Prospective study of vitamins C, E, and A and carotenoids and risk of oral premalignant lesions in men

Abstract

Case–control studies indicate that vitamins C, E, A and carotenoids decrease risk of oral premalignant lesions (OPLs) and oral cancer, but clinical trials have failed to find protective effects of β-carotene and suggest that vitamin E may increase risk. The authors prospectively evaluated the association between intake of vitamins C, E, A and carotenoids and incidence of OPL. Participants were 42,340 men in the Health Professionals Follow-up Study who provided information on supplement use and diet every 2–4 years by food frequency questionnaire. The authors confirmed 207 clinically or histopathologically diagnosed OPL events occurring between 1986 and 2002 by medical record review. Multivariate-adjusted relative risks (RR) of OPL were calculated with proportional hazards models. Total intake of vitamin C, vitamin A or carotenoids was not significantly associated with OPL risk. Dietary vitamin C was significantly associated with reduced risk (quintile 5 vs. 1, RR = 0.52, 95% CI 0.31–0.85, p_trend = 0.04), but no association with supplemental vitamin C was observed. Inverse associations were apparent for β-cryptoxanthin and α-carotene intake. No clear relationship emerged with β-carotene, lycopene or lutein/zeaxanthin. Vitamin E was associated with increased risk (quintile 5 vs. 1, RR = 1.86, 95% CI 1.06–3.19), particularly among current smokers and with supplemental intake (current-smokers, supplement dose tertile 3 vs. 1, RR = 3.07, 95% CI 1.28–7.34, p_trend = 0.01). For current smokers, β-carotene also increased risk. Vitamin C from dietary sources, but not supplements, was associated with a reduced risk of OPL. The observed increased risk for current smokers with high vitamin E or β-carotene intake should be explored further. © 2006 Wiley-Liss, Inc.

A prominent contributor to the onset and development of oral cancer is damage to DNA and other cellular molecules by reactive oxygen species. Tobacco use, the primary known risk factor for oral cancer and precancerous lesions, also increases oxidative stress and therefore enhances the possibility of cancer-causing mutations, oxidization of lipids and proteins and alteration of signal transduction pathways that damage cells.1, 2 Dietary micronutrients, such as vitamin E, vitamin C, β-carotene and lycopene, have potential chemopreventive properties that can neutralize or block reactive oxygen species and protect against cellular damage.3, 4, 5

Observational studies suggest that intake of antioxidants, primarily through consumption of fruits and vegetables, is associated with a decreased risk of oral cancer, but the evidence has been inconsistent.6, 7, 8, 9, 10, 11 Recent large clinical trials, however, saw significantly increased risks of head and neck cancer recurrence and second primary tumors with α-tocopherol (vitamin E) supplementation,12 and failed to find any benefits of supplementation with vitamin A in oral cancer patients.13 Such results suggest that more compelling evidence is needed before large-scale, randomized studies are initiated. Epidemiological studies of nutrients and oral cancer have been limited to the case–control design, which is prone to selection and recall bias that often lead to an overestimation of protective effects. Furthermore, oral cancer has a relatively noticeable presentation; malignant lesions may be symptomatic, affecting the physical sensation and palatability of food and thus causing changes in nutrition that can similarly bias results.14, 15

All oral malignancies arise from premalignant precursors, which are most often asymptomatic and unlikely to affect diet.16, 17 Epidemiologic studies have found that 16–40% of oral premalignant lesions (OPLs) transform to cancer.18, 19, 20, 21 An investigation of the relationship between nutrient intake and OPL may therefore clarify the role of nutrients in cancer development. Thus far, few epidemiological studies have considered intake of specific nutrients and OPL risk, and all have used the case–control approach in populations where supplements are rarely used but tobacco is widespread. The limited evidence from these studies suggests that β-carotene and vitamin C may be inversely related to risk.22, 23, 24, 25, 26, 27 To clarify the possible role of vitamin C, vitamin E, vitamin A and the individual carotenoids in the risk of OPL, we prospectively evaluated nutrient intake in a large cohort of men in the United States. We also examined whether the association with OPL varied by dietary or supplement intake, tobacco use and timing of diet assessment.

OPL, oral premalignant lesion; HPFS, Health Professionals Follow-up Study; RR, relative risk; CI, confidence interval; USDA, US Department of Agriculture; FFQ, food frequency questionnaire.

Material and methods

Study population: Health Professionals Follow-Up Study

The Health Professionals Follow-Up Study (HPFS) is an ongoing prospective cohort study of 51,529 US male health professionals (58% dentists, also optometrists, osteopaths, pharmacists, podiatrists and veterinarians) who were aged 40–75 year when the study was initiated in 1986. At baseline, these men completed detailed questionnaires assessing dietary intake, lifestyle factors and medical history. Follow-up questionnaires were mailed to participants every 2 years to update exposure information and ascertain newly diagnosed disease. Deaths were reported by family members or ascertained from the National Death Index. This study received institutional approval by the Human Subjects Committee at the Harvard School of Public Health.

