点击显示 收起
1 From the Centre for Ageing and Public Health, London School of Hygiene and Tropical Medicine
2 Supported by the UK Department of Health. Additional funding for homocysteine analyses was provided by the European Union (contract no. BMH 4-98-3549). The Trial of Assessment and Management of Older People is funded by the UK Medical Research Council and Departments of Health. 3 Reprints not available. Address correspondence to A Fletcher, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, United Kingdom. E-mail: astrid.fletcher{at}lshtm.ac.uk.
ABSTRACT
Background: Older persons are at risk of both poor nutrition and
increased oxidative stress. Plasma ascorbate concentrations fall
with increasing age, and concentrations of other antioxidants may
also be reduced.
Objective: The goal was to examine the association between antioxidants and mortality in older persons.
Design: We randomly selected persons aged 75-84 y from the lists of 51 British family practitioners taking part in a randomized trial of assessment of older persons. A total of 1214 participants provided a blood sample and were interviewed about their usual diet with the use of a food-frequency questionnaire. Statistical analyses were based on deaths after a median of 4.4 y of follow-up, and hazard ratios were estimated for quintiles of dietary or blood antioxidants.
Results: We found strong inverse trends for blood ascorbate concentrations with all-cause and cardiovascular disease mortality, which were only marginally reduced after adjustment for confounders or supplement use. Those in the lowest fifth (< 17 µmol/L) had the highest mortality, whereas those in the highest fifth (> 66 µmol/L) had a mortality risk nearly half that (hazard ratio = 0.54; 95% CI: 0.34, 0.84). Similar results were found after the exclusion of those subjects with cardiovascular disease or cancer at baseline (hazard ratio = 0.51; 0.28, 0.93). In fully adjusted models, there was no evidence for an influence of -tocopherol, ß-carotene, or retinol on total mortality. Dietary antioxidants measured by the food-frequency questionnaire were not associated with all-cause or cardiovascular disease mortality.
Conclusion: Low blood vitamin C concentrations in the older British population are strongly predictive of mortality.
Key Words: Antioxidant vitamins vitamin C older persons prospective study mortality
INTRODUCTION
Studies conducted predominantly in middle-aged populations show associations between antioxidant vitamins and all-cause or cardiovascular disease (CVD) mortality, although
much debate remains about the role of specific antioxidants and
the benefit of supplementation (1, 2). Relatively few studies
have been conducted in older age groups, however, especially
in persons aged =" BORDER="0"> 75 y, who are at risk of both poor nutrition
and increased oxidative stress (3). Blood concentrations of
potent antioxidants, in particular plasma ascorbate (vitamin C),
decrease with increasing age (4) such that by late life increasing proportions of the older population have concentrations
indicating deficiency. In the British population, nearly 1 in 5
community-dwelling persons aged =" BORDER="0"> 75 y have ascorbate
concentrations indicating severe biochemical depletion, and
this figure increases to nearly one-half of those aged > 85 y
living in institutions (5). Low dietary intakes are considered to
be the main cause of the reduced plasma concentrations observed in later life (4), and conditions of acute free radical
generation, such as infection and inflammation, may dramatically reduce tissue stores of ascorbate (6, 7). The few studies
that have investigated the associations of antioxidant concentrations in late life with mortality were mostly small and
underpowered and showed varying results for the importance
of vitamin C, ß-carotene, or vitamin E (8-13). In the present
article, we report the results for plasma antioxidants and subsequent mortality in participants aged 75-84 y in the nutrition
substudy of a randomized trial conducted in Britain.
SUBJECTS AND METHODS
The Medical Research Council Trial of Assessment and
Management of Older People in the Community is a cluster
randomized trial conducted among family medical practitioners
(known as general practitioners in the United Kingdom). The
aim of the trial is to compare the cost-effectiveness of different
methods of assessment and clinical management of older persons. A total of 106 general practices from the Medical Research Council General Practice Research Framework were
recruited and selected to provide a representative sample of
mortality experience (standardized mortality ratio) and deprivation level (a measure of the socioeconomic characteristics of
the region) of the United Kingdom. The deprivation level was
measured with the use of Jarman scores derived from the 1991
census data (14 75 y on each general
practitioner's list were invited to participate in the trial unless
they were resident in a long-stay hospital or nursing home or
were terminally ill.
