Development and validation of specific carotene food composition tables for use in nutritional epidemiologic studies for Japanese populations.

Assessment of dietary intake is important to understand the relationship between nutrition and health. Although the role of specific carotenoids has recently been of great interest, there are no comprehensive food composition tables for intake of specific carotenoids in Japan. We have therefore developed a new carotene food composition table that shows the alpha- and beta-carotene values based on an extensive review of the literature (FCT1). Using a 14- or 28-day diet record data of sample population (n=188), we selected 12 important foods to two carotene intakes. We analyzed the carotene contents of the foods, and developed the another composition table in which the food contents were replaced by the analytical values (FCT2). Carotene intakes of the population were significantly different between these two composition tables. However, the correlations between the dietary intake and the serum concentrations were almost identical, i.e., partial correlations using FCT1/FCT2 were 0.32/0.30 and 0.33/0.36 for alpha-carotene and 0.28/0.28 and 0.30/0.29 for beta-carotene in men and women. The similar correlations with the serum concentrations may indicate an comparable value for ranking individuals between the two tables. However, the results were inconclusive for the estimation of absolute intakes.


INTRODUCTION
Assessment of dietary intake is important to understand of the relationship between nutrition and health by epidemiological investigation. Dietary assessment depends on the quality of the food composition tables used for nutrient calculation.
Japanese food composition tables, however, do not always provide sufficient information such as sampling methods, detailed analytic methods and explanations of missing values.
In Western countries, although findings have not been consistent, several recent epidemiologic studies have reported the relationship between specific carotenoids in foods and the incidence of cancer at several sites 1-6), cardiovascular disease 7), and age-related macular degeneration 8). In Japan, however, there is no report to date of a comprehensive food composition table of specific carotenoids.
In this article, we describe the development of a new carotene food composition table that indicates the values for alpha-and beta-carotene. Firstly, we developed a carotene food composition table using published data on carotene food composition. Secondly, we sampled and analyzed the compositions of foods with high contribution to carotene intakes. Thirdly, the analyzed values were replaced by the literature-based values. Lastly, they were applied to a sample population, and validated by diet-serum relationships.
We also collected food composition tables which was used in the published epidemiologic studies on the association between carotene intake and diseases conducted in Western countries.
We obtained 6 articles published in Japan 9-14) , 3 in Finland 15 17) and where C is the estimated specific carotene content of a food, ST is the total carotene content in the standard food composition table of the food, R; is the specific carotene content of the food obtained from the literature, i is alpha-carotene or betacarotene, and Rt is the total carotene (sum of alpha-and betacarotene) content obtained from the literature.
When specific carotene data were unavailable for a particular food, we used the composition ratio of alpha-and betacarotene of the botanically similar foods with known specific carotene content (Method B). In reports published in Japan, most of the data were analyzed using raw foods. Thus, the carotene content ratios of raw foods were adapted to that of cooked foods. We estimated specific carotene compositions for foods with carotene contents >0 ug/100 g portion. Foods categorized as cooked and prepared foods, the so-called "food group 18" in the standard food composition table, were not included in this study. Among confectioneries, contained at least some carotene, mostly from milk, egg, butter, and color additives. Based on our literature search, these ingredients contain only beta-carotene. Total carotene was thought to be betacarotene in this food group (Method C). Although green leafy vegetables were considered carotene-rich foods, we could obtain data only for a part of these foods. Most green leafy vegetables do not contain alpha-carotene according to the published reports. Total carotene was thought to be beta-carotene in green leafy vegetables (Method C). Algae are also carotenerich foods, but reliable information on the alpha-and betacarotene composition was unavailable. "Nuts and seeds" and most "seasonings and spices" also contain carotene, but we could not obtain their specific carotene composition.

Selection and laboratory analysis of important foods
Carotene compositions were analyzed for the important foods selected by the following criteria. First, specific carotene intake levels of the four sample populations (see "sample population and data collection" in detail) were calculated with the developed carotene composition table. Next, we selected foods in which alpha-or beta-carotene contributed more than 3% of the overall intake in each area. The food samples were purchased between June and July in 1999. Three samples were obtained for each item from 3 different stores. Most samples were from local markets, and the cheapest one was selected.
Only the edible part of the foods was used for laboratory analysis. Fifty grams of each sample, 50 mL of water, and 5 mL of pyrogallol were mixed together. Ten grams of samples were homogenized and extracted in an Ultra Turrax by 150 mL of n-hexane / acetone (2:1), then filtered and the food residue extracted several times with n-hexane / acetone until the yellow became colorless. The combined carotene extract was mixed with 100 mL of water for 30 min at room temperature. After phase separation, the lower phase was removed and shaken again with hexane. The recombined n-hexane phases were dried by water-free Na2SO4 and reduced in the rotation evaporator at 30 C . Then 25,u L of the solution was applied to a high-performance liquid chromatography (HPLC) (Waters, Milford, Mass). The HPLC conditions were as follows: 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. The mean value was calculated with the data from triplicate analysis.