We excluded participants from analyses if they reported an implausible daily energy intake (outside the range of 800–4,200 kcal/day) or omitted 70 or more of the 131 dietary questions asked at baseline (3% of men). In addition, we excluded those who had been diagnosed with any cancer other than nonmelanoma skin cancer (4%) or who reported OPL (0.04%) prior to the baseline date of January 1986. After exclusions, 42,340 men remained in the study.

Assessment of OPLs

A specific question on the lifetime occurrence of leukoplakia or any other OPL was first included in the 1996 survey, and any new diagnosis was reported on subsequent biennial questionnaires. Participants who reported a professional diagnosis of OPL from 1986 to 2002 were mailed an additional questionnaire to confirm the diagnosis and provide consent for records release. We collected dental records and pathology reports from treating dentists and oral surgeons. A study investigator blinded to the exposure status of participants reviewed records to confirm diagnoses.

Primary outcome definition was based on a clinical diagnosis, because dysplasia is not necessary to define OPL. Lesions meeting one of the diagnosis criteria described in Table AI were included. Oral malignancies that were not first diagnosed as premalignancies were included because they likely developed from premalignant precursors16, 17; omitting these cases would have resulted in selective exclusion of premalignancies, particularly those with greater malignant potential. Also, participants whose medical records indicated oral squamous cell carcinoma-in situ were considered cases because severe oral epithelial dysplasia is consistent with cell carcinoma-in situ. During follow-up, we verified a total of 207 new OPL events (Table AI). Nondentist participants who attested to their OPL on the additional questionnaire, but for whom dental/medical records were unobtainable, were considered “probable” OPL cases (22% of cases).

The main analyses were conducted including all OPL events (n= 207), and then repeated with various restrictions to verify overall findings. First, we conducted an analysis with only confirmed OPL cases (n = 162). Then, to further minimize potential misclassification and detection bias, we restricted the entire sample of participants to the 24,619 men who were dentists by profession (94 events). An analysis aiming to examine a longer induction period also served to ensure the prospective quality of the investigation; here, we included only OPL events that occurred after 1996, which was the first year the diagnosis appeared on a HPFS questionnaire (102 events). We also ran analyses excluding 19 cases diagnosed as oral lichen planus or 46 cases that were first diagnosed as oral cancer upon detection. Finally, because the clinical definition of OPL is prone to subjectivity and is waning in popularity, we analyzed the outcome of histologically confirmed oral epithelial dysplasia or cancer (91 events). Because results were similar across these various restricted analyses, the results we report are from analyses including all the probable OPL events, unless otherwise specified.

Assessment of nutrient intake

A semiquantitative food frequency questionnaire (FFQ) was included in 1986, 1990, 1994 and 1998 in the HPFS. The baseline FFQ contained a list of 131 food and beverage items, including 15 fruit, 30 vegetable and 3 potato items. Nine mutually exclusive response possibilities were provided for frequency of intake over the previous year, with choices ranging from “almost never or less than once per month” to “six or more times per day.” In addition, every 2 years we asked about use of vitamin supplements, including the type, dose, duration and specific brand.

This study examined intake of vitamin C, vitamin E, vitamin A (the sum of preformed vitamin A retinols and carotenoids according to their potential vitamin A activity) and the following carotenoids: α-carotene, β-carotene, β-cryptoxanthin, lycopene and lutein/zeaxanthin. Total intake for each nutrient was computed by multiplying the consumption frequency of each supplement or food by the nutrient content of the specified portion size, using composition values from US Department of Agriculture (USDA) sources and additional information from food manufacturers. When aggregating items to compute average intakes, we considered missing values for individual items within FFQs that were substantially complete to imply no intake (<1% of the data for most items).28 Our calculations for vitamin E intake, expressed as mg of α-tocopherol, take into account the lower activity of synthetic vitamin E often used in supplements. Our calculations for total vitamin A, expressed as retinol equivalents, take into account the poor absorption and bioavailability of carotenoids. Carotenoid counts were based on the USDA-National Cancer Institute database,29, 30 and carotenoid content of tomato-based foods were updated from the USDA.31 For multivitamins, a comprehensive multivitamin preparation database, which includes the dose of individual vitamins in each preparation, has been developed and updated biennially by the HPFS team. Participants who reported using supplements but did not report the dose or duration were assigned the median values for dose and duration observed among the users.

The reproducibility and validity of the FFQ has been demonstrated by comparison with dietary records and plasma samples in a random sample of participants.32, 33, 34 Compared with 2 one-week diet records completed 6 months apart, the Pearson correlation coefficients for the FFQ were 0.92 for vitamin C and vitamin E, and 0.64 for total carotene.32 The correlations between plasma concentrations and FFQ or diet record were 0.51 for α-tocopherol, 0.47 for α-carotene and lycopene, 0.35 for β-carotene, 0.43 for β-cryptoxanthin and 0.40 for lutein.33, 34 Considering that multiple factors contribute to between-person variations in plasma vitamin levels, the FFQ reasonably reflects the long-term nutrient intake of study participants.

