Carotenoids, Vitamins C and E, and Mortality in an Elderly Population

Transcription

1 American Journal of Epidemiology Copyright O 1996 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Vol. 144, No. 5 Printed in US.A Carotenoids, Vitamins C and E, and Mortality in an Elderly Population Nadine R. Sahyoun, Paul F. Jacques, and Robert M. Russell In , the nutritional status of 747 noninstitutionalized Massachusetts residents aged 60 years and over was assessed. Nine to 12 years later, the vital status of these subjects was determined. The data of a subset of 725 community-dwelling volunteers was used to examine associations between mortality and the nutrient antioxidants (carotenoids and vitamins C and E) in plasma, diet, and supplements. Results indicated that subjects with plasma vitamin C levels in the middle and high quintiles had a lower overall mortality (relative risk () =, 95% confidence interval (Cl) and = 0.54, , respectively) than those in the lowest quintile even after adjustment for potential confounders. These associations were largely due to reduced mortality from heart disease. Subjects in the highest quintile of total intake of vitamin C also had a significantly lower risk of overall mortality ( = 0.55, ) and mortality from heart disease ( = 0.38, ) than did those in the lowest quintile after potential confounders were controlled for. Intake of vegetables was inversely associated with overall mortality (p for trend = 0.003) and mortality from heart disease (p for trend = 0.04). No other significant associations were observed. In conclusion, the results indicate that high intakes and plasma levels of vitamin C and frequent consumption of vegetables may be protective against early mortality and mortality from heart disease. Am J Epidemiol 1996; 144: aged; ascorbic acid; biochemistry; carotenoids; diet; food; mortality; vitamin E Cardiovascular disease and cancer are the two leading causes of death among the adult population in the United States (1). It is believed that oxidant damage to DNA, protein, and other macromolecules is responsible to a large extent for these diseases and other age-related degenerative conditions such as cataracts and cognitive dysfunction (2, 3). Maintenance of tissue integrity is dependent on the balance between the production and neutralization of toxic oxidants. The possible protective effects of vitamins E and C and carotenoids against the onset of chronic diseases are currently under investigation. It is believed that these nutrients function as antioxidants and act as scavengers of free radicals, either independently or as part of larger enzyme systems. Vitamin E is concentrated in the lipid regions that are exposed to the highest partial pressure of oxygen, such as cells lining the outer surface of the lung and red blood cell membranes (4). Beta-carotene is located in the membranes and organelles that are exposed to the lowest Received for publication August 23, 1995, and in final form May 16, Abbreviations: Cl, confidence interval;, relative risk. From the Jean Mayer USDA Human Nutrition Research Center on Aging at Tufts University, Boston, MA. Reprint requests to Dr. Nadine R. Sahyoun, National Center for Health Statistics, Office of Analysis, Epidemiology and Health Promotion, 6525 Belcrest Road, Room 775, Hyattsville, MD partial pressure of oxygen (4), but its action may not be restricted to such locations, as seen by its possible protection against lung cancer (5). Vitamin C is located in the water-soluble component of the body. The fact that many of these reactions are linked in cascades or interrelated processes indicates that any single antioxidant may be less effective alone than in combination with complimentary agents (6). Vitamin C is believed to be the first line of defense (7, 8) and appears to have a role in sparing or reconstituting the active forms of vitamin E. Vitamin C was also shown, in vitro, to spare carotenoids (8). However, few studies have examined the relation between all three nutrient antioxidants and morbidity or mortality. In fact, most studies have examined single antioxidants and those from single sources of measures such as blood, dietary intake, or food groups. This study was undertaken to examine associations between mortality and all three antioxidants (carotenoids and vitamins C and E) in plasma, diet, supplements, and food groups over a 12-year follow-up period in an elderly population. MATERIALS AND METHODS The Nutrition Status Survey was conducted in Boston, Massachusetts, from 1981 to Complete nutrition assessment information was collected for

2 502 Sahyoun et al. noninstitutionalized subjects aged 60 years and over. Subjects were self-selected and recruited through presentation of the study to secular and religious groups and through contacts with community housing projects and neighborhood health clinics. The subjects were free of terminal, wasting disease and severe metabolic disorders. The medical exclusions to participating in the study included uncontrolled diabetes, chronic renal or liver failure, active alcoholism, uncontrolled malabsorption syndrome, and untreated diseases such as hyper- or hypothyroidism, malignant neoplasm, severe chronic inflammatory or infectious disease, psychiatric illness, and dementia. Subjects Of the 747 community-dwelling subjects who participated in the study, 67 were either unable to keep a food record or provided records that were unclear or incomplete. Therefore, complete intakes of nutrients, supplements, and food groups were available for 