The practices were randomly divided into 2 groups: targeted or universal screening. In the universal screening practices, all participants underwent a detailed assessment conducted by the study nurse. In the targeted practices, only participants with a predetermined number and type of problems at the brief assessment underwent the detailed assessment. The detailed assessment included questionnaires for several health conditions, including the Rose Chest Pain Questionnaire (15). Patients were also asked about their current and past alcohol intake and smoking habits, sociodemographic factors (including marital status, living circumstances, and housing tenure), and medical history, including past and recent history of heart attack, stroke, cancer, and diabetes. Measurements taken at the detailed assessment included 2 measurements each of blood pressure (sitting and standing), height (with use of a stadiometer), and weight (with use of Soehnle scales; Leifheit AG, Nassau, Germany). Study nurses attended a 2-d training session. Participants were registered for mortality follow-up with the Office of National Statistics, who provided cause of death by using the 9th International Classification of Disease. The trial was approved by the relevant local research ethics committees.
The nutrition study was designed as an add-on study to the trial and was separately funded. Its main objective was to examine the association between blood concentrations of antioxidant vitamins and lipids (total, HDL, and LDL cholesterol) and mortality, especially from CVD. Dietary information was also collected. Written informed consent was obtained from the participants, and all relevant research ethics committees gave ethical approval for the nutrition study. Presented here are the results of the association of antioxidants with mortality.
Sample size
The nutrition study aimed to invite 3000 persons (with an
expected response rate of 70%) aged 75-84 y from the universal screening arm of the trial because all of the participants in
this arm would receive a full health assessment and would be
a representative group of older persons. Patients in the targeted
screening arm who had a detailed assessment were not included
in the nutrition study because they were a selected group. The
sample size was chosen to provide 80% power at a 5% significance level to detect a protective effect of 0.6 between the
highest and the lowest quintile of antioxidant distribution,
assuming a mortality rate of 60/1000 person-years. The upper
age limit of 84 y was chosen because it was judged that an
additional interview might be too burdensome in the oldest age
groups (those aged =" BORDER="0"> 85 y). The nutrition and physical activity
interview and nutrition study blood results were not part of the
trial interventions, and no attempt was made to provide participants with any information or advice regarding their blood
results or diet.
Sample selection
General practitioners in the 53 practices in the universal
screening arm of the trial were invited to take part in the
nutrition substudy. For each participating practice, persons
with birth dates within the eligible range were identified from
the age-sex registers, and a systematic random sample was
taken by applying a sampling fraction that had been predetermined for each practice to obtain a similar number of participants for each practice. At the detailed assessment, the nurse
invited the selected participants to take part in a further interview (to be held 4 wk after the assessment) and to provide an
additional 8 mL blood. To take account of seasonal variations
in diet, which might influence vitamin concentrations, the
invitation dates were evenly spread across the year.
Blood collection
All nurses in the nutrition study were instructed how to take
and process the blood samples. The nurses were asked to draw
the blood under subdued lighting, to collect the sample into 7-
or 9-mL EDTA-containing tubes, and to place the tubes immediately into an ice pack to be stored in a domestic refrigerator until shipment on ice in an insulated container to a local
laboratory within 4 h of collection. At the laboratory, the blood
was immediately centrifuged and divided into 7 aliquots. Metaphosphoric acid at a concentration of 10% was added to 2 of
the aliquots to prevent oxidation of vitamin C. The aliquots
were frozen to -80 °C. The local laboratories stored the blood
samples for a maximum of 6 mo until they were sent in batches
by 24-h courier on dry ice to Rowett Research Laboratories
(Aberdeen, United Kingdom) for further storage at -70° C.