Data analysis

Each eligible participant contributed person-time of follow-up from the date of return of the baseline questionnaire to the month of diagnosis of an oral premalignant or malignant lesion, death or the end of follow-up (January 31, 2002), whichever occurred first. Participants who reported an OPL or oral cancer or who died were excluded from subsequent follow-up.

Nutrient intake values were adjusted for total energy intake using the residual model method, in which the nutrient intakes of individuals are regressed on their total energy intakes, and then standardized to a total energy intake of 2,000 kcal/day.35 To best represent long-term nutrient intake and to reduce within-person variation, the main analysis used the cumulative average daily intake of nutrients from all available questionnaires up to the start of each 2-year follow-up interval.36 If participants were diagnosed with cancer during follow-up, intake was not updated after the beginning of the interval in which they developed the diagnosis. Because the induction period for any relationship between antioxidants and OPL is not known, in additional analyses, we examined the possibility of a long latency for OPL development, by using only the baseline (1986) intake, and then by using the baseline intake with a lag period of at least 10 years, excluding the 105 cases that occurred before 1996. Participants were grouped into quintiles of cumulative average daily intake of each nutrient, with the lowest quintile as the reference category. Linear relationships and tests for trend were assessed using the median values of deciles of intake to represent the exposure of all participants in the same decile to minimize the influence of outliers.36 For vitamins C, E, A and β-carotene, additional analyses separately considered supplemental and dietary intake.

We used multivariate Cox proportional hazards models with time-dependent covariates to calculate hazard ratio estimates and 95% CI as estimates of relative risk (RR). We examined the proportional hazards model assumptions by including interaction terms with time variables; nonsignificant interaction terms indicated no violation of proportional hazards. All nonnutrient covariates were updated in the analysis using simple updating for each 2-year period. Multivariate models adjusted for age (months), time period (2-year intervals), cigarette use (never, former, current), age at start smoking (<15, 15–19, 20–29, ≥30 years), quantity smoked during years of active smoking (1–4, 5–14, 15–24, 25–34, 35–44 or ≥45 cigarettes/day), time since quitting among past smokers (<10 or ≥10 years), pipe or cigar use (never, former, current), ever use of chewing tobacco (yes/no), alcohol consumption (g/d), total energy intake (quintiles), profession (dentist vs. nondentist) and total intakes of vitamin E, vitamin C and β-carotene (quintiles for each). In the analysis of nutrient intake from dietary sources, we adjusted for supplemental intake (categorical dose) of that nutrient, and in the analysis of supplemental intake, we adjusted for dietary intake (quintiles) of that nutrient. We also included the following factors and assessed change in estimates, but they did not have an impact on the results and so were not included as confounders in the final multivariate analyses: race/ethnicity, family history of cancer, having had a recent physical exam, number of natural teeth, intake of other micronutrients or flavonoids and an interaction term between tobacco use and alcohol consumption.37 Report of a recent routine physical exam was used as a surrogate for frequency of oral examinations, because data on oral examination frequency was not collected for the entire study cohort. A subanalysis excluded men taking any vitamin supplements. We considered differences in the effect of nutrient intake by categories of tobacco use in stratified analyses, using tertiles of nutrient intake due to the smaller number of participants and cases in each stratum, and also by including interaction terms between linear intake and smoking status in the multivariate models. All statistical tests were two-sided, performed at an alpha level of 0.05 and conducted using SAS software, version 9.1 (SAS Institute, Cary, NC).

Results

The distribution of risk factors for OPL differed somewhat by vitamin E, vitamin C and β-carotene intake (Table I). Men in the top quintiles of intake—most of whom used multivitamin or specific vitamin supplements—were more likely to be dentists by profession and to have had a recent routine physical exam. Men in the lower quintiles of nutrient intake were younger, consumed more alcohol per day and were more likely to use tobacco. The average percentages of current supplement users were 19% for vitamin E, 36% for vitamin C, 8% for vitamin A and 42% for multivitamins. As expected, intake levels of all dietary nutrients were significantly correlated (p < 0.0001), with particularly strong correlations between dietary vitamin C and β-cryptoxanthin (Spearman correlation coefficient, r = 0.70), dietary β-carotene and lutein/zeaxanthin (r = 0.69) and dietary β-carotene and α-carotene (r = 0.79).

Table I. Number of Oral Premalignant Lesion Cases from 1986 to 2002 and Age-Standardized Characteristics at Baseline, by Quintiles of Total Intake of Vitamin E, Vitamin C and β-Carotene, among 42,340 Men in the Health Professionals Follow-up Study1

