There is evidence of an inverse relationship between fruit and vegetable consumption and the risk of CVD and cancers(Reference Lock, Pomerleau, Causer, Altmann and McKee1–5). Consequently, promoting the consumption of at least five portions of fruits and vegetables (400 g) per d has become a national health priority in many countries(6).
Several mechanisms have been proposed to explain this positive association, including an increased intake of antioxidant compounds that are widely found in fruits and vegetables. Trials in which antioxidant micronutrients were taken at high doses over long periods have not confirmed this potential beneficial effect(7–Reference Vivekananthan, Penn, Sapp, Hsu and Topol11) and even suggested harmful effects(7, Reference Bjelakovic, Nikolova, Gluud, Simonetti and Gluud12, Reference Omenn, Goodman and Thornquist10). However, at nutritional doses, antioxidant supplementation reduced the risk of cancer in men in the Supplementation with Antioxidant Vitamins and Minerals (SU.VI.MAX) study(Reference Galan, Briancon and Favier13, Reference Hercberg, Galan, Preziosi, Bertrais, Mennen, Malvy, Roussel, Favier and Briancon14) and reduced cancer risk and total mortality in the Linxian study(Reference Blot, Li and Taylor15, Reference Li, Taylor and Li16).
It is now considered that whole fruit and vegetables would confer a greater protection than that achieved by supplementation with a single component, through an interaction effect between several nutrients(Reference Norman, Go and Butrum17, Reference Donaldson18).
Given the strong relationship between fruit and vegetable consumption and health, it would be useful to understand which serum antioxidants and what concentrations are associated with the recommended consumption of fruit and vegetables. In addition, among all fruit and vegetables, it should be clarified whether fruit, vegetables or specific foods from these groups are more strongly associated with these serum concentrations.
Finally, few studies have examined the relationships between biological serum concentrations of antioxidants and fruit and vegetable consumption in European populations(Reference Al Delaimy, Ferrari and Slimani19, Reference Ferrari, Al Delaimy and Slimani20). The objective of the present study was to investigate the relationship between fruit and vegetable (amount and type) consumption estimated by repeated 24 h records and vitamin serum concentrations, in a large sample of middle-aged adults from France.
Methods
Subjects
Subjects were participants of the SU.VI.MAX study, a randomised, double-blind, placebo-controlled, primary-prevention trial designed to evaluate the effect of a daily antioxidant supplementation at nutritional doses on the incidence of cancers and IHD. The design, methods and rationale of the SU.VI.MAX study have been reported elsewhere(Reference Hercberg, Galan, Preziosi, Bertrais, Mennen, Malvy, Roussel, Favier and Briancon14, Reference Hercberg, Preziosi, Briancon, Galan, Triol, Malvy, Roussel and Favier21). Briefly, 13 017 participants (7876 women aged 35–60 years and 5141 men aged 45–60 years) were included in 1994–1995 with a planned follow-up of 8 years. To be included in the study, subjects should not have declared using vitamin supplements. Participants were invited annually for either a clinical or a biological examination and were asked to regularly provide information on health events and dietary habits.
The SU.VI.MAX study was approved by the Ethical Committee for Studies with Human Subjects of Paris-Cochin Hospital (CCPPRB no. 706) and the Comité National Informatique et Liberté (CNIL no. 334641).
Dietary assessment
The subjects were asked to provide a 24 h dietary record every 2 months, for a total of six records per year, so that weekdays, weekends and all seasons were covered. Information was collected via computerised questionnaires using the Minitel Telematic Network containing specific software. The Minitel was a small terminal widely used in France as an adjunct to the telephone at the beginning of the study. Participants were assisted by the conversational features of the software and used an instruction manual for coding food portions. This reference manual includes previously validated photographs(Reference Le Moullec, Deheeger, Preziosi, Monteiro, Valeix, Rolland-Cachera, Potier De Courcy, Christides, Galan and Hercberg22) of more than 250 foods (corresponding to 1000 generic foods) represented in three different portion sizes. Subjects could also choose from two intermediate or two extreme portions, for a total of seven different possible portion sizes.