680 subjects. Analysis of carotenoids was initiated a few months after the start of the study, and so complete biochemical data were available on 515 subjects. The characteristics of subjects with and those without biochemical data were very similar and therefore should not affect the results of the study. A total of 725 subjects had biochemical data, data on dietary and supplemental intakes, or both. The subjects included both sexes and ranged in age from 60 to 101 years. The subjects in our study were healthier and better educated than the general population. This was reflected in their serum levels and in their dietary intake, which were, on average, more adequate than those of the general population. However, there was enough variability in nutrient values to examine high versus low values with the long-term outcome. Since the intent of our study was to examine the protective effect of antioxidants rather than the outcome of deficiency states, we believe that this population provided an opportunity to examine the impact of high levels of the nutrients. The results may have been more marked had there been more subjects in the lower end of the distribution. Data collection The following information was obtained on all volunteers. Biochemical profile. An overnight fasting blood sample was collected and analyzed for complete blood count, nutrient biochemical parameters, and routine clinical analysis (sequential multiple analyzer, SMA- 20, Technicon Corporation, Tarrytown, New York). All blood samples were transported immediately to the laboratories, where they were processed and frozen. The samples were then analyzed in batches throughout the study. Vitamin E was measured by high-performance liquid chromatography (9), vitamin C was determined by a colorimetric method (10), and total plasma carotenoids were measured by a standard spectrophotometric method on a petroleum ether extract (11). Dietary intake. The subjects kept a 3-day food record. They were instructed by a dietitian in the proper manner of recording everything they ate or drank for 3 consecutive days, including health food supplements. The food records were then collected and reviewed by the dietitian for accuracy and completeness. The Grand Forks Nutrient Database (12) was used to analyze the dietary data for its nutrient content, and the United States Department of Agriculture/National Cancer Institute's carotenoid food composition database (13) was used to analyze the dietary data for its carotenoid content. For our analyses, carotenoid intake was derived from the sum of five carotenes: alphacarotene, beta-carotene, lutein plus zeaxanthin, lycopene, and beta-cryptoxanthin. Dietary content of vitamin E is not complete in the nutrient database. However, the vitamin E in food is so much lower than that in supplements that the dietary content would not substantially influence the outcome of the analysis. Foods were subdivided into related groups and subgroups by modifying the Grand Forks groupings as described by Tucker and Rush (14). The food subgroups citrus fruit/juices and dark green/orange vegetables were considered the largest contributors of vitamin C and carotenoids to the diet, respectively. The consumption of these food subgroups and of the food groups vegetables and fruit was examined for their association with mortality. Anthropometric and physical evaluation. Heights and weights were measured, and body mass index was calculated as weight (kg)/height (m) 2. In addition, a brief physical examination was conducted to judge overt general health status. This included measurement of blood pressure while seated, pulse rate, and a urine sample to measure sugar and protein. Medical and medication history. A nurse-practitioner obtained information on the subjects' medical history by prompting the volunteers for specific medical conditions and major surgical procedures. These medical conditions included stroke, heart attack, heart failure, and claudication. The presence of diabetes and hypertension was determined on the basis of intake of medication or of dietary modifications for these conditions, as reported by volunteers. The subjects were categorized according to their disease status as those

3 Carotenoids, Vitamins C and E, and Mortality 503 TABLE 1. Key characteristics of the Nutrition Status Survey noninatitutionalized population, Massachusetts, Population characteristics Men (n = 254) Women Vital status (% alive) Age (mean years (SD*)) 72.3 (7.4) 73.7 (8.1) Marital status (%) Married Widowed Single Divorced/separated Education (%) Grade school or less school College/graduate school Cigarette use (%) Never Ever Chewing problems (% yes) CXietelet body mass index (mean kg/m> (SD)) Disabilities affecting shopping (% yes) History of disease (%) Heart failure Heart attack Claudication Stroke Diabetes Hypertension Disease status (%) None 1 22 Medication use (%) None Serum cholesterol (mean mg/dl (SD)) Plasma nutrients (mean (SD)) Plasma vitamin C (mg/dl) Plasma vitamin E (mg/dl) Total plasma carotenoids (ng/dl) Total nutrient intakes (mean (SD)) Vitamin C intake (mg) Vitamin E Intake (mg) Dietary carotenoids (jig) ,912 (4.1) (47.5) (0.41) (0.55) (49.5) (592.4) (158.5) (5,336) ,173 (5.1) (53.2) (0.41) (0.56) (52.7) (515.5) (156.1) (5,238) Supplement use (%) MultJvitamin (with vitamins E and C) Vitamin C Vitamin E Food groups (mean (SD)) All vegetables (g) Dark green/orange vegetables (g) All fruit (g) Citrus fruit/juices (g) (119.6) (42.9) (199.6) (120.6) (116.2) (49.5) (169.2) (107.7) 1 SD, standard deviation.