Development of table with analytical (FCT2)
We computed alpha-and beta-carotene composition using the formula (1) for the foods with analytical values. But R; is the analytical specific carotene content of the food and Rt is the sum of analytical alpha-and beta-carotene content of the food (Method D). Botanically similar foods were also replaced by analytical values (Method E). The composition for these foods in FCT 1 were replaced by the new compositions. We referred to this composition table as FCT2.

Sample population and data collection Subjects
The sample population was a sub-population (n=221, aged 44-63 years) from the Japanese Public Health Center-based Prospective Study (JPHC Study) 20) living in Iwate, Akita, Nagano, and Okinawa prefectures who kept 28-or 14-day dietary records (DR) in 1994 or 1995. The subjects were orally informed the purpose and procedure of this study and agreed to the participation by providing diet record and blood sample.

Diet record
Semi-weighted dietary records from four different seasons (two seasons, i.e., winter and summer only in Okinawa) over seven consecutive days were collected by a method used in the National Nutrition Survey 21) with some modifications 22, 23). Research dieticians instructed the subjects to record all foods and beverages prepared and consumed in a specially designed booklet. The participants were asked to provide detailed descriptions of each food, including the methods of preparation and recipes whenever possible. The dieticians checked the records at each participant's home during the survey and reviewed them in a standardized way after recording.

Blood samples
A total of 25 mL serum was collected by venipuncture from all subjects just before the DR in winter and just after the DR in summer. Fasting for at least 5 hours was requested before blood collection. The samples were stored at -80 C until analysis. Serum carotenes were determined by HPLC. Total cholesterol in serum was analyzed enzymatically with an autoanalyzer.

Anthropometric data
The body mass index (BMI) was computed in all subjects as self-reported weight (kg) divided by height (m) squared.

Validation of two composition tables
We calculated specific carotene levels of the sample population with the two developed carotene composition tables. The first step was to compare the distributions of each of the aforementioned composition tables. We evaluated the agreement between the two different composition table estimates by comparing mean and median intakes. Because the distributions of the paired differences were not normally distributed, a nonparametric test, the Wilcoxon signed-rank test, was used to determine whether or not the median difference between FCT1 and FCT2 carotene estimates were significantly different from zero. The second step was to validate the two composition tables by comparing the diet-serum relationship for the alphaand beta-carotene using Spearman rank correlation coefficients. Multiple regression analysis was conducted on the serum carotene concentrations and the estimated intakes including possible confounding factors such as BMI, serum cholesterol concentration, alcohol intake, number of cigarettes smoked per day, and age [24][25][26][27]. Serum total-and beta-carotene showed significant correlations with BMI and alcohol intake. Alpha-carotene showed significant correlations with BMI, alcohol intake, and serum total cholesterol. Therefore the partial correlation coefficients adjusting for these significant variables were computed in each model and presented here. SAS software, release 6.12 (SAS Institute, Cary NC) was used for all statistical computations.
Three subjects (1 man and 2 women) who took carotene containing supplements were excluded from the analysis. Serum carotene reflects the relatively long-term intake such as several weeks 28,29). In addition, there are large day-to-day variation in dietary carotenoid intake 30-32 ). Thus, we used serum measurement in summer and 14-day DR data combining two 7-day DR data obtained in winter and summer respectively. Table 1 shows the number of foods by substitution methods. Among 553 foods in which carotene is >0 /*g/100 g portion, substitution methods were found for 341 foods. A suitable substitution method was not found for the remaining 212 foods. The 81.5%, 11.1%, and 7.3% of foods were substituted by Methods A, B and C, respectively. The food compositions of alpha-and beta-carotenes for the 341 foods are listed in Appendix A. Appendix B shows there was only beta-carotene. Table 2 presents the contribution rate of food for intake of total-, alpha-, and beta-carotene in the sample population. Ninjin (Daucus carota) was the most important food both for alpha-and beta-carotene intakes in all areas (about 90% for alpha-carotene, and 32 to 50% for beta-carotene). No apparent area-difference was observed for alpha-carotene except in the Okinawas, where Nigauri (Momordica charantia) was also important for alpha-carotene intake. In all the areas , hourensou (Spinacia oleracea) mainly contributed to beta-carotene intake. In the Nagano area, Nozawana (Brassica campestris) also largely contributed to beta-carotene intake. In the Okinawa area, mango (Mangifera indica) also acconuted for much betacarotene intake. According to the criteria, we selected 12 foods and collected 72 samples in each area. Ninjin and hourensou were collected in 4 areas. Nira (Allium tuberosum) and tomato (Lycopersicon esculentum) were collected in the Iwate, Akita , and Nagano areas. Komatsuna (Brassica campestris) was collected in the Akita and Nagano areas . Shungiku (Chrysanthemum coronarium) was gathered only in the Akita area. Seiyou-Kabocha (Cucurbita mexima) was obtained from the Iwate and Nagano areas. Nozawana and Ingenmame (Phaseolus vulgaris) were collected only in the Nagano area . Karashina (Brassica juncea), Nigauri, and mango were gathered only in the Okinawa area. Analytical contents of alpha-and beta-carotene were shown in Table 3. Although slight differences were found for the contents between areas, there were no apparent differences in the content ratio of alpha-and beta-carotene between samples.

