational exposure (OR, 3.25; 95% CI, 1.32 to 8.00), but lower levels of recreational sun exposure (OR, 0.50; 95% CI, 0.21 to 1.19). LM and LMM patients reported higher occupational exposure and lower recreational sun exposure than trunk melanoma patients, although this was not significant. We found no significant differences between the groups for school-age sun exposures.
Conclusion Melanomas developing at different body sites are associated with distinct patterns of sun exposure. Melanomas of the head and neck are associated with chronic patterns of sun exposure whereas trunk melanomas are associated with intermittent patterns of sun exposure, supporting the hypothesis that melanomas may arise through divergent causal pathways.
Footnotes
-
Supported by grants from the Queensland Cancer Fund and the National Cancer Institute (CA 88363-01A1). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute.
Presented in part during an invited plenary session at the 6th World Congress on Melanoma, Vancouver, British Columbia, Canada, September 8, 2005 (D.C.W.).
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
- Received February 9, 2006.
- Accepted April 21, 2006.
-
- You must be logged in to reply to this topic.