A Randomized, 12-Year Primary-Prevention Trial of Beta Carotene Supplementation for Nonmelanoma Skin Cancer in the Physicians' Health Study

Context  Although basic research provides plausible mechanisms for benefits of beta carotene supplementation on nonmelanoma skin cancer (NMSC) primarily consisting of basal cell carcinoma (BCC) and squamous cell carcinoma (SCC), observational studies are inconsistent. Randomized trial data are limited to 1 trial of secondary prevention that showed no effect of beta carotene on the incidence of NMSC after 5 years.

Objective  To test whether supplementation with beta carotene reduces the risk for development of a first NMSC, including BCC and SCC.

Design  Randomized, double-blind, placebo-controlled trial with 12 years of beta carotene supplementation and follow-up.

Setting  Physicians' Health Study in the United States.

Participants  Apparently healthy male physicians aged 40 to 84 years in 1982 (N = 22 071).

Intervention  Beta carotene, 50 mg, on alternate days.

Main Outcome Measure  Relative risk (RR) and 95% confidence interval (CI) for a first NMSC, BCC, and SCC.

Results  After adjusting for age and randomized aspirin assignment, there was no effect of beta carotene on the incidence of a first NMSC (RR, 0.98; 95% CI, 0.92-1.05), BCC (RR, 0.99; 95% CI, 0.92-1.06), or SCC (RR, 0.97; 95% CI, 0.84-1.13). There was also no significant evidence of beneficial or harmful effects of beta carotene on NMSC by smoking status (current, past, or never).

Conclusion  This large-scale, randomized, primary prevention trial among apparently healthy well-nourished men indicates that an average of 12 years of supplementation with beta carotene does not affect the development of a first NMSC, including BCC and SCC.

BASAL CELL carcinoma (BCC) and squamous cell carcinoma (SCC), usually referred to as nonmelanoma skin cancer (NMSC), are the most common malignant neoplasms among white patients. Worldwide, including in the United States, the incidence of NMSC is increasing.^1,2 Although NMSC has an extremely low case fatality rate, it is a major cause of morbidity, including notable disfigurement, particularly as most lesions occur on the face, head, or neck.³

Carcinogenesis of the skin is thought to be initiated by UV light–induced DNA mutations.⁴ It has been suggested that antioxidant vitamins could prevent cancer-causing damage to DNA by scavenging free radicals or excited oxygen molecules. Beta carotene can quench free radicals or inhibit their formation and may therefore reduce DNA damage.^5,6 Several other plausible biological mechanisms have been hypothesized to explain the possible chemopreventive properties of beta carotene.⁷ These include stimulation of the immune system⁸ and enhancement of cell-to-cell communication.⁹ Beta carotene might also affect cell proliferation and differentiation through its retinoid precursor function.¹⁰

Animal studies of UV light–induced skin cancer have provided consistent evidence of an anticancer effect of carotenoids. Epstein,¹¹ Mathews-Roth,¹² Mathews-Roth and Krinsky,¹³ and Santamaria et al¹⁴ found that beta carotene supplementation significantly decreased the number of skin tumors induced by exposing mice to UV irradiation. Lambert et al^15,16 reported a beneficial effect of dietary beta carotene against 2-stage carcinogenesis in mice, causing a delayed onset and a reduced number of skin cancers.

Observational studies on the relationship of dietary intake or blood levels of beta carotene with subsequent risk for NMSC incidence have yielded inconsistent results. Although some authors^17,18 have suggested that individuals who select diets high in beta carotene have a lower incidence of NMSC; others^19-23 have shown no effect. One previous randomized trial²⁴ of secondary prevention showed no effect of beta carotene supplementation on the incidence of NMSC after 5 years. A recently completed trial²⁵ of beta carotene and sunscreen in primary prevention of NMSC among 1626 participants treated for 4.5 years in Nambour, Australia, also showed no effect of beta carotene.