Selection and analysis of important foods
The number of foods in the revised composition table by the substitution method was shown in Table 1 (in parentheses). Carotene contents were revised with analytical values for 32 foods (Methods D and E).

Validation of two composition tables
Mean, standard deviation, and median estimates of daily carotene intakes based on the developed composition tables in the sample populations by sex are shown in Table 4. The estimated alpha-carotene intakes were higher in FCT1 than in FCT2. Conversely, the estimated beta-carotene intakes were lower in FCTI than in FCT2. All the median values are significantly different between FCT1 and FCT2 by Wilcoxon signedrank test. We also calculated and examined them by area, but no apparent area-difference was observed (data not shown). Table 5 shows Spearman correlation coefficients for specific carotenes between the intakes and the corresponding serum concentrations. In both sexes, carotene intakes moderately cor-related with the serum concentrations. Only small differences were found between the results based on FCT1 and FCT2.   Table 3. Analytical contents of alpha-and beta-carotene (*g/100 g food portion) .
a Not detected .
Composition ratio of alpha-or beta-carotene to total carotene .
See Table 2 for detailed information.
Carotene Composition   a See Table 4 for the detailed definitions . b Same as FCT1 (see text for details) . ` Partial correlation coefficients adjusted for body mass index (BMI) and alcohol intake for total-and beta-carotene, and BMI, alcohol intake, and serum total cholesterol for alphacarotene.
could not sufficiently consider the data quality of the measuring and sampling methods used in them. Mangels  Total carotene, alpha-and beta-carotene were mainly derived from Ninjin. Total-carotene and beta-carotene were derived from Hourensou and Nozawana, and alpha-carotene was derived from Tomato. Ninjin and Hourensou were also maior sources of total carotene intake in two previous studies in Japan 44,45) Both analytical and reference values varied widely between sampled foods because many factors such as varietal differences, variable growth and storage, different geographic locations, and seasons, affect the carotene composition in foods 9, 36). This may seriously distort the results, especially in the development of a composition table using reference values rather than analytical values obtained from the foods actually consumed by a target population. Despite this assumption, the two composition tables similarly ranked the sample subjects according to each carotene intake when the serum carotene concentration was used as the gold standard ( In conclusion, we developed a carotene composition table with food compositions obtained from the literature and analytical values for use in nutritional epidemiologic studies for Japanese populations. FCT1 is readily available for non-profit use only on the authors' website (http://www.east.ncc.go jp/epi). They request, however that this article be cited when a study in which the data, even in part, have been published or made available to the public. The composition table may be useful for studies on the association between alpha-and beta-carotene intakes and health status in Japanese populations. Its reliability for estimation of estimate absolute intakes, however, is not yet confirmed.

Appendix
A. Developed (substituted) alpha-and beta-carotene food composition table.
Food code used in the standard tables of food composition in Japan, the fourth revised edition 19) n See text for methods of substitution .