Plasma ascorbate was analyzed by HPLC by using the assay procedure of Ross (16). The CV was 3.7%. Retinol, -tocopherol, and ß-carotene were also analyzed by HPLC by using the method of Hess et al (17); the percentage relative error compared with National Institute of Standards and Technology (Gaithersburg, MD) reference values for high, medium, and low concentrations of the fat-soluble vitamins were < 5% for 11 of 12 comparisons. Serum total and HDL cholesterol were by measured using the KONE instruments kit (KoneLab Corporation, Helsinki). Blood results were sent to the London School of Hygiene and Tropical Medicine and were merged with the main data files. Additional funding from the European Union Biomed Programme permitted the analysis of homocysteine at the University Department of Pharmacology, University of Oxford, by using a fluorescence polarization immunoassay on an Abbott IMx autoanalyzer (Abbott Laboratories, Abbott Park, IL).
Interviews
The nutrition and physical activity interviews were carried
out by the UK government's Office for National Statistics
Social Survey Division and took place after the blood collection (median time of 28 d). The nutrition interview used the UK
EPIC study (European Prospective Investigation into Cancer
and Nutrition) version of a food-frequency questionnaire (FFQ)
originally developed by Willett et al (18). We made some
minor modifications to the questionnaire; in particular, we
enquired about seasonal consumption of certain fruit and vegetables. The questionnaire included 139 food groups with a
choice of 1 of 9 response codes of frequency. Respondents
were asked about consumption over the previous year. Information on supplement use and type was also collected. Estimates of daily intakes of vitamin C, vitamin E, ß-carotene, and
retinol were calculated from the food groups by using the
nutrient databank that was made available to us by the investigators of the British National Diet and Nutrition Survey of
Older Adults (NDNS; 5); estimates of portion sizes were based
on the equivalent age group from the NDNS. Intakes from
supplements could not be included in the estimates of daily
intake because insufficient information on doses was obtained.
Because no appropriate physical activity questionnaire for this
UK age group was available, we adapted questions from the
Allied Dunbar National Fitness Survey (19) to cover a range of
activities from leisure activities, housework, and home maintenance. These were further categorized according to time
spent and intensity of effort.
Statistical analysis
Analyses were performed for respondents with a full set of
blood results, health assessments, and nutrition interviews to
allow for adjustment for potential confounders. The results
presented are weighted to allow for the differing planned
probabilities of selection in the practices. Analyses were performed using STATA 6 software (20). Mortality analyses used
Cox regression models and modified Wald tests for statistical
significance. Analyses took account of the cluster design in the
estimation of 95% CIs (21). Hazard ratios were estimated for
quintiles of plasma antioxidants referent to the lowest quintile
and were adjusted for age and sex. The P values for test for
trend were obtained from a logistic regression model in which
the quintiles were scored from 1 to 5 and a log-linear model for
odds of the outcome assumed. Further models took account of
possible confounders from data collected at the detailed assessment and described above. Smoking was classified as current,
ex, and never and pack-years were calculated from information
on lifetime smoking consumption. -Tocopherol concentrations were divided by total cholesterol to estimate the
-tocopherol ratio, and all analyses were based on the ratio of
-tocopherol to cholesterol. CVD was classified as a death
from an underlying cause due to hypertensive disease (ICD
401-405), ischemic heart disease (ICD 410-414), other heart
diseases (ICD 420-429), cerebrovascular disease (ICD 430-439), and diseases of the arteries (440-447). The analyses were
based on deaths reported by September 2001; the median
follow-up time was 4.4 y.
RESULTS
Of 53 practices in the universal arm of the study, 51 agreed
to participate in the nutrition study. A total of 2959 persons
were randomly sampled for the nutrition study, of whom 2167
attended the detailed assessment. One hundred twenty-seven
were not invited either because of delays in obtaining ethics
approval or because of administrative errors. The overall response rate to the invitation to the interview was 68% (1387)
and that to the blood collection was 75% (1529). A total of 120
blood samples were unusable or were not transported by the
local laboratory. The number of participants with both interviews and blood samples was 1214. Taking into account all of
the subjects sampled for the nutrition study, the response rate
was 47% for the interview and 52% for the blood sample.
The characteristics of those who provided a blood sample and completed an interview were not significantly different from those of either subjects in the universal arm of the study in the comparable age range or subjects sampled to take part in the nutrition study (Table 1). Of those with both a blood sample and a completed interview, 290 (24%) had died by September 2001, of whom 44% had died of CVD.