Variable	Vitamin E intake		Vitamin C intake		β-Carotene intake
	Quintile 1	Quintile 5	Quintile 1	Quintile 5	Quintile 1	Quintile 5
1 Quintile 1 is the quintile with lowest intake; quintile 5 is the quintile with highest intake. Total n of events = 207. 2 Recent routine physical exam was asked in 1988 questionnaire. 3 Quantity smoked was calculated among current smokers only.
Oral premalignant lesion cases, n (207 total)	37	50	55	50	38	24
Mean daily intake
Vitamin E (mg of α-tocopherol)	7	202	14	141	38	74
Vitamin C (mg)	197	966	92	1,240	314	603
β-carotene (μg)	3,536	6,164	3,469	6,217	1,941	10,486
β-cryptoxanthin (μg)	64	89	34	99	56	109
α-carotene (μg)	715	1,058	694	1,071	352	2,082
Lutein and Zeaxanthin (μg)	2,675	4,259	2,649	4,407	1,717	6,508
Lycopene (μg)	7,945	10,700	7,856	11,165	6,805	13,649
Total Vitamin A (μg in retinol equivalents)	1,434	4,726	1,456	4,438	1,801	4,139
Folate (μg)	306	704	291	708	397	601
Alcohol (g)	13.3	11.3	13.2	10.9	13.7	9.0
Use of supplements (%)
Multivitamin	9	83	10	82	38	49
Vitamin E	0	93	2	57	14	28
Vitamin C	15	83	0	96	28	46
Risk factors
Age (mean y)	52.1	55.9	52.0	54.6	51.7	55.8
Dentist by profession (%)	53	60	51	64	50	65
Recent routine physical exam (%)2	49	54	48	54	49	55
Chewing tobacco ever-user (%)	5	4	6	3	6	3
Pipe or cigar user (%)	7	6	7	6	8	6
Cigarette smoking status (%)
Never	44	46	40	46	43	48
Past	41	43	43	43	42	42
Current	11	7	13	7	12	6
≥ 35 cigarettes/d3	18	16	18	15	19	15

In the multivariate analyses for linear trends, total intake (from both dietary and supplemental sources) of vitamin E, vitamin C, vitamin A or carotenoids was not significantly associated with risk of OPL (Table II). However, β-cryptoxanthin showed trends toward decreased risk and had inverse associations across all intake categories. Inverse associations were also apparent for most categories of α-carotene. In contrast, the trend test for lutein/zeaxanthin showed a significant increased risk with greater intake of these carotenoids, though there were no significant associations across quintiles. Vitamin E was associated with increased risk across all quintiles. The linear trend of increased risk with total vitamin E intake was significant (p_trend = 0.03) before we controlled for vitamin C and β-carotene, and even after adjusting for these other nutrients, there was a significantly increased risk for men in the highest vitamin E intake category compared with the lowest. No clear relationship emerged with β-carotene.

Table II. Multivariate Relative Risks (RR) and 95% CI of Oral Premalignant Lesions, by Cumulative Average Intake of Vitamin E, Vitamin C, Vitamin A and Carotenoids, Among 42,340 Men and 207 Events in the Health Professionals Follow-up Study1

Variable	RR (95% CI) per quintile of intake				Test for trend p	Median daily intake
	2	3	4	5		Quintile 1	Quintile 5
1 Total intake includes both dietary and supplemental sources. Risks are calculated for quintile categories (quintile 1 is the referent). Test for trend are calculated from linear models that use median values of deciles of intake to represent the intake of all participants in the same decile. All risk estimates are adjusted for age, follow-up cycle, pipe/cigar smoking status, chewing tobacco use, cigarette smoking status, years since quitting among past smokers, cigarettes smoked/day among current smokers, age at start smoking, alcohol use, total energy intake, dental profession and intakes of vitamin E, vitamin C and β-carotene. 2 p ≤ 0.05. 3 RR additionally adjusted for supplemental intake of that nutrient in dose categories. 4 p ≤ 0.01.
Total vitamin E	1.36 (0.84–2.20)	1.36 (0.82–2.26)	1.57 (0.94–2.62)	1.84 (1.06–3.19)2	0.17	8.1 mg	189 mg
Dietary vitamin E3	1.21 (0.77–1.89)	1.42 (0.90–2.23)	1.41 (0.87–2.26)	1.48 (0.91–2.42)	0.09	7.6 mg	13.3 mg
Total vitamin C	0.83 (0.52–1.32)	0.99 (0.62–1.59)	0.83 (0.50–1.37)	0.96 (0.56–1.64)	0.45	106 mg	1,056 mg
Dietary vitamin C3	0.64 (0.43–0.96)2	0.58 (0.38–0.89)4	0.64 (0.41–1.00)2	0.52 (0.31–0.85)4	0.04	86 mg	252 mg
Total vitamin A	0.87 (0.55–1.37)	0.77 (0.47–1.26)	0.83 (0.50–1.39)	0.77 (0.43–1.38)	0.81	1,032 μg	4,603 μg
Dietary vitamin A3	1.07 (0.70–1.63)	0.87 (0.54–1.39)	0.80 (0.48–1.33)	0.94 (0.55–1.61)	0.80	902 μg	2,670 μg
Total β-carotene	1.48 (0.95–2.31)	1.61 (1.02–2.53)2	1.41 (0.87–2.27)	1.17 (0.70–1.95)	0.58	2,240 μg	9,195 μg
Dietary β-carotene3	1.32 (0.86–2.03)	1.47 (0.95–2.27)	1.21 (0.75–1.93)	1.16 (0.70–1.90)	0.59	2,170 μg	8,105 μg
Total β-cryptoxanthin	0.86 (0.58–1.26)	0.56 (0.35–0.88)4	0.67 (0.43–1.05)	0.64 (0.40–1.00)	0.18	19.7 μg	202 μg
Total α-carotene	0.69 (0.44–1.08)	0.97 (0.63–1.50)	0.56 (0.33–0.94)2	0.80 (0.45–1.41)	0.61	327μg	1,792 μg
Total lutein/zeaxanthin	0.85 (0.53–1.36)	0.99 (0.61–1.59)	0.98 (0.59–1.62)	1.44 (0.85–2.44)	0.02	1,411 μg	6,407 μg
Total lycopene	0.77 (0.50–1.19)	0.89 (0.58–1.36)	0.96 (0.62–1.48)	0.83 (0.52–1.32)	0.97	3,912 μg	16,903 μg