4 504 Sahyoun et al. with no history of disease, those who were diagnosed with one chronic disease, and those who suffered from two or more of the chronic diseases listed in table 1. The subjects were also questioned about whether they had received medical or dental care or had been hospitalized during the year previous to their participation in the survey. Past and present medication and nutrient supplement use were also obtained. History of supplement use was collected by the nurse-practitioner at the interview or by the dietitian from the food record if it was a health food supplement. Information was obtained on past and present supplement use, dosage, duration, and frequency. Demographic information and health practices. Subjects were administered a questionnaire on demographic characteristics, health practices, and eating habits. Information collected included sex, race, marital status, education, employment and current/last job, living arrangement, eating habits, group meals, cooking facilities, dental care and denture wear, eating problems, medical conditions affecting either appetite or ability to eat, and medical conditions affecting the ability to cook or shop. Also collected was information on alcohol consumption and tobacco use. Vital status. The vital status of subjects as of January 1993 was ascertained. This represents 9-12 years of follow-up. The mortality of the study subjects was determined by examining the annual index of the deaths that occurred in the Commonwealth of Massachusetts from 1981 to January Death certificates were obtained to extract information used to confirm the match between the deceased and the corresponding study subjects. The information used to confirm the match included last name, first name and middle initial, date of birth, address of residence, country of birth, occupation, marital status, and race. The National Death Index, a central, computerized index of the death certificates filed in the vital statistics office of each state of the United States, was used to search for deaths that may have occurred outside Massachusetts. The data obtained from the National Death Index were reviewed, and the death certificates of all of the names that appeared to match or closely match the study subjects were obtained from the vital statistics office of every state where the deaths potentially occurred. The same information described above was used to confirm the match between the deceased and the corresponding subjects in the study. The date of death and the immediate, underlying, and contributing cause(s) of death were abstracted from the death certificates and coded according to the rules of the International Classification of Diseases, Injuries, and Causes of Death, Ninth Revision (15). Statistical analysis Cox's proportional hazard survival analysis was used to examine the association between mortality and vitamins C and E and carotenoids in plasma and in food as absolute values and as food groups contributing vitamin C and carotenoids to the diet. For these analyses, we categorized subjects into quintiles on the basis of their nutrient levels. Quintiles of nutrients were represented in the proportional hazard models as indicator variables. Regression coefficients and standard errors were used to estimate relative risks and 95 percent confidence intervals. The values in the first quintile (below the 20th percentile) represented low levels and were used as the reference category, quintiles 2-4 represented the middle range, and the values in the fifth quintile (above the 80th percentile) were the high levels. Serum cholesterol was included in all models that examined the association between plasma nutrients and mortality. Daily mean gram intakes of food groups were also examined in quintiles. All fruits and then separately citrus fruit/juices (the food subgroup containing foods with the highest contributions of vitamin C to the diet) and all vegetables and then dark green/orange vegetables (the food subgroup representing foods with the highest contributions of carotenoids to the diet) were examined for their association with mortality. Tests for trend were performed by fitting the variables in their continuous form in the proportional hazards model. All two-nutrient interaction terms were included in the models to test for synergistic effects between the three nutrients in both plasma and total intake. Associations between nutrients and overall mortality and mortality from cancer, cardiovascular disease, and other causes were examined. When mortality from one of these three diagnostic categories was fitted into a model, deaths from the other two categories were treated as censored data. RESULTS Selected characteristics of the 725 subjects for whom we had serum or dietary data or both are presented in table 1 for men and women separately. The study population was composed of 254 (35 percent) men and 471 (65 percent) women, with an average age of 72.3 years and 73.7 years, respectively. A total of 72.3 percent of the men and 73.0 percent of the women were still alive as of January Underlying causes of death included cardiovascular disease (n = 108), cancer (n = 64), and other causes (n = 45). Other causes included deaths from respiratory disease (n = 20), endocrine/metabolic disease (n = 8), genitourinary disease (n = 5), digestive disease (n = 4),