Blood sampling
Blood samples were collected at enrolment (before supplementation) from participants, who had been fasting for 12 h, in a mobile medical unit or a preventive health centre. All biochemical measurements were performed in only one laboratory for each nutrient being analysed. All measurements were carried out between November and April, which corresponds to a period of low fruit and vegetable consumption.
Blood samples for β-carotene and α-tocopherol measurements were centrifuged immediately, then serum was frozen and kept at − 80°C until HPLC determination using Arnaud's method(Reference Arnaud, Fortis, Blachier, Kia and Favier23), with minor modifications. The HPLC system was a Biotek-Kontron (Montigny le Bretonneux, France) including a 525 gradient pump, a 402 Peltier Oven, a 465 auto sampler, a Geminyx data station and a 540 diode array detector.
To assess vitamin C concentration (dehydroascorbic acid and ascorbic acid), the serum was diluted (1:10) immediately in 10 % metaphosphoric acid(Reference Bourgeois, Chartois, Counstans and George24, Reference Margolis and Duewer25) and kept frozen at − 80°C until determination by spectrofluorimetry(Reference Bourgeois, Chartois, Counstans and George24). A Technicon continuous flow analysis apparatus (Bayer-Technicon, Pittsburgh, PA, USA) that was equipped with a RF-530 Shimatzu fluorescence detector (Shimatzu Corporation, Kyoto, Japan) was used.
The laboratory quality insurance included analysis of serum from standard pools with each run and, if available, international standards (markers ProBioQual). CV were determined by measuring concentrations in aliquots five times on the same day and on ten different days: 9·02 % for β-carotene; 3·52 % for vitamin C; 3·54 % for α-tocopherol. Total cholesterol and TAG were measured by enzymic methods (Technicon DAX, Tarrytown, NY, USA).
Other data collection
Information about physical activity (irregular, < 1 h walking/d, ≥ 1 h walking/d), smoking status (non-smoker, former smoker, current smoker) and education level (primary school, high school, university or equivalent) was self-reported by participants on a questionnaire at enrolment. BMI was calculated using weight and height measured during a clinical examination 1 year after enrolment. When these measures were not available (n 482), self-reported weight and height from the enrolment questionnaire were used.
Inclusion criteria
A sub-sample of the population of 8280 subjects had β-carotene, vitamin C and α-tocopherol serum concentrations measured at inclusion. Subjects with at least six or more 24 h records completed in the 2 years following inclusion were included for the present analyses (n 4195). An additional 21 subjects were excluded because more than one-third of their 24 h records were lower than 2090 kJ (women) or 3334 kJ (men), leaving 4174 subjects.
Fruit and vegetable consumption can vary greatly according to season. Therefore, participants were selected if at least one-third of their reported 24 h records were from the low season (November to April) and another third from the high season (May to October) of fruit and vegetable intakes, leaving 3861 individuals. A total of 360 subjects were further excluded because at least one adjustment variable was missing, leaving a total of 1487 men and 2034 women included in the present analysis.
Statistical methods
Fruits and vegetables from composite dishes were included, but potatoes as well as dried fruits and legumes were excluded. Subgroups of fruit and vegetables as described in Table 1 were investigated. However, fruit and vegetable groups in which intake levels were too small to be significant (e.g. cabbage) were not considered. Groups of fruit mainly consumed during the summer (i.e. stone fruit, melons, berries) were not studied. Their consumption is not close enough to the blood sampling period and therefore would not be able to explain serum antioxidant concentrations measured.