When we examined nutrient intake from only dietary sources of vitamins E, C, A or β-carotene (excluding intake from multivitamins or specific supplements), dietary vitamin C was significantly associated with reduced risk of OPL (Table II). Men with the highest intake had approximately half the risk of men with the lowest intake, with a significant linear trend. Considering the strong correlation between dietary vitamin C and β-cryptoxanthin, we mutually adjusted for each nutrient in a subanalysis; dietary vitamin C retained significance to a greater extent than did β-cryptoxanthin, suggesting that the inverse association with β-cryptoxanthin may partly be explained by its correlation with vitamin C. Dietary β-carotene and vitamin E were associated with increased OPL risk, but risk estimates were not statistically significant. In the subanalysis excluding participants taking any multivitamin or individual vitamin supplements (leaving 61 events among 17,773 men), there was a linear positive association with dietary β-carotene (data not shown). Results for other dietary nutrients were similar when supplement users were excluded.

We saw no significant trends in risk with supplemental use of vitamin C or β-carotene (Table III). Vitamin E supplements seemed to increase OPL risk, not only with higher doses, but also longer duration, as men who had used vitamin E supplements for 10 years or more had a RR of 1.53 (95% CI, 0.93–2.52, p = 0.10) compared with never users, and the linear trend in duration was near significance (p_trend = 0.06). Interestingly, we saw a significant risk reduction for vitamin A supplement use in lower doses, but suggestions for increased risk with higher doses. The overall test for trend was nonsignificant, but supported the notion that risk increased as vitamin A dose increased.

Table III. Multivariate Relative Risks and 95% CI of Oral Premalignant Lesions, by Dose of Vitamin Supplements1

Variable		Dose of supplement			Test for trend p
1 All risk estimates are adjusted for age, follow-up cycle, pipe/cigar smoking status, chewing tobacco use, cigarette smoking status, years since quitting among past smokers, cigarettes/d among current smokers, age at start smoking, alcohol use, total energy intake, dental profession, dietary intake of the nutrient and total intake of other nutrients: vitamin E, vitamin C and/or β-carotene. Tests for trend are calculated from linear models that use median values of deciles of intake to represent the intake of all participants in the same decile. 2 p = 0.05.
Vitamin E	Never users	≤250 IU/d	300–500 IU/d	≥600 IU/d
	1.0	1.11 (0.60–2.06)	1.12 (0.80–1.58)	1.47 (0.84–2.57)	0.25
Vitamin C	Never users	<400 mg/d	400–750 mg/d	≥750 mg/d
	1.0	0.77 (0.31–1.94)	1.24 (0.89–1.74)	1.31 (0.82–2.07)	0.15
Vitamin A	Never users	<8,000 IU/d	≥ 8,000 IU/d
	1.0	0.14 (0.02–1.00)2	1.29 (0.91–1.82)		0.16
β-carotene	Never users	<8,000 IU/d	≥8,000 IU/d
	1.0	0.43 (0.11–1.77)	0.96 (0.63–1.44)		0.82

Because supplemental vitamin C had no association with OPL risk, while dietary vitamin C had significant inverse associations, we examined the potential that other components of dietary sources of vitamin C may be involved in lowering OPL risk. Inverse associations between consumption of vitamin C-rich fruits and vegetables and OPL risk38 were virtually unaffected by adjustment for dietary vitamin C or β-cryptoxanthin.

Overall, results from the analyses using baseline intake with or without a 10-year lag period before diagnosis were similar to the main analyses (data not shown). When we restricted the analyses to participants who were dentists (94 events among 24,465 men), the reduction in risk was somewhat greater for dietary vitamin C (quintile 5 vs. 1, RR = 0.47, 95% CI 0.23–0.98, p = 0.05), while the increased risk with vitamin E intake became highly significant across all quintiles (e.g., quintile 5 vs. 1, RR = 3.65, 95% CI 1.46–9.15, p = 0.006), with a significant trend for dietary vitamin E (p_trend = 0.04). Results were further confirmed in the analysis of cases with oral epithelial dysplasia (91 events among 42,274 men), but estimates were not as precise (data not shown).