5 Carotenoids, Vitamins C and E, and Mortality 505 injuries/poisoning (n = 4), nervous system disease (n 3), and infectious/parasitic disease (n = 1). A total of 33.1 percent of the men and 43.7 percent of the women took supplements containing vitamin C, and 28.0 percent of the men and 40.3 percent of the women used supplements containing vitamin E. However, the majority of these nutritional supplements were in the form of multivitamins rather than individual supplements of vitamin C (9.9 and 12.7 percent for men and women, respectively) or vitamin E (4.8 and 9.3 percent for men and women, respectively). Table 1 also presents mean plasma concentrations and total intakes of carotenoids and vitamins C and E and mean gram consumption of food groups. Values identifying the lowest and highest quintiles of plasma, total intakes of carotenoids and vitamins C and E, supplemental intake of vitamins C and E, and food groups containing carotenoids and vitamin C are presented in table 2 for men and women combined. Plasma antioxidants The age-, sex-, and serum cholesterol-adjusted associations between low, medium, and high plasma nutrient levels and mortality are presented in table 3. Table 3 also presents the associations adjusted for additional potential confounders, such as presence of TABLE 2. Cutoff values for lowest and highest quintiles, Massachusetts, Nutrient/Tood Plasma measures Plasma vttamin C (mg/dl) Plasma vttamin E (mg/dl) Plasma carotenoids (ng/dl) Dietary plus supplement intake Vitamin C (mg) Vitamin E (mg) Carotenoids (ng) Food groups All fruit (g) Citrus fruit (g) All vegetables (g) Dark green/orange vegetables (g) Supplemental Intaket Vitamin C (mg) Vitamin E (mg) 20th , * Percent! es 80th , * 33.1 percent of the subjects had 0 g Intake of this vegetable subgroup. t Percentile distribution of supplement use among supplement users: 40 percent were vttamin C users, and 39 percent were vitamin E users. existing diseases and disabilities affecting the ability to shop, and for the other two antioxidant nutrients under study. Other variables, such as smoking, body mass index, alcohol consumption, and medication use, were examined to assess their potential as confounders. However, none of these variables were observed to alter the associations between plasma nutrient levels and mortality. Participants with plasma vitamin C levels in the middle and high quintiles had lower mortality than did those in the lowest quintile {p for trend < 0.005). The data in table 3 show that these associations held true even after controlling for age, sex, serum cholesterol, presence of existing diseases, disabilities affecting shopping, and plasma levels of vitamin E and carotenoids (p for trend < 0.03). The relative risks of mortality and plasma vitamin C adjusted for all potential confounders did not markedly change when deaths that occurred within the first 3 years of participation in the Nutrition Status Survey were excluded from the analysis (relative risk () = 0.73, 95 percent confidence interval (CI) and = 0.51, 95 percent CI , for middle and high plasma vitamin C values, respectively, compared with lower values). An inverse association was found between plasma carotenoid levels and mortality even after controlling for age, sex, and serum cholesterol (p for trend < 0.008). Relative risk of death was 0.67 (95 percent CI ) and 0.56 (95 percent CI ) for participants in the middle and highest quintiles, respectively, compared with those in the lowest quintile. However, the association was no longer significant after adjustment for potential confounders such as existing diseases, disabilities affecting shopping, and the other two antioxidant nutrients under study. No associations were found between plasma vitamin E and mortality (table 3), and no two-way interactions were found between the plasma antioxidant nutrients. The associations between plasma nutrients and mortality were examined by causes of death, controlling for age, sex, serum cholesterol, disease status, and disabilities affecting shopping (table 4). The results indicated that plasma vitamin C was inversely associated with mortality from heart disease (p for trend = 0.07). Those subjects with medium levels of plasma vitamin C had approximately one half the risk of heart disease compared with those who had lower levels ( = 0.51, 95 percent CI ). Subjects in the highest quintile had a relative risk similar to those in the middle quintiles but with a wider confidence interval because of a smaller number of subjects ( = 0.53, 95 percent CI ). Plasma carotenoids were associated with mortality from other causes (p

6 506 Sahyoun et al. TABLE 3. Adjusted relative risks (), 95 percent confidence Intervals (Cl), and p values for trend between plasma nutrients and overall mortality, Massachusetts, Levels of plasma nutrients Nutrients Adjusted for Low ()* Medtumt * pfortrend Plasma vitamin C, serum cholesterol Multivariatell Multivariatell plus vitamin E and carotenoids Plasma carotenoids Plasma vitamin E, serum cholesterol Muittvariatell Multivariatell plus vitamin C and vitamin E, serum cholesterol Multivariatell Multivariatell plus vitamin C and carotenoids * Referent category (quintjie 1). t Qutntites 2-4. t Qulntile 5. Tests for trend were performed by fitting the variables In their continuous form in the proportional hazards model. II Adjusted for age, sex, serum cholesterol, disease status, and disabilities affecting shopping. TABLE 4. Adjusted relative risks* () and 95 percent confidence Intervals (Cl) between plasma nutrients and mortality by causes of death, Massachusetts, Nutrients Plasma vitamin C p for trendt Plasma carotenows p for trendt Plasma vitamin E p for trendt MortaBty from cancer (n = 45) Mortality from heart disease (n = 75) Mortality from other causes (n = 33) * Risk versus low nutrient quintjie. Adjusted for age, sex, serum cholesterol, disease status, and disabilities affecting shopping. t Tests for trend were performed by fitting the variables in their continuous form in the proportional hazards model for trend < 0.03). No other significant associations were observed (table 4). To examine whether supplement users were responsible for the associations between mortality and the plasma nutrients, subjects who used multivitamins and vitamin C and E supplements were excluded from the analysis. The trend in risk of mortality was weaker across the more limited range of plasma vitamin C ( = 0.77, 95 percent Cl and = 0.72, 95 percent Cl for medium and high levels of plasma vitamin C compared with subjects in the lowest quintile) than trends including supplement users. The relation between plasma vitamin E and mortality was not appreciably affected by exclusion of supplement users. No participants reported use of supplements containing carotenoids, and exclusion of supplement users had no influence on the association between plasma carotenoids and mortality.