Table 1 Groups of fruits and vegetables
The relationship between antioxidants and all fruit and vegetables was assessed using Spearman's rank correlation. β-Carotene and α-tocopherol concentrations were adjusted on plasma cholesterol and TAG concentration using the standard value of 5·18 mmol/l for cholesterol and 1·25 mmol/l for TAG(Reference Willett26, Reference Jordan, Brubacher, Moser, Stahelin and Gey27). Fruit and vegetables were adjusted on energy intake(Reference Willett26). Generalised linear models were used to perform multivariate analyses. In the multivariate analyses, logarithmic transformation of the distribution was used because of skewed distribution of serum antioxidant concentrations. Adjusted geometric means of serum antioxidants are presented for each quintile of fruit and vegetable groups. Trend tests using fruit and vegetable intakes as continuous variable were performed. The multivariate analyses were adjusted for potential predictors of antioxidant concentration: sex; age at inclusion (continuous); BMI (continuous); physical activity (three categories); smoking status (three categories); education level (three categories); energy intake (continuous); alcohol consumption (continuous); month of blood sampling (one category for each month); cholesterol and TAG concentrations (continuous). When analysing groups of fruit or vegetables, all other fruit or vegetables were included in the model. We used four different types of models. In addition to adjusted variables, the first model included ‘fruit and vegetables’, the second, ‘fruit’ and ‘vegetables’, the third, ‘fruit juices’, ‘fruit without juices’ and ‘vegetables’ and the last included all subgroups of vegetables. Statistical analyses were performed using SAS software (The SAS Institute Inc., Cary, NC, USA).
Results
Characteristics of selected men and women at baseline are presented in Table 2. As expected by the sampling method, mean age was higher in men than women. Men had a higher mean BMI, higher physical activity and a lower educational level than women. The percentage of current smokers was similar in both groups. Women consumed less energy and less alcohol. Women consumed 24 g/d more fruit and 15 g/d more vegetables than men after energy intake adjustment. The most consumed vegetables were fruiting vegetables and root vegetables for both men and women, while the most consumed fruit were pome fruit and stone fruit. Fruit juices represented 17·0 % of fruit intake in men and 17·8 % in women. Quintiles medians of crude intake are shown in Table 3.
Table 2 Characteristics of the study population†
(Mean values and standard deviations)
χ2 for qualitative variables and t test for quantitative variables.
Fruit and vegetable consumption is adjusted on energy.
Plasma antioxidant level is adjusted on cholesterol and TAG.
Values were significantly different: *P < 0·05; **P < 0·01; ***P < 0·001.
† For details of subjects and procedures, see Methods.
‡ Not included in the analysis on the relationship with antioxidants.
Women had higher β-carotene and vitamin C serum concentrations than men, but lower α-tocopherol serum concentration (Table 1). Serum β-carotene and vitamin C concentrations were correlated with fruit and vegetable consumption, as well as with most of the food groups tested (Table 4). In contrast, serum α-tocopherol correlated poorly with a few of the food categories tested. Correlations for β-carotene ranged from − 0·03 to 0·25 and were correlated with both vegetables (r 0·24) and fruit (r 0·18). Root vegetables and citrus fruit were particularly correlated with β-carotene. Correlations for vitamin C ranged from 0·03 to 0·34 and were greater for fruit than for vegetables. Vitamin C correlated with fruit as well as with fruit+fruit juices and with citrus fruit. Correlation for α-tocopherol ranged from 0·00 to 0·09. Finally, α-tocopherol was significantly but poorly correlated with vegetable (0·05) and fruit (0·04) consumption. Correlations with vegetable groups were mostly non-significant, with the exception of leafy vegetables, bulb and stem vegetables and citrus fruit.
Table 4 Spearman's correlation coefficients and significance for the correlation between plasma antioxidant concentration adjusted for cholesterol and TAG and intake of vegetables, fruits and juices adjusted on energy (n 3521)†
Values were significantly different: *P < 0·05; **P < 0·01; ***P < 0·001.
† For details of subjects and procedures, see Methods.
