Validity of a Self-administered Food Frequency Questionnaire Used in the 5-year Follow-up Survey of the JPHC Study Cohort I to Assess Carotenoids and Vitamin C Intake: Comparison with Dietary Records and Blood Level

We compared carotene and vitamin C intake assessed with our 138-item food frequency questionnaire (FFQ) against 28-day weighed dietary records among a subgroup of JPHC Study Cohort I (102 men and 113 women), and the corresponding serum carotenoid levels or plasma vitamin C levels (86 men and 100 women). Correlation coefficients between carotenoids or vitamin C intake estimated from FFQ and intakes estimated from DR were as follows in men and women, respectively: alpha-carotene, r=0.47 and r=0.46; beta-carotene, r=0.40 and r=0.30; lycopene, r=0.18 and r=0.22; vitamin C, r=0.44 and r=0.31. Correlation coefficients between carotenoids or vitamin C intake estimated from FFQ and the corresponding serum carotenoids levels or plasma vitamin C levels were as follows: alpha-carotene, r=0.38 and r=0.30; beta-carotene, r=0.28 and r=0.11; lycopene, r=0.30 and r=0.19; vitamin C, r=0.07 and r=0.06; in men and women, respectively. These data indicated carotenoid and vitamin C intakes estimated from FFQ were associated with intake from DR, although the association was weak for lycopene. Carotenoid intake estimated from FFQ were associated with corresponding serum carotenoid levels in men, but the correlation was weak in women except for alpha-carotene. Both in men and women, no association was observed for plasma vitamin C levels.

partly due to antioxidant nutrients such as carotenoids or vitamin C, which might protect biological structures from oxidative damage and reduce lipid peroxidation. 4 In Japan, until recently no comprehensive food composition database for individual carotenoids has been available, making it difficult to assess the possible relationship between specific carotenoid intake and the risk of disease.
In the present study, we estimated the alpha-carotene, betacarotene and lycopene intake from both our 138-item food frequency questionnaire (FFQ) and 28-day weighed dietary records (DR) among a subgroup of JPHC Study Cohort I using our newly-developed food composition database for individual carotenoids5. We compared the carotenoid intake and vitamin C intake estimated from both the FFQ and DR with the corresponding serum carotenoid level or plasma vitamin C level. Serum carotenoids were quantified by high-performance liquid chromatography (HPLC) using a previously-described modified method.10 Briefly, the components were extracted by shaking and mixing for 30 min after addition of each of the following: 250* L of water, 500* L of ethyl alcohol, 4 mL of n-hexane, and 400 ,* L of serum. After centrifugation, the n-hexane layer was dried with N2 gas. To determine carotenoids (alpha-carotene, beta-carotene, lycopene),the residue was dissolved in 30* L of dichloromethane and 200* L of the mixed solvent. Then, 200* L of the solution was applied to a HPLC (Waters, Milford, Mass) with the following conditions: C18 reverse phase column, Wakopak (Wako, Osaka, Japan); mobile phase, methanol: acetonitrile: water=60:40:1; UV wavelength, 450; flow rate, 1.0 ml/min.

Statistical Analysis
The subjects of this study were 215 persons (102 men and 113 women). Both their FFQ for the validation study and the complete DR (14-day records in Ishikawa PHC area and 28-day records on other 3 areas) were included in this analysis. As for the analysis using the serum carotenoid level, 186 subjects (86 men and 100 women) were measured twice (Feb. and Aug.). For their plasma ascorbic acid use, 185 subjects (86 men and 99 women) were measured (Feb.).
The mean, SD and median intakes for carotenoids (alphacarotene, beta-carotene, lycopene) and vitamin C from the FFQ and DRs were calculated by sex and area. The Spearman rank correlation was used to assess the association between carotenoid and vitamin C intake from FFQ and from DR. Serum carotenoids and plasma vitamin C were presented as mean, SD and median by sex and area. The Spearman correlation was again used, this time to examine the association between intakes (carotenoids and vitamin C assessed by FFQ or DR) and blood levels (serum carotenoids and plasma ascorbic acid). The mean intake of carotenoids and vitamin C assessed by DR and the mean biochemical indicator concentrations were calculated according to quintile of intake assessed with FFQ. All statistical analyses were performed using the SAS statistical software package.12 Table 1 shows carotenoid and vitamin C intakes by sex and area estimated from the DR. In men, alpha-carotene and betacarotene intakes were highest in the Ishikawa PHC area. Lycopene intakes were highest in the Ninohe PHC area. In women, no apparent area-difference was observed for carotenoid intake except that alpha-carotene intake was the highest in the Ishikawa PHC area. No apparent area-difference was observed for vitamin C intakes both in men and women. or women. Table 3 shows mean and median serum carotenoid and plasma vitamin C levels by sex and area. In men, the serum beta-carotene level was highest in the Saku PHC area. In women, the serum alpha-carotene level was highest in the Ishikawa PHC area and the serum beta-carotene level was highest in the Saku PHC area.