The Physicians' Health Study (PHS), a large-scale, randomized, double-blind, placebo-controlled trial among 22 071 apparently healthy men, provided a unique opportunity to assess any effects of 12 years of beta carotene supplementation (50 mg every other day) on risks for development of a first NMSC.

The PHS was a randomized, double-blind, placebo-controlled, 2 × 2 factorial trial that tested the effect of aspirin and beta carotene in the primary prevention of cardiovascular disease and cancer among 22 071 apparently healthy male US physicians. The trial design, subjects, and methods have been detailed in previous reports.^26-28 Participants were aged 40 through 84 years in 1982, with no history of cancer (except possibly NMSC), myocardial infarction, stroke, or cerebral ischemia, and were not currently taking aspirin or other platelet-active medications or vitamin A supplements. Subjects with prevalent NMSC were not excluded from the trial, because this diagnosis was not specified as a primary end point of the PHS, which was chiefly focused on the prevention of noncutaneous malignant neoplasms and cardiovascular disease. Participants were randomly assigned to take 325 mg of aspirin (Bufferin; Bristol-Myers Squibb Company, New York, NY) or placebo every other day along with 50 mg of beta carotene (Lurotin; BASF Corporation, Ludwigshafen, Germany) or its placebo on the alternate days. A total of 11 036 physicians were assigned to active beta carotene treatment, and 11 035 to beta carotene placebo treatment. The randomized aspirin component of the trial was terminated early principally owing to the emergence of a statistically extreme (P<.00001) 44% reduction in the risk for a first myocardial infarction in the aspirin group.^27,28 The randomized beta carotene component of the trial continued uninterrupted until its scheduled termination on December 31, 1995. After 12 years of supplementation with beta carotene, there was no significant evidence of benefit or harm concerning cancer or cardiovascular disease, including among current smokers and nonsmokers.²⁶ In this primary analysis of the trial data, NMSC was not considered, because it was not included in the definition of the primary study end point for the PHS trial.²⁶

By the end of 1995, participants had been taking beta carotene or placebo for an average of 12 years (range, 11.6-14.2 years). At the end of 11 years of follow-up (the last year completed for all participants) 99.7% were still providing morbidity information, and vital status was known for all except 1 participant. In the beta carotene and placebo groups, 81.4% of participants who responded to the last questionnaire were still taking an average of 97% of their study pills. The remaining 18.6% of participants in each group were not taking any study pills. Overall, 78% of the study pills were reported as being taken in the beta carotene group, and 6.4% of the physicians in the placebo group reported taking nonstudy beta carotene or vitamin A supplements.²⁶

Participants were sent follow-up questionnaires at 6 months and annually thereafter, asking about the occurrence of any relevant events as well as their compliance with the assigned pill-taking regimen. By using 3 mailings plus an additional telephone contact of nonrespondents, response rates at each survey were greater than 98%. Information on occurrence of any NMSC was obtained on the annual questionnaires. All reported diagnoses of cancer except the numerous reports of BCC were followed up by review of the medical records, which were requested from hospitals and treating physicians after receiving the participant's consent. An end points committee of physicians reviewed all pathological reports.

To assess the validity of self-report of BCC in the PHS, records were obtained for 29 participants who reported a diagnosis of NMSC. Among these, 16 reports of BCC were all confirmed. The 13 remaining reports included 1 SCC and 1 keratoacanthoma, which were confirmed on medical record review, and 1 report of 2 BCCs and 1 SCC in the same participant, which were confirmed as 3 BCCs. The other 10 reports of a nonspecific diagnosis of skin cancer were confirmed as 8 BCCs and 2 SCCs. Based on these results, occurrence of BCC was assessed by self-report.

In this report, the main end point was any first NMSC, defined here as BCC or SCC. Because BCC and SCC may arise through different biological pathways,²⁹ and because risk factors may differ,³⁰ we also assessed both histological types separately. Reports of Bowen disease, an early or in situ SCC, were classified as SCC. The few cases of mixed BCC and SCC were only considered in analyses of total first NMSC.