After adjusting for confounding factors, higher fruit and vegetable consumption was associated with higher serum concentrations of β-carotene and vitamin C with a linear dose–effect relationship and significant P for trend (Table 5). Specifically, the serum concentration of β-carotene was related to vegetables in general and to the root, leafy and fruiting vegetable groups as well as green beans and peas. In addition, fruit, fruit + juices and particularly pome and citrus fruit were related to β-carotene status. Serum vitamin C concentrations were related to the same food products with the addition of tropical fruits and fruit juices and the exclusion of green beans and peas. Vitamin C was more strongly associated with fruit than with vegetables. On the other hand, no significant relationship was found for α-tocopherol. Analysis of adjusted geometric means of antioxidant concentration suggested a linear dose–response relationship for all groups of fruit and vegetables associated with antioxidants.
Table 5 Adjusted geometric means of plasma levels within quintiles of intake (n 3521)*
Adjusted for gender, age, BMI, physical activity, smoking status, education level, energy intake, alcohol consumption, blood sampling month, cholesterol and TAG.
Fruit and vegetable groups are adjusted for all other groups.
* For details of subjects and procedures, see Methods.
Discussion
The present results indicate that intakes of fruit and vegetables are moderately correlated with serum concentrations of β-carotene and vitamin C but not with α-tocopherol. Other factors influencing antioxidant concentrations, such as sex, age, smoking status, season, dietary intake and location of residence, have been identified in previous analyses from the SU.VI.MAX study(Reference Faure, Preziosi, Roussel, Bertrais, Galan, Hercberg and Favier28, Reference Galan, Viteri and Bertrais29).
In the current study, mean fruit and vegetable intakes for both sexes were higher than described in other studies performed on the French population(Reference Guilbert and Perrin-Escalon30–Reference Touvier, Kesse, Volatier, Clavel-Chapelon and Boutron-Ruault32). This may be partly due to the sample selection, as participants who completed at least six dietary records might be specifically interested in health and diet and may thus show a better dietary profile. The participants' higher level of education may also partly explain high fruit and vegetable consumption(Reference Irala-Estevez, Groth, Johansson, Oltersdorf, Prattala and Martinez-Gonzalez33).
A number of studies have examined the correlation between fruit and vegetable intakes and plasma nutrients. β-Carotene was consistently found to be associated with intakes of fruit and/or vegetables in a number of observational(Reference Block, Norkus, Hudes, Mandel and Helzlsouer34–Reference van Kappel, Steghens, Zeleniuch-Jacquotte, Chajes, Toniolo and Riboli41) and intervention(Reference Brevik, Andersen, Karlsen, Trygg, Blomhoff and Drevon42–Reference Zino, Skeaff, Williams and Mann49) studies. However, Al Delaimy et al. (Reference Al Delaimy, Ferrari and Slimani19, Reference Al Delaimy, Slimani and Ferrari50) suggested that the correlation of plasma β-carotene with fruit and vegetable intake was low compared with other carotenoids. Although at the lower end of the range, the correlation coefficients observed in the present study (fruit+vegetables+juices: r 0·25; fruit: r 0·19; vegetables: r 0·24) were comparable with those found in previous observational studies, ranging from 0·17 to 0·45 for fruit and vegetables, from 0·18 to 0·45 for fruit and from 0·18 to 0·34 for vegetables, using either Pearson or Spearman correlation coefficients. In the present investigation, fruits were found to have comparable associations with serum concentrations of β-carotene rather than vegetables. These results are in agreement with those of Drewnowski et al. (Reference Drewnowski, Rock, Henderson, Shore, Fischler, Galan, Preziosi and Hercberg36), who showed similar correlation coefficients for fruit and vegetables. However, they contrast with previous studies(Reference Campbell, Gross, Martini, Grandits, Slavin and Potter35, Reference Jansen, van Kappel, Ocke, Van't Veer, Boshuizen, Riboli and Bueno-de-Mesquita37, Reference van Kappel, Steghens, Zeleniuch-Jacquotte, Chajes, Toniolo and Riboli41), which found greater correlation coefficients for fruit than for vegetables. In the current study, fruit juices had a weak or no association with serum concentrations according to the regression analysis. Few authors have also investigated the importance of fruit juice contributing to nutrient serum concentrations. Two studies found a limited positive(Reference Jansen, van Kappel, Ocke, Van't Veer, Boshuizen, Riboli and Bueno-de-Mesquita37, Reference Pollard, Wild, White, Greenwood, Cade and Kirk38) association and one study reported an inverse association(Reference van Kappel, Steghens, Zeleniuch-Jacquotte, Chajes, Toniolo and Riboli41). In addition, McEligot et al. (Reference McEligot, Rock, Shanks, Flatt, Newman, Faerber and Pierce51) found no significant difference in serum values of β-carotene between subjects consuming juice and those not consuming any juice.