7 Carotenoids, Vitamins C and E, and Mortality 507 Dietary and supplemental intake of antioxidants Table 5 presents the association between mortality and dietary carotenoids and dietary plus supplemental intakes of vitamins C and E. Although no statistically significant trend was observed between total vitamin C intake and mortality (p for trend < 0.22), participants in the highest quintile had a significantly lower risk of death relative to those in the lowest quintile even after multivariate adjustment ( = 0.55, 95 percent CI ). Other potential confounders such as calorie intake, measurement of overall dietary adequacy, total vitamin A and folate intake, multivitamin use, chewing problems, and cigarette smoking were examined. These adjustments did not alter the observed association between vitamin C intake and mortality. Crude dietary carotenoid intake was inversely associated with mortality (p for trend = 0.04). However, once adjusted for age and sex, the association was no longer statistically significant (p for trend = 0.16). Neither crude nor adjusted vitamin E intake was significantly associated with mortality. No two-way interactions were found between the intakes of nutrients under study. The associations between nutrient intakes and mortality were also examined by cause of death, controlling for age, sex, and other potential confounders (table 6). There was an inverse association between vitamin C intake and mortality from heart disease. Relative risks of death from heart disease for participants in the middle and highest quintiles relative to those in the lowest quintile were (95 percent CI ) and 0.38 (95 percent CI ), respectively, although the test for trend was not statistically significant (p for trend = 0.22). Carotenoid intake appeared to be inversely related to mortality from other causes of death. Relative risks of death from causes other than cancer and heart disease were 0.43 (95 percent CI ) for medium intakes and 0.38 (95 percent CI ) for high intake relative to low intake. Vitamin E intake was marginally associated with death from heart disease. The relative risks for the medium and high quintiles of intake relative to the lowest quintile were 0.68 (95 percent CI ) and 0.75 (95 percent CI ), respectively. To examine whether supplement use was responsible for the associations between overall mortality and vitamin C, subjects who used multivitamin-containing vitamin C and vitamin C supplements were excluded from the analysis. The cutoff levels for vitamin C were changed since 99 percent of the population consumed less than 308 mg of vitamin C from their diet. When the cutoff levels of less than 45, , and greater than 250 mg of vitamin C were used, no associations with mortality were found. The relative risks were 0.84 (95 percent CI ) and 8 (95 percent CI ) for medium and high vitamin C intakes, respectively, compared with low values. We also examined associations between intake of vitamin supplements and mortality. No association TABLE 5. Adjusted relative risks (), 95 percent confidence Intervals (CI), and p values for trend between nutrient Intakes and overall mortality, Massachusetts, Nutrients Vitamin C intake Carotenoids intake Adjusted for MultJvariatell MultJvariatell plus vitamin E and carotenoids MultJvariatell Muttivariatell plus vitamin C and vitamin E Low ()' Levels ol nutrient Intake Medlumf 95% CI * 95% CI pfortrerkl Vitamin E intake MultJvariatell Multivariatell plus vitamin C and carotenoids * Referent category (quintile 1). t Quintiles 2-4. t Quintile 5. Tests for trend were performed by fitting the variables in their continuous form in the proportional hazards model. II Adjusted for age, sex, serum cholesterol, disease status, and disabilities affecting shopping

8 508 Sahyoun et al. TABLE 6. Adjusted relative risks* () and 95 percent confidence intervals (Cl) between antioxidant nutrient Intakes and mortality by causes of death, Massachusetts, Nutrients Mortality from cancer (n=57) Mortality from heart disease (n=101) Mortality from other causes (n-41) Vitamin C Intaket p for trendy Carotenoid Intake p for trendy Vitamin E intake p tor trendt * Risk versus lowest quinrjle. Adjusted for age, sex, disease status, and disabilities affecting shopping, t Vitamins C and E intakes represent dietary plus supplementary intakes. t Tests for trend were performed by fitting the variables in their continuous form in the proportional hazards model. was found between subjects who took vitamin C supplement and mortality compared with those who used no multivitamin supplements or vitamins C and E supplements ( = 0.88, 95 percent Cl ). In addition, no association was found between subjects who took multivitamins and mortality compared with those who used no multivitamin supplements or vitamins C and E supplements ( = 0.89, 95 percent Cl -1.23). Finally, subjects who took vitamin E supplements were examined in association with mortality compared with subjects who did not take multivitamin or