Modification of effects by tobacco use

As tobacco has great potential to both distort results through confounding and also modify effects of nutrients through biological interactions, we performed analyses separately for never users, past users and current users of tobacco (cigarettes, cigars, pipes or chewing tobacco) (Table IV). Among never users, no significant trends emerged, but there were inverse associations for vitamin C and β-carotene intake. Among past tobacco users, an inverse association indicating approximately 50% lower OPL risk with high dietary vitamin C was significant (p = 0.02). However, past smokers with high intake of vitamin C from supplements (≥750 mg/day) had twice the risk of OPL, compared with past smokers who never used vitamin C supplements (p = 0.05).

Table IV. Multivariate Relative Risks (RR) and 95% CI of Oral Premalignant Lesions for High Intake of Vitamin E, Vitamin C, Vitamin A and Carotenoids, by Tobacco use (Chewing Tobacco, Cigarette, Pipe, Cigar) Status1

Variable	Tobacco use status, RR (95% CI)			Interaction p-value4
	Never, 46 cases	Past2, 92 cases	Current3, 69 cases
1 All risk estimates are adjusted for age, alcohol use, dental profession and total intakes of energy, vitamin E, vitamin C and β-carotene. Risk estimates for total and dietary intake compare tertile 3 (highest intake) to tertile 1 (referent, lowest intake). Risks for dietary intake are additionally adjusted for supplemental intake of that nutrient in dose categories. Risk estimates for supplemental intake compare the highest dose category to nonusers of the supplement, and all estimates are additionally adjusted for dietary intake of the nutrient. β-cryptoxanthin, α-carotene, lutein and lycopene use total intake values. 2 Past use of cigarette, cigar, pipe or chewing tobacco. Risks are additionally adjusted for age at start of smoking and years since quitting cigarettes. 3 Current use of cigarettes, pipe or cigar. Risks are additionally adjusted for age at start of smoking and cigarettes/d. 4 Overall interaction effect between linear trend in nutrient intake and smoking (0 = never, 1 = past, 2 = current), corresponding to a chi-square test with 1 df based on the likelihood ratio of models with and without an interaction term. 5, 5, p ≤ 0.05. 6 Test for linear trend p ≤ 0.05. For total or dietary intake, linear models used median values of deciles of intake to represent the intake of all participants in the same decile. For supplemental intake, linear models used the median values of dose categories. 7 Test for linear trend 0.05 ≤ p < 0.10, calculated as in footnote 6. 8 p ≤ 0.01.
Vitamin E
Total	1.27 (0.54–3.02)	0.99 (0.52–1.87)	1.78 (0.87–3.64)7	0.4
Dietary	1.15 (0.52–2.57)	1.13 (0.64–1.99)	1.92 (0.99–3.73)5	0.04
Supplement	0.73 (0.16–3.37)	1.09 (0.45–2.61)	3.07 (1.28–7.34)8	0.08
Vitamin C
Total	0.66 (0.26–1.66)	1.26 (0.68–2.34)	1.57 (0.74–3.34)	0.2
Dietary	0.80 (0.37–1.77)	0.51 (0.29–0.91)5,6	0.83 (0.42–1.64)	0.6
Supplement	0.96 (0.37–2.47)	2.02 (1.01–4.06)5,6	1.18 (0.55–2.55)	0.4
Vitamin A
Total	2.04 (0.74–5.63)	0.70 (0.35–1.37)	0.92 (0.42–1.99)	0.2
Dietary	0.72 (0.29–1.81)	0.78 (0.42–1.44)	1.18 (0.59–2.37)	0.6
Supplement	0.68 (0.30–1.54)	1.13 (0.68–1.88)	1.73 (0.96–3.13)6	0.05
β-carotene
Total	0.88 (0.41–1.87)	0.85 (0.49–1.47)7	1.59 (0.83–3.05)	0.3
Dietary	0.93 (0.43–1.99)	0.70 (0.40–1.21)	1.94 (1.01–3.74)5	0.5
Supplement	0.51 (0.20–1.30)	0.93 (0.51–1.68)	1.44 (0.69–2.99)	0.10
β-cryptoxanthin	1.62 (0.71–3.66)	0.61 (0.35–1.03)	0.47 (0.23–0.95)5	0.6
α-carotene	1.50 (0.65–3.15)	0.72 (0.36–1.44)	0.79 (0.37–1.67)	0.9
Lutein/zeaxanthin	1.65 (0.68–4.00)	1.44 (0.78–2.68)	1.32 (0.65–2.69)	0.4
Lycopene	1.84 (0.83–4.05)	0.89 (0.51–1.54)	0.73 (0.40–1.33)	0.5

Among current users of tobacco, the only nutrient that was significantly associated with decreased OPL risk was β-cryptoxanthin (p = 0.03), slightly more so than in its association among past tobacco-users. On the other hand, current smokers had an elevated risk of OPL with greater intakes of β-carotene, vitamin E or supplemental vitamin A. Smokers in the top tertile of dietary β-carotene or dietary vitamin E had approximately twice the risk of OPL as smokers in the lowest tertile. Smokers taking ≥600 IU/day of vitamin E had 3 times the risk of smokers who were nonusers of vitamin E supplements. Fewer men took supplements of β-carotene, thereby limiting our analytical power, but the linear trend for dose of vitamin A supplements (which often contain β-carotene) was significant.