Only a few studies looked at possible differences in serum nutrient concentrations for particular fruit and vegetables. In the present study, the root vegetables category, which includes carrots, was shown to be specifically associated with serum concentrations of β-carotene. In addition, there was a stronger association for citrus and pome fruit compared with tropical fruit. Similarly, the vegetables most associated with β-carotene status were root vegetables(Reference Al Delaimy, Ferrari and Slimani19, Reference Al Delaimy, Slimani and Ferrari50) or more specifically carrots(Reference Campbell, Gross, Martini, Grandits, Slavin and Potter35, Reference Tucker, Chen, Vogel, Wilson, Schaefer and Lammi-Keefe40), in subjects with a wide variety and regular consumption of vegetables and fruits. In another study, both carrots and tomatoes were associated with β-carotene status(Reference Michaud, Giovannucci, Ascherio, Rimm, Forman, Sampson and Willett52), whereas Jansen et al. (Reference Jansen, van Kappel, Ocke, Van't Veer, Boshuizen, Riboli and Bueno-de-Mesquita37) showed the specific importance of tomatoes and cabbages. It should be noted that only fruits consumed in autumn and winter were considered in the current study, as blood sampling was carried out exclusively during this period. Therefore, the association between berries, stone fruit and melons with antioxidant serum concentrations could not be investigated. In addition, it is likely that the association with fruit or vegetables mostly consumed during these periods, such as citrus fruit, is overestimated.
A possible increase of cancer risk with an antioxidant supplementation has been emphasised in intervention trials(7, Reference Omenn, Goodman and Thornquist10). However, a concentration of 0·53 μmol/l found in the present study for β-carotene is far lower than concentrations of 5·6 μmol/l(7) or 3·9 μmol/l(Reference Omenn, Goodman and Thornquist10) observed in these trials where high doses of pharmacological supplements were given.
The relationship found between vitamin C status and usual consumption of fruit and vegetables has been consistently found in observational studies(Reference Block, Norkus, Hudes, Mandel and Helzlsouer34, Reference Drewnowski, Rock, Henderson, Shore, Fischler, Galan, Preziosi and Hercberg36, Reference Pollard, Wild, White, Greenwood, Cade and Kirk38, Reference Ness, Khaw, Bingham and Day53). The correlation coefficient found for vegetables+fruit+juices (r 0·28) is similar to the value found by Drewnowski et al. (Reference Drewnowski, Rock, Henderson, Shore, Fischler, Galan, Preziosi and Hercberg36) for fruit and vegetables (r 0·29), but lower than the one found by Block et al. (Reference Block, Norkus, Hudes, Mandel and Helzlsouer34) (r 0·59). These results are supported by interventional studies, where subjects consuming increased amounts of fruit and vegetables had higher concentrations of plasma vitamin C(Reference Broekmans, Klopping-Ketelaars, Schuurman, Verhagen, van den, Kok and van Poppel43–Reference Le Marchand, Hankin, Carter, Essling, Luffey, Franke, Wilkens, Cooney and Kolonel46, Reference Zino, Skeaff, Williams and Mann49). Fruits were found to be more associated with serum concentrations of vitamin C than vegetables, as observed in previous studies(Reference Drewnowski, Rock, Henderson, Shore, Fischler, Galan, Preziosi and Hercberg36, Reference Ness, Khaw, Bingham and Day53). Citrus fruits were more strongly associated with serum concentrations of vitamin C in the present study, whereas pome fruit, tropical fruit and fruit juices showed lower correlation coefficients. However, the regression analysis indicated that all fruits and fruit juices analysed highly contributed to antioxidant concentrations. Of the different vegetable groups, fruiting, leafy and root vegetables provided the highest contribution. Results from the Su et al. study(Reference Su and Arab39) supported the contribution of salad consumption, which is an important component of the leafy vegetable group.