Of the 22 071 physicians, 187 reported an NMSC diagnosed before their randomization into the trial. These physicians were excluded from further analyses. Data from 21 884 men with no history of NMSC were analyzed on an intention-to-treat basis according to randomized treatment assignment. Person-time was censored at the time of the first recorded NMSC, an unrefuted report of death, or the date of the last completed follow-up questionnaire, whichever was earliest. Age-specific incidence rates were calculated within 5-year age groups and standardized to the age distribution of the 1988-1992 US white male population.³¹

Relative risks for development of NMSC among subjects randomized to beta carotene supplementation compared with those assigned to placebo were estimated using Cox proportional hazards models adjusted for aspirin assignment and age at baseline. Despite the randomized design and large sample size, we conducted further analyses, which included adjustments for smoking history (never, past, or current) and state of residence (north, middle, or south) based on the latitude. Latitudes were taken as those of each state's capital city. We also considered whether these factors might modify a potential effect of beta carotene on the incidence of NMSC. To evaluate the possibility of a delayed effect of beta carotene supplementation, we subdivided the follow-up time into 4-year time periods (0-4, 5-8, and ≥9 years). In an additional analysis, NMSC diagnosed within 2 years after randomization were excluded, since many of these may have been present as preclinical disease at baseline.

Because medical records for BCC (most NMSC in this study) were not reviewed, we performed a sensitivity analysis to determine to what extent misclassification of end points might have influenced our findings. Since the data were from a large, double-blind, randomized trial, we used the approach described by Greenland³² for calculation of nondifferential disease misclassification in person-time follow-up data.

Table 1 shows the baseline characteristics of the study participants according to randomized assignment to beta carotene. As expected in this very large trial, the distribution of baseline characteristics was virtually identical in both treatment groups. At baseline in 1982, 49.5% of the participants had never smoked, 39.2% were past smokers, and 11.2% were current smokers. The distribution of the participants with regard to geographic areas was almost equally balanced (Table 1). During an average of 12 years of supplementation and follow-up, we documented 3607 cases of NMSC. Among these, 1786 were reported in the beta carotene group and 1821 in the placebo group (relative risk [RR], 0.98; 95% confidence interval [CI], 0.92-1.05) (Table 2). Direct adjustment to the 1988-1992 US white male population revealed an age-specific incidence rate of 533 per 100 000 person-years (95% CI, 516-550) for all NMSC combined. The corresponding rates for a first BCC were 464 per 100 000 person-years (95% CI, 448-480); for a first SCC, 95 per 100 000 person-years (95% CI, 88-102). There was no significant effect of beta carotene for the end points of BCC (RR, 0.99; 95% CI, 0.92-1.06) or SCC (RR, 0.97; 95% CI, 0.84-1.13). As expected in a large randomized trial, adjustment for age, randomized aspirin assignment, smoking, and latitude did not substantially alter these results (data not shown). Similarly, exclusion of the first 2 years after randomization to eliminate possible preexisting malignant neoplasms had no effect on these results. There was also no significant evidence of benefit or harm due to supplementation with beta carotene in the different time periods.

Among the subgroups of smokers, we did not observe any indication of an effect of beta carotene on risk for NMSC, including BCC and SCC (Table 3). Although NMSC increases markedly with decreasing latitude,^30,33 there was no evidence of an association between beta carotene and incidence of NMSC in data stratified by region of residence (data not shown).

In sensitivity analyses to assess the potential impact of nondifferential misclassification of disease end points, we continued to observe a null result (Table 4). Assuming a sensitivity of 75% and 100 misclassified cases in beta carotene and placebo groups showed an unchanged RR of 0.98 for development of a BCC in the beta carotene vs the placebo group. Even if we assumed that more than half of the reports in each group were false-positive findings and that we failed to detect 25% of true cases, the RR changed only minimally to 0.97. Results of sensitivity analyses were similar for SCC and the combined number of all NMSC. We did not perform sensitivity analyses for differential misclassification because, given the randomized, double-blind, placebo-controlled design of the trial, this type of misclassification is extremely unlikely.