vitamins C or E. No association with mortality was observed ( = 1.44, 95 percent Cl ). Fruits and vegetables There were no significant associations between fruit consumption and mortality (table 7). However, the results indicated that overall mortality in the middle and high quintiles of consumption of all vegetables and in the highest quintile of consumption of dark green/orange vegetables were significantly lower than overall mortality in the lowest quintile (table 7). This association appeared to be due to lower mortality from heart disease and lower mortality from other causes (table 8). Intake of vegetables was inversely associated with mortality after controlling for age, sex, existing diseases, disabilities affecting shopping, and intake of citrus fruit and juices (p for trend < 0.01) (table 7). Other potential confounders, such as problems biting/ chewing, living alone, and calorie intake, did not account for these associations. The exclusion of supplement users did not markedly affect the relative risks of either food group. The correlation coefficient between intake of citrus fruit and juices and dietary vitamin C was 0.70 (p < ), and that between this food group and dietary carotenoids was 0.08 (p < 0.04). The correlation coefficient between the food group dark green/ yellow vegetables and dietary carotenoids was 0.55 (p < ), and that between this food group and dietary vitamin C was 0.32 (p < ). DISCUSSION In this study, we found an inverse association between overall mortality and both plasma levels and total intake of vitamin C. The inverse associations were specifically observed for mortality from heart disease. Vitamin C intake greater than approximately 400 mg was associated with lower mortality compared with vitamin C intake below 90 mg. It appears that the associations may be attributed mainly to higher dietary plus supplemental intake since vitamin C supplements without consideration of diet were not associated with mortality. We also failed to observe any association between mortality and vitamin C intake among nonsupplement users. Consideration of each component of intake individually may result in misclassification of intakes. The plasma vitamin C data, a more direct and objective assessment of vitamin C status than dietary intake, indicate that even medium values of vitamin C were inversely associated with mortality compared with low levels.

9 Carotenoids, Vitamins C and E, and Mortality 509 TABLE 7. Adjusted relative risks (), 95 percent confidence intervals (Cl), and p values for trend between fruit and vegetable consumption and overall mortality, Massachusetts, Levels of fruit and vegetable consumption Food groups Adjusted for Low ()* Medlurrrt Wgrcf pfortrend All fruit and Juices Age,sex MuttJvariatell MuttJvariatell plus dark green/orange vegetables Citrus fruit and iulces All vegetables Dark green/orange vegetables Mutttvariatell MurrJvariatell plus dark green/orange vegetables Muttivariatell MultJvariatell plus citrus fruit/juices MultJvariatell Muttivariatell plus citrus fruit/juices O Referent category (quintfle 1). t Quintiles 2-4. t Quintile 5. Tests for trend were performed by fitting the variables In their continuous form in the proportional hazards model. II Adjusted for age, sex, disease status, and disabilities affecting shopping. Other studies have also observed associations between vitamin C, overall mortality, and mortality from heart disease. Gey et al. (16) observed from a crosssectional survey that plasma status of vitamin C near marginal deficiency (median of 0.45 mg/dl) may be correlated with an increased risk of ischemic heart disease. In our study, the lowest quintile of plasma vitamin C was 0.91 mg/dl, well above the acceptable lower limit of 0.6 mg/dl, with only 8.8 percent of the population below this value. Enstrom et al. (17) examined the relation between vitamin C intake and mortality in the First National Health and Nutrition Examination Survey Epidemiologic Follow-up Study cohort. The authors reported strong inverse trends between high dietary plus supplemental intake of vitamin C, mortality from all causes, and mortality from cardiovascular diseases for both sexes. However, these authors did not exclude the possibility of confounding by other nutrients. Studies have shown inverse associations between dietary intake and supplemental intake of vitamin E and reduced risk of coronary heart disease (18-20). In our study, we did not find an association between intake of vitamin E and overall mortality even when we compared subjects who used vitamin E from supplements with those who did not take any multivitamin or vitamin E supplement. Although we did not observe a statistical significant association, our data are consistent with an inverse association between total vitamin E intake and heart disease mortality. Some studies have also reported inverse associations between plasma vitamin E and onset or mortality from heart disease (21, 22) and cancer (23, 24), however, the