In the present study, vitamin C concentrations were more strongly associated with overall fruit and vegetable intake (r 0·28) than β-carotene concentrations (r 0·20). These results are slightly different from those of Drewnowki et al. (Reference Drewnowski, Rock, Henderson, Shore, Fischler, Galan, Preziosi and Hercberg36), who reported a higher correlation between energy adjusted fruit and vegetable intake and β-carotene (r 0·36) than with vitamin C concentrations (r 0·29) in another French population. However, the current findings support the results from the Block et al. (Reference Block, Norkus, Hudes, Mandel and Helzlsouer34) study, in which vitamin C was found to be considerably more associated with fruit and vegetable intakes (r 0·59) than β-carotene (r 0·35). Intervention studies have shown a greater association between fruit and vegetable intakes and vitamin C concentrations compared with β-carotene(Reference Broekmans, Klopping-Ketelaars, Schuurman, Verhagen, van den, Kok and van Poppel43, Reference Zino, Skeaff, Williams and Mann49) or have at least shown a similar association(Reference John, Ziebland, Yudkin, Roe and Neil45). However, the fruit and vegetables category cannot be considered in its entirety, since correlations can vary greatly between these two foods. In the present study, vitamin C was consistently shown to be related to fruit intakes compared with β-carotene. On the other hand, β-carotene was found to be more associated with vegetable intakes compared with vitamin C. Therefore, the proportion of fruit in the diet when compared with vegetables might significantly affect the results.
Fruit and vegetables as well as some fruit and vegetable groups were shown to be significantly associated with serum α-tocopherol. However, these relationships were very weak and probably significant due to the large number of subjects. Although fruits and vegetables are sources of these two antioxidants, their contribution to serum concentrations is negligible: the main sources of α-tocopherol are oils, nuts, seeds and wholegrain cereals. Thus, the correlation observed between leafy vegetables and serum α-tocopherol may be mainly explained by the dressing that is often added to this type of food. These findings confirm the results of observational(Reference Block, Norkus, Hudes, Mandel and Helzlsouer34, Reference Campbell, Gross, Martini, Grandits, Slavin and Potter35)and intervention(Reference Brevik, Andersen, Karlsen, Trygg, Blomhoff and Drevon42, Reference Broekmans, Klopping-Ketelaars, Schuurman, Verhagen, van den, Kok and van Poppel43, Reference John, Ziebland, Yudkin, Roe and Neil45, Reference Le Marchand, Hankin, Carter, Essling, Luffey, Franke, Wilkens, Cooney and Kolonel46, Reference Zino, Skeaff, Williams and Mann49) studies on α-tocopherol, which included a relatively small number of subjects.