View this table:
TABLE 1. . Characteristics of persons eligible for the nutrition sample, of those selected for the present study, and of those who provided a blood sample and
completed the questionnaire1
Inverse associations existed between increasing blood ascorbate concentrations and age, the proportion of men, and current
smoking, and positive associations existed with the proportion
of homeowners, levels of physical activity, supplement use,
alcohol consumption, and total and HDL cholesterol (Table 2). Ascorbate concentrations were associated with dietary vitamin
C and ß-carotene intakes and with the proportion consuming
=" BORDER="0"> 5 fresh fruit and vegetables daily. There were fewer trends
for associations with -tocopherol (Table 3). Trends similar to
those observed for ascorbate were seen for ß-carotene (Table
4), but fewer were seen for retinol (Table 5).
View this table:
TABLE 2. . Characteristics of the participants by quintiles of plasma ascorbate
View this table:
TABLE 3. . Characteristics of the participants by quintiles of plasma -tocopherol
View this table:
TABLE 4. . Characteristics of the participants by quintiles of plasma ß-carotene
View this table:
TABLE 5. . Characteristics of the participants by quintiles of plasma retinol
We observed a strong inverse trend for all-cause mortality
and blood ascorbate, which was only marginally reduced after
adjustment for confounders, including sex and supplement use
(Table 6). Forty-three percent of the subjects reported taking
some form of vitamin supplement; the most common was cod
liver oil, which was taken by 25%, mainly on its own (21%).
Vitamin C alone was reported by only 5%, and the total taking
vitamin C either alone or as part of a multivitamin preparation
was 17%. Adjustment for any use of supplements had little
effect on the results. Separate analyses using sex-specific quintiles confirmed the associations for men and women with
ascorbate and mortality (data not shown); there was no evidence for a differential effect for men compared with women
(P value for test for interaction in the fully adjusted model =
0.83).
View this table:
TABLE 6. . Hazard ratios (95% CI) for all-cause and cardiovascular disease mortality by quintile (Q) of plasma antioxidant referent to the first quintile
-Tocopherol and retinol were not associated with mortality,
and a nonlinear association with ß-carotene disappeared after
adjustment for confounders. The results were essentially similar after the exclusion of those with prevalent CVD or cancer
at baseline (Table 6).
The results for CVD mortality also showed decreasing hazard ratios with increasing ascorbate concentrations, with no extra benefit in the top fifth compared with the fourth. For -tocopherol, there was weak evidence for a protective effect on CVD mortality for concentrations > 3.92 µmol/mmol cholesterol in the fully adjusted models, whereas for ß-carotene, a nonsignificant reduction in risk with increasing concentrations disappeared when adjusted for confounders. The hazard ratios for a 20-µmol/L increase in ascorbate were 0.82 for all-cause and CVD mortality, respectively, and 0.76 and 0.74, respectively, after the exclusion of those with prevalent CVD or cancer at baseline (Table 7). The hazard ratios for ascorbate were little altered by including all plasma antioxidants in the model, either in adjusted or unadjusted analyses. For example, in a fully adjusted model (adjusted for age, sex, BMI, cholesterol, systolic blood pressure, smoking, alcohol, diabetes, history of CVD or cancer, physical activity, housing tenure, and supplement use) that included -tocopherol, ß-carotene, and retinol, the hazard ratios by increasing fifths of ascorbate compared with the lowest quintile were 0.97 (0.72, 1.31), 0.59 (0.40, 0.86), 0.63 (0.36, 1.08), and 0.54 (0.34, 0.85) (P for trend = 0.004). None of the other antioxidants was significantly associated with mortality in this model.