During the trial, no major side effects were associated with the average of 12 years of beta carotene supplementation.²⁶ Minor adverse effects included yellowing of the skin (1745 [15.9%] in the beta carotene vs 1535 [14.0%] in the placebo group) and minor gastrointestinal tract symptoms such as belching (275 [2.5%] in the beta carotene vs 124 [1.1%] in the placebo group). These findings have been described previously with this dose and formulation of beta carotene.²⁴

This large-scale randomized trial among apparently healthy, well-nourished men provides substantial evidence of no significant beneficial or adverse effects of beta carotene supplementation during a 12-year time period on the risk for NMSC, including BCC and SCC. Furthermore, there was no evidence of a trend toward beneficial or harmful effects during the follow-up period. Results were also similar regardless of smoking status.

Among previous observational studies, a retrospective case-control study among Australian veterans showed that high blood levels of beta carotene were associated with a significantly lower risk for NMSC.¹⁷ A small, nested case-control study in Taiwan showed that patients with lower blood levels of beta carotene had a significantly higher risk for arsenic-induced NMSC compared with healthy controls.¹⁸ Several other prospective, nested case-control studies using baseline blood levels of beta carotene or food questionnaires showed no significant relationship of beta carotene with NMSC,^19-23 including the largest of all such studies.²¹ Observational studies tend to compare subjects with the highest vs the lowest levels in the range of usual dietary intake. In contrast, the dose of beta carotene in randomized trials tends to be several times higher than the usual intake (less than 3 mg daily in the US population³⁴) and sufficient to put the subjects into the top small percentage of usual intake. The only reported randomized trial of secondary prevention, the Skin Cancer Prevention Trial, tested daily supplementation with 50 mg of beta carotene in 1805 patients with a previously diagnosed NMSC. Patients assigned to beta carotene therapy had no decrease in the occurrence of new NMSC during a 5-year period, nor was there any adverse effect.²⁴ However, because there is a long latent period between UV exposure and the development of human skin cancer,² a 5-year treatment duration might not have been sufficient to show any effect of beta carotene supplementation.

Sunlight reportedly acts twice to cause skin cancer: first to mutate the p53 gene and then to set up conditions for the unrestrained growth of the altered cell line.^35,36 In animal models of skin cancer, beta carotene was fed before, during, and shortly after exposure to UV light. Possibly, the early onset of feeding supplementary beta carotene prevented the first step of carcinogenesis in these studies. Thus, timing of beta carotene supplementation might be an important consideration. Despite the long duration of our study, supplementation may have been initiated too late to prevent carcinogenesis in our trial population. Consequently, it remains to be determined whether beta carotene might be effective in preventing the earlier stages of NMSC development in humans. If so, a small to moderate benefit could still emerge with continued follow-up of the PHS cohort, which is ongoing.

It is also possible that any benefit of supplementation might be restricted to those with low blood levels of beta carotene at baseline. In this regard, beta carotene supplementation was associated with lower rates of prostate and total cancer in a subgroup of PHS participants with low baseline blood levels of beta carotene.³⁷ In the Chinese Cancer Prevention Trial, a randomized trial among a poorly nourished rural population in Linxian, China, those assigned to a daily supplement of combined beta carotene (15 mg), vitamin E (30 mg), and selenium (50 µg) had a significant decrease in total mortality, largely resulting from a reduction in stomach cancer.³⁸ With respect to NMSC specifically, Hsueh et al^18,39 observed that poor nutritional status, including inadequate beta carotene intake, might be related to the development of arsenic-induced NMSC. Karagas et al²³ found that individuals with the highest compared with the lowest quartile of blood beta carotene levels had a possible but nonsignificant lower risk for NMSC (RR, 0.73; 95% CI, 0.38-1.41). In contrast, however, the single reported randomized trial of secondary prevention of NMSC showed no benefit among the subgroup with the lowest baseline blood levels of beta carotene after 5 years of supplementation.²⁴ In the setting of primary prevention, it seems plausible that any benefit of beta carotene might be confined to subjects with the lowest blood levels of beta carotene at baseline. We were not able to test this hypothesis directly, since we currently do not have baseline blood levels of beta carotene for all study participants.