The present study included a very large sample of subjects from a free-living population and with a large diet diversity, even though subjects were recruited on a voluntary basis(Reference Hercberg, Preziosi, Briancon, Galan, Triol, Malvy, Roussel and Favier21). This is an advantage when compared with other studies realised in this field, which often included a relatively limited number of individuals. In addition, the design of the present study provides precision of both biochemical and diet measurements. The evaluation of antioxidant serum concentrations were all performed in one laboratory, in the morning, and in fasting subjects, therefore reducing the variability of measures due to methodology and diurnal variations(Reference Comstock, Burke, Hoffman, Norkus, Gross and Helzlsouer54). Assessing fruit and vegetable intake is a frequently encountered problem. Studies investigating the relationships between serum concentrations of antioxidants and diet most often use FFQ(Reference Campbell, Gross, Martini, Grandits, Slavin and Potter35, Reference Jansen, van Kappel, Ocke, Van't Veer, Boshuizen, Riboli and Bueno-de-Mesquita37, Reference Tucker, Chen, Vogel, Wilson, Schaefer and Lammi-Keefe40, Reference van Kappel, Steghens, Zeleniuch-Jacquotte, Chajes, Toniolo and Riboli41) and sometimes a single 24 h recall(Reference Su and Arab39). Al Delaimy et al. (Reference Al Delaimy, Slimani and Ferrari50) concluded that fruit and vegetable intakes assessed by 24 h recalls were generally better able to predict plasma carotenoids than FFQ on an ecological concentration. In addition, three 24 h recalls provide a particularly good estimation of the diet, when compared with other methods if subjects are well educated(Reference Resnicow, Odom, Wang, Dudley, Mitchell, Vaughan, Jackson and Baranowski55). The present study includes well-educated volunteers, with about two-thirds of subjects having achieved high school. Therefore, the use of at least six 24 h records per person provides added precision and validity.
In this supplementation trial, diet assessment followed blood sampling. Consequently, the timing between intake and serum variables is not consistent with a cause-and-effect relationship. However, in the present study we aimed to evaluate the relationship between usual diet and serum concentrations of antioxidants. In this cohort, diet has been shown to be very stable over the years (data not shown), and diet measured over 2 years is therefore representative of the usual diet. Furthermore, blood sampling has been performed over a period of 7 months, corresponding to autumn and winter seasons, and was compared with usual fruit and vegetable consumption over the year. Fruit and vegetable groups with high seasonal variation of consumption are therefore unlikely to be related to antioxidant concentrations in the blood sampling period. Fruit and vegetables mainly consumed in summer were therefore not considered. However, consumption of the other fruit and vegetable groups studied was relatively stable over the period of blood sampling (data not shown), suggesting that the correlation measured is representative of the relationship with usual fruit and vegetable consumption in the autumn/winter period. The fact that the blood sampling period does not fully cover the dietary measurement period could partly explain why observed correlations are lower than in previous studies. Finally, the single blood sample available in the present study did not allow us to take into account day to day variation in antioxidant levels.
In conclusion, serum concentrations of β-carotene and vitamin C were associated with fruit and vegetable intake in the current study population. Vitamin C serum concentrations were particularly associated with fruit intakes. These results support the findings of other studies, including a more limited number of subjects, and investigating the relationship between nutrient serum concentrations and fruit and vegetable intakes. However, correlations found in the present study may be too weak to support the use of β-carotene and vitamin C serum concentrations as biomarkers of dietary intake or the validation of a dietary assessment method but changes in serum antioxidants could potentially be used to validate self-reported changes in fruit and vegetable intake in intervention studies(Reference Le Marchand, Hankin, Carter, Essling, Luffey, Franke, Wilkens, Cooney and Kolonel46). The specific importance of root vegetables and citrus fruit for β-carotene status was emphasised in the present study, as well as the importance of citrus fruit for vitamin C. Variations in associations between specific groups of fruit or vegetables and serum concentrations of antioxidants could be taken into account when making public health recommendations.
Acknowledgements
None of the authors had any personal or financial conflict of interest. We would like to thank Stacie Yung for reviewing and editing the manuscript. The SU.VI.MAX project received funds from the Direction Générale de la Santé, the Ministère de la Santé and the Institut Virtuel de Recherche en Santé Publique (groupe cohorte) INSERM.