View this table:
TABLE 7. . Hazard ratio per 20-µmol/L change in plasma ascorbate
The results were also essentially unchanged in other analyses
substituting HDL for total cholesterol or adjusting for folate
concentrations or using pack-years of smoking (data not
shown). In analyses restricted to 417 never smokers, the fully
adjusted (age, sex, BMI, cholesterol, systolic blood pressure,
alcohol, diabetes, history of CVD or cancer, physical activity,
housing tenure, and supplement use) hazard ratios for a 20-µmol/L increase in ascorbate were 0.73 (95% CI: 0.57, 0.93;P = 0.012) for all-cause mortality. We found a small attenuation of the estimates in analyses that additionally adjusted for
homocysteine (available for 78%). The fully adjusted hazard
ratios for a 20-µmol/L increase in ascorbate were reduced to
0.88 (95% CI: 0.76, 1.02) for all-cause mortality and 0.84 (95%
CI: 0.66, 1.08) for CVD mortality in the subset of those with
homocysteine results. Dietary antioxidants were not associated
with either all-cause or CVD mortality (Table 8).
View this table:
TABLE 8. . Hazard ratios (95% CI) by quintile (Q) of dietary antioxidant referent to the first quintile1
DISCUSSION
In the present study of an older British population, we found
strong associations between plasma vitamin C concentrations
and mortality, with individuals having higher concentrations
also having considerably reduced mortality. In common with
the results of the British NDNS (5) and with other studies,
which were conducted mainly in younger population groups (4,
22, 23), we found lower plasma vitamin C concentrations in
men, in current smokers, in those of lower socioeconomic
status, and in those with lower levels of physical activity.
However, we found no clear associations with diabetes, history
of CVD or cancer, systolic blood pressure, or BMI, for which
associations were reported in some but not all studies, although
in inconsistent directions. We observed a positive association
between ascorbate and total cholesterol, as in the National
Health and Nutrition Examination Survey (23) but in contrast
with the Norfolk EPIC study, which reported an inverse association (22). For HDL cholesterol, we also observed a positive
association with ascorbate concentrations that was similar to
the association reported in some studies (22), including those in
elderly persons (4). However, adjustment for these variables
only marginally affected our results. Homocysteine concentrations were negatively correlated with ascorbate (r = -0.27),
but adjustment for homocysteine led to only a small attenuation
of the hazard ratios.
The response rate in our study was 75% of those invited; taking into account the original numbers sampled, the response rate was just > 50%, equivalent to that of the Norfolk EPIC cohort of 45% (22). Our respondents were similar both to those randomly selected for the nutrition study and to the participants in the universal arm of the trial and are therefore likely representative of the British elderly in this age range. Ascorbate concentrations in our sample were comparable with those of community-dwelling older persons of similar ages in the British NDNS, in which median values were 36 µmol/L for men and 47 µmol/L for women (5) compared with 35 and 47 µmol/L, respectively, in our study. Our study group, however, included fewer persons with very low concentrations (< 11 µmol/L): 13% of men and 9.5% of women in the present study compared with 19% of men and 17% of women in the NDNS.
Ascorbate concentrations in the present study were also lower than in the Norfolk EPIC cohort (which had a mean age 60 y), but the association with mortality was similar (Table 7). The Norfolk EPIC investigators were unable to adjust for socioeconomic status or physical activity because they did not have this information. We showed that taking these variables into account had little effect on the results.
We found no clear evidence from the dietary data of an association between vitamin C or other dietary antioxidants and all-cause or CVD mortality. Although plasma antioxidant concentrations were correlated with the corresponding dietary antioxidant intakes, our correlations were lower than those reported by others [in our study the correlation between ascorbate and dietary vitamin C was 0.3 compared with 0.6 in the NDNS (5) and 0.4 in the Norfolk EPIC study (22)]. Those 2 studies used 4-d weighed intakes or 7-d diet histories, respectively, whereas our source of dietary vitamin intakes was an FFQ that asked about usual food intake over the previous year. The median time between blood collection and completion of the questionnaire in our study was 1 mo, so it is unlikely that the weaker correlations were substantially affected by this time difference. The Norfolk EPIC investigators also reported a correlation of 0.28 for ascorbate with dietary vitamin C intakes derived from an FFQ (24) and concluded there would be substantial measurement error in estimates derived from FFQs. Our experience also suggests that FFQs may not reliably categorize dietary antioxidant intakes in older age groups. Compared with the subjects of the NDNS, both the men and the women in our study had higher dietary intakes of vitamin C (72 mg/d in men and 85 mg/d in women compared with 50 and 43 mg/d, respectively, for the equivalent sample in the NDNS). Our respondents may have overreported their intakes, because plasma concentrations were similar in the 2 samples. We did not find strong evidence for associations for -tocopherol, ß-carotene, or retinol with mortality.