Formal tests for interactions between tobacco use and nutrients were statistically significant for vitamin A supplements and dietary vitamin E (Table IV). An exploratory analysis of interactions between nutrients showed a significant interaction leading to an even greater risk of OPL with higher doses of both vitamin E and vitamin A supplements (p = 0.01), or both vitamin E and β-carotene supplements (p = 0.03), only among current smokers. No such interactions between nutrients were apparent in nonsmokers.

Discussion

In this prospective study, high consumption of vitamin C from dietary sources, but not supplements, was associated with a reduction in risk for OPLs in men. Inverse associations were also apparent for β-cryptoxanthin and α-carotene, though trends were nonsignificant. Conversely, greater intakes of vitamin E and β-carotene were associated with increased risk of OPL, particularly among current users of tobacco, where exploratory analyses showed significant interactions leading to increased risk for smokers taking both vitamin E and vitamin A supplements. Total vitamin A intake seemed unrelated to OPL risk, possibly as a consequence of the opposing directions of associations between various carotenoids and OPL, and the null association between preformed vitamin A and OPL. Lycopene was also unrelated to OPL risk.

Our findings of inverse associations with dietary vitamin C are consistent with results from previous case–control studies of OPL and oral cancer.7, 8, 10, 23, 25 Although vitamin C has well-established antioxidant properties,4 it is possible that the protection seemingly offered by its dietary intake is actually partly due to some other component of vitamin C-rich foods. Earlier case–control studies of diet and oral cancer risk suggested that other dietary components may be at work, because inverse associations observed with fruit intake were not completely accounted for by vitamin C, carotene or fiber in fruits (other micronutrients were not associated with oral cancer in the study).6 Similarly, in our study, risk reductions with vitamin C-rich fruits and vegetables were not explained by vitamin C or β-cryptoxanthin intake.38 Flavonoids, which are plant pigments found in vitamin C-rich fruits and vegetables, may be involved,39 but we examined intake of flavonoids and saw no association between flavonoids and OPL risk in our study.

Our results regarding supplemental use of vitamin C support the notion that vitamin C alone may not be the primary protective agent. Rather, vitamin C from supplements was associated with an increased risk of OPL, especially in past tobacco users, although this finding may be an artifact of residual confounding by other supplements or the relatively low numbers of cases in past smokers. It is less likely that differences in the bioavailability of ascorbic acid used in supplements explain these results, because natural and synthetic ascorbic acid are chemically identical, and studies have not found any differences in bioavailability.4, 40, 41, 42, 43, 44 The median intake for participants in the lowest quintile of vitamin C from foods was close to the RDA (90 mg/day). Thus, we cannot exclude the possibility that use of vitamin C supplements may reduce risk of OPL for people deficient in vitamin C.

We observed an increased risk of OPL with greater intake of vitamin E and β-carotene, particularly among current smokers. While conflicting with many earlier case–control studies of OPL and oral cancer,7, 8, 10, 26 our results are consistent with recent clinical trials. Two major placebo-controlled trials of β-carotene in populations predominantly composed of cigarettes smokers—the β-Carotene and Retinol Efficacy Trial and the α-Tocopherol, β-Carotene Lung Cancer Prevention Study—observed an unexpected increased risk of lung cancer in the intervention arms during the supplementation period.45, 46, 47 In a subgroup of participants in the α-Tocopherol, β-Carotene study, Liede et al. examined the risk of leukoplakia and failed to find any associations with use of either α-tocopherol (50 mg/day) or β-carotene (20 mg/day), despite a 7-fold greater concentration of β-carotene in oral mucosal cells in the intervention arm.48, 49 More recently, 2 randomized, placebo-controlled clinical trials designed specifically for oral cancer, using populations composed of past and never smokers as well as current smokers, confirmed the complexity of chemopreventive strategies: supplementation with vitamin A (150,000–300,000 IU/day retinyl palmitate) had no benefit in patients with oral cancer,13 but supplemental α-tocopherol significantly increased risks of head and neck cancer recurrence and second primary tumors, indicating 2- to 3-fold greater risks with vitamin E use (400 IU/day) during the supplementation period (2nd primary cancer RR = 2.9, 95% CI 1.6–5.3).12

Although the mechanisms by which α-tocopherol may act as a prooxidant are unclear,50 evidence of the harmful properties of β-carotene when given in isolation to individuals exposed to environmental carcinogens is convincing.51 β-carotene may act as a cocarcinogen through different mechanisms, which include decreasing tissue concentrations of retinoic acid and retinoic acid receptor-β while increasing activator-protein-1, or increasing the levels of reactive oxygen species.51, 52, 53, 54, 55 A recent study demonstrated that β-carotene exacerbates DNA oxidative damage and modifies p53-related pathways of cell proliferation and apoptosis in cultured cells exposed to tobacco smoke, but has no significant effect in cells unexposed to tobacco.56 Overall, it seems that β-carotene behaves as an antioxidant at low oxygen partial pressures, but as a pro-oxidant at higher oxygen pressures.51 Analyses aiming to clarify the association between β-carotene, α-tocopherol and risk of OPL therefore need to carefully consider individuals exposed to carcinogens such as tobacco smoke.