overall evidence is variable, hi our study, no associations were found between plasma vitamin E and overall mortality or mortality from heart disease or cancer. Plasma carotenoids were inversely associated with overall mortality. The magnitude of this association was reduced once existing diseases and disabilities affecting shopping were taken into account but remained marginally significant. The possibility of a small protective association cannot be ruled out. The association with carotenoids appeared to be largely confined to death from conditions other than heart disease and cancer. The major cause of death in this category was respiratory disease. Intake of fruit and the subgroup of citrus fruit and juices containing high vitamin C food sources was not associated with mortality. This is consistent with our observation that vitamin C intake from foods alone was not significantly associated with mortality. It may be possible that with food groups as with dietary

10 510 Sahyoun et al. TABLE 8. Adjusted relative risks* () and 95 percent confidence intervals (CI) between fruit and vegetable Intakes and mortality by causes of death, Massachusetts, Mortality from cancer MortaBty from heart disease Mortality from other causes Food groups 96% CI 95% CI 95% CI All fruit and juices pfortrendt Citrus fadt and juices pfortrendt All vegetables pfortrendt Dark green/orange vegetables p for trendt * Risk versus lowest quintile. Adjusted for age, sex, disease status, and disabilities affecting shopping, t Tests for trend were performed by fitting the variables In their continuous form In the proportional hazards model. vitamin C, 3-day food records may not be the best indicator of usual consumption of vitamin C. In fact, Basiotis et al. (25) reported that a minimum of 33 days and 19 days is required to estimate confidently the true average intake of vitamin C of a small group of men and women, respectively. The vegetable food group was inversely associated with mortality, but it is unclear which substances in this food group are responsible for this association (5, 26). The literature has often attributed such findings to carotenoids in food. However, the dietary data collection methods often used in those studies were not sensitive enough to reach such conclusions (27-30). For example, in one study, carotene intake was estimated from six foods by using a food frequency questionnaire. Only two of the foods were significantly associated with lower mortality, and these two foods are more important contributors of vitamin C than carotenes (27). In addition, Shekelle et al. (28) developed an approximate carotenoid index from summarized diet histories that collected retinol intake. Other studies used food questionnaires of limited scope (29, 30). Recently, more complete data have become available on carotenoid content of food, which we used in analyzing the food records in our study (13). Nevertheless, it is possible that 3-day food records do not accurately reflect carotenoid intake, resulting in misclassification. It is also possible that carotenoid intake might only be a marker for some other substance in vegetables that is associated with reduced mortality. This is supported by our observation that the protective association for all vegetables is stronger than the association for the dark green and orange vegetables, which are better sources of carotenoid intake. Some of our results may have been limited by the relatively small sample size. This study did not find associations between any of the nutrients and mortality from cancer. A large number of studies examined associations between antioxidants and cancer mortality but focused mainly on specific cancer sites rather than on the overall incidence of cancer. The results of the studies were quite varied (24, 31, 32). In our study, deaths from cancer were too few in number (<60) to examine site-specific cancers. We also looked for interactions among the nutrient antioxidants in their association with mortality but did not find any. It may be possible that our study had low power to detect interactions. Another limitation of this study is that plasma and dietary nutrients were measured at only one point in time. However, although replicates would help to characterize better the long-term diets and serum levels, the lack of replications would only tend to weaken any relations and would not explain the associations observed in this paper. The main advan-