RESULTS
Both in men and women, the plasma vitamin C level was highest in the Ninohe PHC area. Table 4 shows intake levels of carotenoids and vitamin C esti-mated from DR and FFQ in 4 areas. Both in men and women, total carotene, beta-carotene, lycopene and vitamin C intakes estimated from FFQ were higher than intake from DR except for lycopene in women. Table 4 also presents the relation between carotenoid and vitamin C intake estimated from FFQ, and carotenoids and vitamin C intake estimated from DR. Spearman rank correlation coefficients between carotenoid or vitamin C intake estimated from FFQ and intakes estimated from DR were as follows: alpha-carotene, r=0.47 and r=0.46; beta-carotene,   r=0.40 and r=0.30; lycopene, r=0 .18 and r=0.22; vitamin C, r=0.44 and r=0.31; in men and women , respectively. No improvement in the correlation coefficient was observed when intakes were expressed as energy-adjusted value. Table 5 shows the mean carotenoid and vitamin C intake estimated from DR within quintiles of corresponding carotenoid and vitamin C intake estimated from FFQ. In men, the mean intake in the highest quintile was 1.5 times higher or more than in the lowest quintile for any carotenoid and vitamin C. In women, only alpha-carotene was more than 1.5 times higher. A steady increase in mean intake from the lowest to the highest quintile was observed for alpha-carotene in men and beta-carotene in women. The result of cross-classification of the subjects by quintiles from carotenoids and vitamin C intake from DR and carotenoids and   Table 6. Both in men and women, more than 60% of them were classified in the adjacent quintiles and less than 6% of them were classified in the extreme quintiles. Table 7 shows the mean, SD and median serum carotenoid levels and plasma vitamin C level in 4 areas. All serum carotenoid and plasma vitamin C levels were higher in women than in men. 3 For vitamin C, women (n=99).   2 Data on subjects in Ishikawa PHC (14-day data) were counted twice for 28-day data. r=0. 19. As for the plasma vitamin C, no significant correlation was observed either in men (r=-0.07, r=-0.01 for FFQ and DR, respectively) or women (r=0.06, r=0.07 for FFQ and DR, respectively). No improvement in the correlation coefficient was observed when intakes were expressed as an energy-adjusted value. Table 8 shows the mean serum carotenoid and plasma vitamin C levels within quintiles of corresponding carotenoid and vitamin C intake estimated from FFQ. The ratio of the mean intake in the highest to the lowest quintile in men was higher than in women . The result of cross-classification of the subjects by quintiles from serum carotenoid level or plasma vitamin C level and carotenoids and vitamin C intake from FFQ are shown in Table 9. Both in men and women, more than 50% of the subjects were classified in the same or adjacent quintiles, and less than 10% of the subjects were classified in the extreme quintiles. Table 10 and Table 11 show the cumulative % contribution of the top 20 foods for alpha-carotene and beta-carotene assessed by DR. Carrots and tomato contributed more than 92% of the total alpha-carotene intake in both sexes. Carrots were the greatest contributor, and many kinds of leafy green vegetables were important contributors of beta-carotene.

DISCUSSION
In the present study, we examined the correlation between dietary alpha-carotene, beta-carotene, lycopene and vitamin C intakes estimated from FFQ and DR. We also examined the possible correlation between the carotenoid and vitamin C intake and blood levels of carotenoid and vitamin C. Both in men and women, dietary alpha-carotene, beta-carotene, and vitamin C intakes estimated from FFQ were significantly correlated with intake from DR. However, the association between dietary lycopene intakes estimated from FFQ and DR was weak. Dietary carotenoid intake was associated with serum carotenoid level in men. However, there was no association in women except for alpha-carotene. Both in men and women, no significant correlation was observed for the plasma ascorbic acid level.
The correlation between dietary alpha-or beta-carotene intake estimated from the FFQ and DR was moderate. The alphacarotene correlation coefficient was within the range of reported correlations (0.34-0.49) in earlier studies as was the beta-carotene correlation coefficient (0.20-0.50).13.14 Because carotenoids are not energy-provided nutrition, the correlation between carotenoid intakes estimated from FFQ and both intake from DR and serum carotenoid level was not improved when expressed as an energyadjusted value.
The correlations between carotenoid intake estimated from FFQ and the corresponding serum carotenoid levels were higher in men than in women. Also, in men, correlations between carotenoid intake estimated from FFQ and corresponding serum carotenoid levels were higher than between intakes from DR and the corresponding serum carotenoid levels. In many previous studies, the correlations between dietary alpha-carotene intakes and the corresponding serum or plasma levels ranged from 0.25 to 0.53, while dietary beta-carotene intakes and the corresponding serum or plasma levels ranged from 0.15 to 0.36.1-19 Our observed correlation for men was within the range of these reported correlations. These results suggest that the data on carotenoid intake assessed with FFQ used in the JPHC study might be valid for men. However, it seemed reasonable to think that the bioavailability of carotenoids was different in men and women.
A low correlation was observed between dietary lycopene intake estimated from FFQ and those estimated from DR. This might be due to the inadequate portion size of tomatoes which contribute to more than 50% for the cumulative percentage of lycopene intake. In addition, because the lycopene database was quotation from the Department of Agriculture (USDA) food composition tables,20 it may be difficult to estimate lycopene intake in Japanese.
No meaningful correlation was observed between Vitamin C intake estimated from both DR and FFQ and plasma vitamin C levels. It is considered that vitamin C intake may vary in response to the availability of seasonal foods, and also the plasma vitamin C levels do not reflect habitual intake because of its short halflife. In addition, the estimated vitamin C intake in this study did not take account of cooking loss, and this is another reason for no meaningful correlation.
In summary, we observed a moderate correlation between dietary carotenoids and vitamin C intake estimated from the FFQ and DR. A moderate correlation was also found between dietary carotenoid intake and the corresponding serum carotenoid level for men; the correlation, however, was weak for women except for alpha-carotene. No significant correlation was observed with the plasma ascorbic acid level. These results suggest that carotenoid intake estimated from the FFQ is more valid in men than in women.