The prospective nature of the data in this study coupled with the asymptomatic nature of oral premalignancies makes it unlikely that nutrient intakes were biased by disease status. Our ability to use repeated measures through follow-up further minimizes any reporting errors. Nevertheless, errors in the measurement of long-term diet are to be anticipated and create difficulties in assessing the independent effects of nutrients.57 For example, we observed an increased risk of OPL with greater intake of lutein/zeaxanthin, but this may be due to the fact that the primary dietary source of lutein, spinach, is also abundant in β-carotene and vitamin E. Risks with lutein/zeaxanthin were increased only for men in the highest quintile of intake; thus, it may be that the observed linear trends are due to residual confounding in that category of intake, by high intake of vitamin E or β-carotene.

A disadvantage of the asymptomatic nature of OPL is that some cases may have been missed. This may cause bias if the missed cases are related to frequency of oral examinations and frequency of oral exams is related to nutrient intake. Although we had no data on frequency of oral examinations in the entire cohort, in a substudy of 584 HPFS participants (described by Pai et al.58), 80% of participants reported receiving a dental exam at least once per year (unpublished data). To address concern over detection bias, we restricted the main analysis to the 58% of participants who were dentists by profession. Among dentists, the inverse association between nutrient intake and frequency of oral exams was weaker than in the total substudy population, as dentists had a more cohesive frequency of oral exams. Furthermore, dentists were significantly more likely than nondentists to have more than 2 oral exams per year, whereas nondentists were significantly more likely to receive oral exams only when in pain. As expected, results in the dentist-only analysis suggest that the association between nutrients and OPL may be stronger than our main results indicated, particularly for the significant inverse associations with vitamin C and the increased risks with vitamin E. We further confirmed results in the analysis of oral epithelial dysplasia. However, residual detection bias from variation in frequency of exams is still possible.

In conclusion, with control for detailed tobacco and alcohol use characteristics, risk of OPLs was significantly reduced with higher dietary intake of vitamin C for all participants, while an increased risk was observed with greater consumption of dietary or supplemental vitamin E and β-carotene, particularly among current users of tobacco. Although α-carotene or β-cryptoxanthin may reduce the risk of OPL, our data do not provide adequate support for recommending an increased intake of these compounds, and with the relatively small number of participants and cases in each category of tobacco use, findings from the stratified analysis should be viewed with caution. The benefits of a diet that includes vitamin C-rich fruits as well as a variety of nutrients should continue to be emphasized, but the data presented here underscore the need to consider possible detrimental effects of certain isolated supplements, particularly in individuals exposed to environmental carcinogens.

Acknowledgements

The authors thank Drs. Walter Willett, Eric Rimm and Athanasios Zavras for their support and advice; Al Wing, Mira Kaufman and Lydia Liu for statistical programming assistance; Stacey DeCaro, Elizabeth Frost-Hayes, Katherine Smith and Melanie Gerard for administrative support and the participants of the Health Professionals Follow-up Study for their continued cooperation.

Appendix

Table AI. Eligible Oral Premalignant Lesions for Case Definition1

Lesion	Definition	Percent (n)
1 Lesions on the external lip vermilion, which are likely to resemble skin cancer in etiology, and lesions of the oropharynx were ineligible. There were no cases of proliferative verrucous leukoplakia or oral submucous fibrosis. 2 Percent is calculated among the 102 cases who had a clinical diagnosis term assigned. 3 Percent is calculated among total number of cases (n = 207). A biopsy was performed on 142 cases; thus, 64% of biopsied lesions displayed oral epithelial dysplasia.
Clinical diagnosis terms
Leukoplakia	A white patch or plaque that does not rub off, and cannot be characterized clinically or pathologically as any other disease	73 (74)2
Erythroplakia	A red patch that cannot be clinically or pathologically diagnosed as any other condition	3 (3)2
Erythroleukoplakia	An area of leukoplakia that has red patches, also known as “speckled leukoplakia” or“leukoerythroplakia”	2 (2)2
lichen planus	Chronic dermatologic disease that also affects the mucosa of the mouth, of either reticular (interlacing white lines) or erosive (ulceration in center) type	19 (19)2
Histopathological Diagnosis Terms
Oral epithelial dysplasia	Histopathologically verified abnormality of development, in pathology, alteration in size, shape and organization of adult cells	44 (91)3
Oral squamouscell carcinoma	Histopathologically verified malignant neoplasm characterized by the proliferation of anaplastic cells that tend to invade surrounding tissue and metastasize; virtually all arise from premalignant precursor lesions	22 (46)3