Adverse effects of beta carotene were observed among cigarette smokers in other trial populations.^40,41 In our trial, 11.1% of participants were current smokers at baseline, yielding about 29 000 person-years of observation. We observed no significant evidence of beneficial or harmful effects of beta carotene supplementation with regard to NMSC, regardless of smoking status.

We had no information on sun exposure, skin type, or history of sunburns, which are known risk factors for NMSC. However, since our study was a large randomized trial, these factors are expected to be evenly distributed between the study groups, thereby eliminating the potential for confounding. The fact that the distribution of measured baseline characteristics was virtually identical between the beta carotene and placebo groups suggests that confounding by unmeasured factors is extremely unlikely.

Although the diagnosis of BCC for our study was based solely on self-report, we found no evidence of any major misclassification of reported end points in a small validation study among these male physicians. Similarly, in the Nurses' Health Study, a large cohort of female health professionals, Hunter et al⁴² confirmed 27 of 28 cases reported as BCC. Underascertainment of cases is possible, but NMSC incidence rates in this cohort are higher than those previously reported for the United States,³ and low sensitivity does not bias results in a follow-up study if specificity is high. Moreover, the randomized prospective design of the trial, combined with the medical knowledge of participants and the high follow-up rate, suggest that biased ascertainment of cases is unlikely. To rule out nondifferential misclassification as an explanation for our null findings, we estimated the effect of potential underdiagnosis as well as overdiagnosis of BCC on the RR for beta carotene vs placebo in our study. Even under very extreme assumptions regarding the sensitivity and specificity of BCC diagnosis, the RR remained very close to that obtained in our primary results. This provides some assurance that misclassification of end points is an unlikely explanation for the null finding. There are several unique strengths of the PHS, including its large sample size, randomized design, and high compliance and follow-up rates. However, it is the extremely long duration of treatment and follow-up of 12 years that distinguish this from any other completed or ongoing trial. In these data, beta carotene supplementation had no effect on the primary prevention of NMSC, including BCC and SCC, regardless of smoking history. Continued follow-up of this cohort and the other recently completed primary prevention trial²⁵ will yield further useful information on this question.

Support for a recommendation of a diet rich in fruits and vegetables emanates from observational epidemiological data that consistently show a lower risk for a number of site-specific cancers in persons who eat this type of diet.^43,44 In addition, in a randomized trial, increased fruit and vegetable consumption accompanied by a low-fat diet was associated with a significant reduction in the occurrence of NMSC.^45,46 In light of the results of our trial, it appears that any protection against NMSC offered by increased intake of fruits and vegetables is less likely to be related to beta carotene than to other constituents of this diet. Efforts to curtail the incidence of NMSC should concentrate on risk reduction through limiting the major known risk factor, ie, exposure to sunlight and other sources of UV light.

Accepted for publication June 15, 1999.

This study was supported by investigator-initiated research grants CA-34944, CA-40360, HL-26490, HL-34595, and AR-42689 from the National Institutes of Health and K-08 grant EY-00365 from the National Eye Institute (Dr Schaumberg), Bethesda, Md.

Presented in part at the 29th Annual Meeting of the Society for Epidemiologic Research, Boston, Mass, June 12-15, 1997.

We thank the study participants for their continued participation and cooperation, and Mary Breen, Miriam Schvartz, MD, and in particular Vadim Bubes for their outstanding efforts.

Reprints: Debra A. Schaumberg, ScD, MPH, Division of Preventive Medicine, 900 Commonwealth Ave E, Boston, MA, 02215 (e-mail: dschaumberg@rics.bwh.harvard.edu).