11 Carotenoids, Vitamins C and E, and Mortality 511 tage of this study is that it examines in the same population associations between mortality and levels of three antioxidant nutrients measured from plasma, dietary intake, supplements, and food groups. In summary, the data show an inverse association between plasma vitamin C and total intake of vitamin C with overall mortality and mortality from heart disease. The data also suggest that vegetables contain substances that protect against early mortality. ACKNOWLEDGMENTS Funded in part with federal funds from the US Department of Agriculture Research Service under contract 53-3K The authors thank Dr. Gerard Dallal for his input and thoughtful contribution to this paper. REFERENCES 1. Healthy People 2000: national health promotion and disease prevention objectives. Washington, DC: United States Department of Health and Human Services, Public Health Service, (DHHS publication no. (PHS) ). 2. Halliwell B. Oxidants and human disease: some new concepts. FASEB J 1987;l: Block G. The data support a role for antioxidants in reducing cancer risk. Nutr Rev 1992;50: Burton GW, Ingold KU. Beta-carotene: an unusual type of lipid antioxidant. Science 1984;224: Willett WC. Vitamin A and lung cancer. Nutr Rev 1990;48: Block G, Menkes M. Ascorbic acid in cancer prevention. In: Moon TE, Micozzi MS, eds. Nutrition and cancer prevention, investigating the role of micronutrients. New York, NY: Marcel Dekker, Inc., 1989: Frei B, England L, Ames BN. Ascorbate is an outstanding antioxidant in human blood plasma. Proc Natl Acad Sci USA 1989,86: Esterbauer H, Gebicki J, Puhl H, et al. The role of lipid peroxidation and antioxidants in oxidative modification of LDL. Free Radic Biol Med 1992;13: Bieri JG, Evarts RP. Gamma tocopherol: metabolism, biological activity and significance in human vitamin E nutrition. Am J Clin Nutr 1974;27: Roe J, Keuther C. The determination of ascorbic acid in whole blood and urine through the 2,4-dinitrophenyldrazine derivative dehydroascorbic acid. J Biol Chem 1943; 147: Neeld JB Jr, Pearson WN. Macro- and micromethods for the determination of serum vitamin A using trifluoracetic acid. J Nutr 1963;79: Grand Forks Nutrient Database release Grand Forks, ND: USDA-ARS Grand Forks Human Nutrition Research Center, Mangels AR, Holden JM, Beecher GR. Carotenoid content of fruits and vegetables: an evaluation of analytic data. J Am Diet Assoc 1993;93: Tucker K, Rush D. Food choices of the elderly. In: Hartz SC, Rosenberg IH, Russell RM, eds. Nutrition in the elderly. The Boston Nutritional Status Survey. London, England: Smith- Gordon and Company, Ltd., 1992: International classification of diseases. Ninth revision. Clinical modification. Ann Arbor, MI: Commission on Professional and Hospital Activities, Gey KF, Brubacher GB, Stahelin HB. Plasma levels of antioxidant vitamins in relation to ischemic heart disease and cancer. Am J Clin Nutr 1987;45: Enstrom JE, Kanim LE, Klein MA. Vitamin C intake and mortality among a sample of the United States population. Epidemiology 1992;3: Rimm EB, Stampfer MJ, Ascherio A, et al. Vitamin E consumption and the risk of coronary heart disease in men. N Engl J Med 1993;328: Stampfer MJ, Hennekens CH, Manson JE, et al. Vitamin E consumption and the risk of coronary disease in women. N Engl J Med 1993;328: Knekt P, Reunanen A, Jarvinen R, et al. Antioxidant vitamin intake and coronary mortality in a longitudinal population study. Am J Epidemiol 1994; 139: Gey KF, Puska P, Jordan P, et al. Inverse correlation between plasma vitamin E and mortality from ischemic heart disease in cross-cultural epidemiology. Am J Clin Nutr 1991;53:326S- 34S. 22. Riemersma RA, Wood DA, Macintyre CCA, et al. Risk of angina pectoris and plasma concentrations of vitamin A, C, and E and carotene. Lancet 1991;337:l Knekt P, Aromaa A, Maatela J, et al. Vitamin E and cancer prevention. Am J Clin Nutr 1991;53:283S-6S. 24. Chow CK. Evaluation of publicly available scientific evidence regarding certain nutrient-disease relationships: 8C. Vitamin E and cancer. Bethesda, MD: Life Sciences Research Office, Federation of American Societies for Experimental Biology, Basiotis PP, Welsh SO, Cronin FJ, et al. Number of days of food intake records required to estimate individual and group nutrient intakes with defined confidence. J Nutr 1987; 117: Ziegler RG. Vegetables, fruits, and carotenoids and the risk of cancer. Am J Clin Nutr 1991;53:251S-9S. 27. Colditz GA, Branch LG, Lipnick RJ, et al. Increased green and yellow vegetable intake and lowered cancer deaths in an elderly population. Am J Clin Nutr 1985,41: Shekelle RB, Liu S, Raynor WJ, et al. Dietary vitamin A and risk of cancer in the Western Electric study. Lancet 1981;2: Dartigues JF, Dabis F, Gros N, et al. Dietary vitamin A, beta-carotene and risk of epidermoid lung cancer in southwestern France. Eur J Epidemiol 1990;6: Paganini-Hill A, Chao A, Ross RK, et al. Vitamin A, /3- carotene and the risk of cancer a prospective study. J Natl Cancer Inst 1987;79: Ross C. Evaluation of publicly available scientific evidence regarding certain nutrient-disease relationships: 8A. Vitamin A and cancer. Bethesda, MD: Life Sciences Research Office, Federation of American Societies for Experimental Biology, Sauberlich HE. Evaluation of publicly available scientific evidence regarding certain nutrient-disease relationships: 8B. Vitamin C and cancer. Bethesda, MD: Life Sciences Research Office, Federation of American Societies for Experimental Biology, 1991.