The limited evidence for an association between plasma antioxidant concentrations and mortality in older persons is more convincing for vitamin C than for other antioxidants. In a 20-y follow-up of participants in the first Diet and Nutrition Study of the British Elderly, lower death rates from stroke (30% reduction) were found in those with ascorbate concentrations > 28 µmol/L (10). A study in a Massachusetts population (with a mean age of 72 y) showed a lower risk of total mortality and a trend for CVD mortality in the highest and combined middle fifths of ascorbate compared with the lowest, smaller and nonsignificant protective effects for total plasma carotenoids (8), but no association with -tocopherol. A 7-y follow-up of a Dutch cohort of persons aged 65-85 y found the highest risks for total plasma carotenoids and mortality, specifically, for ß-cryptoxanthin and lutein but not for ß-carotene, lycopene, or -tocopherol (11), whereas a Finnish study found no effect of plasma -tocopherol (12). Ascorbate was not measured in either study. We did not have funds to measure carotenoids other than ß-carotene.
We found strong positive correlations between ascorbate concentrations and reporting of consumption of =" BORDER="0"> 5 fresh fruit and vegetables daily. The main foods associated with plasma vitamin C concentrations in our population were oranges, broccoli, and sweet peppers. It is therefore possible that the association between ascorbate and mortality was, at least in part, due to other dietary micronutrients.
Studies showing associations between plasma concentrations and outcomes may be prone to biases resulting from reverse causation, ie, lower concentrations of antioxidant vitamins reflect depletion as a result of concomitant disease, itself a marker for subsequent mortality and morbidity. It is unlikely that reverse causation explains our results, however: the association was linear across the ascorbate distribution, the estimates agreed well with those found in prospective studies of younger cohorts, and the associations remained after the exclusion of those with prior CVD or cancer. The concentrations of vitamin C observed at older ages along with the increased stresses on the antioxidant defense system has led to concerns that concentrations in older populations are too low (25) and to recommendations to increase the dietary allowance for vitamin C (26). The older British population is a special worry because ascorbate concentrations, which are generally low in both the British population and in the older population, are low compared with concentrations in some other countries. The lowest quintile of plasma ascorbate at highest risk in the Massachusetts elderly study (< 52 µmol/L) would have included 60% of our study population, whereas < 10% would have been in the most protected quintile (> 89 µmol/L) (8).
A key question is how to increase concentrations of ascorbate in older age groups. Enthusiasm for vitamin supplementation has been tempered by the negative results from randomized trials, which were conducted predominantly in middle-aged populations, although the largest trial did include persons up to the age of 80 y (27). Supplementation with chemical forms of a few antioxidants cannot substitute for the richness and variety of dietary sources of antioxidants. Antioxidants act in synergy, and many antioxidants (of which vitamin C is one of the most powerful) appear to be involved in a cascade of radical-quenching reactions (28). Thus, the best recommendation for older persons, as for middle-aged and younger persons, is to maintain a diet rich in a variety of antioxidant micronutrients. At older ages, however, several factors, such as reduced appetite and taste, poor dentition, physical and economic barriers to food sources, and lack of motivation, present formidable challenges to this strategy (29). The evidence for the success of interventions to promote healthy eating in elderly persons is weak, at best, and nonexistent for the United Kingdom (30). More attention needs to be given to addressing the barriers to the adoption of healthy diets in old age.
ACKNOWLEDGMENTS
We thank Anna Whyte at the Rowett Institute for advice on the handling
and transport of blood samples, Robert Clarke and C Johnston at the
University of Oxford for homocysteine analysis, and our colleagues from
the ONS: Jill Matheson, then Director of the Social Survey Division;
Michaela Pink, field manager; and the interviewers who administered the
nutrition questionnaire.
AEF and PSS designed the nutrition study. EB coordinated the study and carried out the statistical analyses. AEF and EB specified the analyses, and AEF wrote the paper. PSS and EB contributed critical comments to the paper. None of the authors had a conflict of interest to report.
REFERENCES