The most recent Dietary Guidelines for Americans recommend grains, vegetables and fruits for their nutrient value, including the fat-soluble vitamins (USDA and HHS, 1995). With the current emphasis on obtaining optimal nutriture from foods, it is important to identify food intake patterns that may predispose individuals to inadequate intakes of these vitamins. Prior to undertaking such epidemiologic studies which often rely on a single dietary measure, an understanding of the relationship between dietary intake and the corresponding biochemical marker of nutritional status is required for each of the fat-soluble vitamins.

Facilitated by the advent of HPLC, reliable methods have been developed for determination of each of the fat-soluble vitamins and their metabolites which can be applied to the assessment of nutritional status (Sadowski et al. 1989). Whereas there have been numerous studies of the relationship between dietary intakes and plasma nutrient levels of retinol, carotenoids, -tocopherol (Tangney et al. 1987, Yong et al. 1994), and to a lesser extent 25-hydroxyvitamin D (Jacques et al. 1993), the phylloquinone food composition database has only recently been developed for dietary assessment of this nutrient (Booth et al. 1993 and 1995a).

The magnitude of the correlation between dietary intake of a specific nutrient and its corresponding biochemical variable is in part determined by the reproducibility of each of the measures. It has been reported that the mean of multiple plasma measurements is required to reduce the large within subject to between subject variance ratios of the plasma concentrations of several fat-soluble vitamins, notably retinol, -tocopherol (Tangney et al. 1987), and certain carotenoids (Yong et al. 1994). Similarly, large within subject variation in dietary intakes has been reported for retinol (Bingham et al. 1994), multiple carotenoids (Yong et al. 1994) and -tocopherol (Tangney et al. 1987). We recently reported a within-to-between subject variance ratio of 2.6:1 in phylloquinone dietary intake among postmenopausal women, as estimated from three consecutive days of diet records (Booth et al. 1995b). To date, the within-to-between subject variance ratio for fasting plasma phylloquinone concentrations has not been reported.

In this study, we compare the diet-plasma relationships for the fat-soluble vitamins in 34 healthy adults. In addition, within and between subject variance ratios are presented for both dietary intakes and fasting plasma concentrations for each of the fat-soluble vitamins.

METHODS

Of the original study group, two women were excluded from this study which assessed the relationship between dietary intakes and corresponding plasma concentrations of the fat-soluble vitamins. One older woman had fasting plasma -tocopherol concentrations of 59-88 nmol/L, which were suggestive of vitamin E supplement use, and one younger woman did not complete one set of diet records. The characteristics of the remaining 34 subjects who were included in this study are summarized in Table 1.

Table 1. Characteristics of the study group1 [View Table]

Diet records were coded using the Minnesota Nutrition Data System (NDS) software, developed by the Nutrition Coordination Center, University of Minnesota, Minneapolis, MN (Food Database, version 6A; Nutrient Database, version 21, 1993), with the phylloquinone food values linked to the corresponding codes in the NDS. The development of the phylloquinone database has been previously described (Booth et al. 1995b).

The percentage contributions of individual food items to the total intake of each of the fat-soluble vitamins were calculated as described in detail elsewhere (Block et al. 1985). Vitamin A was defined as µg of retinol equivalents per day, vitamin D was reported in µg per day, vitamin E was defined as mg of -tocopherol equivalents per day, and vitamin K was defined as µg of phylloquinone per day.

RESULTS

The arithmetic mean intakes for retinol equivalents and phylloquinone were greater than the current Recommended Dietary Allowances (RDA), whereas vitamin D and -tocopherol equivalents were within or below the RDA for these age groups (Food and Nutrition Board, 1989) (Table 2). However, subjects were instructed to stop all supplement use for at least one month prior to the completion of the first set of diet records so the dietary intakes reported here were derived exclusively from food sources. Thirteen subjects reported regular use of vitamin and mineral supplements before enrolling in the study.

Table 2. Fasting plasma concentrations and dietary intakes of the fat-soluble vitamins in humans1,2 [View Table]

The older subjects had significantly higher intakes of retinol equivalents (P = 0.008) and vitamin D (P = 0.01) than the younger subjects (Table 2). The men as a group consumed significantly more vitamin D (P = 0.01) and -tocopherol equivalents (P < 0.001) when compared to the women. There were no significant differences in phylloquinone intakes when compared by age or sex, and no significant interactions between the independent variables (age, sex and visit) for any of the dietary intakes.

Only two other effects achieved statistical significance. Plasma 25(OH)D concentrations were 20% higher in the first visit than in the last two visits for all age and sex groups (P < 0.001 for visit). Older subjects had significantly higher plasma -tocopherol concentrations at each visit. A small but significant (P = 0.04) visit by age interaction for plasma -tocopherol was seen, where the relationship with age was greater for visit 3 compared to visits 1 and 2. 

Table 3. Correlation coefficients between fasting plasma fat-soluble vitamin concentrations and the corresponding dietary intakes as estimated from diet records1 [View Table]

An increase in both the number of diet records collected and the number of blood draws improved the diet-plasma relationship for vitamins A, D, and K, with only the relationship for phylloquinone being significant (Table 3). The phylloquinone diet-plasma relationship was not evident in the first visit whereas it was in the second and third visits (Figure 1). The highest between subject correlation measured between dietary intake of phylloquinone and plasma phylloquinone concentrations was for the average of the last two sets of diet records and corresponding blood draws (r = 0.51, P = 0.004). Collectively these data suggest a training effect on the adequacy of completing the diet records.

Table 4. Comparison of within-to-between subject variance ratios for fat-soluble vitamins in diet and plasma [View Table]

Table 5. Primary food sources of each of the fat-soluble vitamins [View Table]

DISCUSSION

In a previous cross-sectional study of 358 postmenopausal women, we found a significant correlation (r = 0.13, P = 0.01) between total dietary intake of phylloquinone as estimated from 3-d diet records and a single plasma phylloquinone concentration (Booth et al. 1995b). The current study confirmed that the correlation between dietary phylloquinone intake and corresponding plasma concentrations is improved with multiple plasma and diet measurements. Given the potential influences of triglyceride concentrations (Sadowski et al. 1989) and apolipoprotein E genotypes (Kohlmeier et al. 1995) which were not controlled for in this study, the diet-plasma relationship for phylloquinone was surprisingly strong.

Non-dietary factors affecting nutrient concentrations can also reduce the magnitude of the diet-plasma correlation (Jacques et al. 1995). In this study and others (McKenna 1992, Olmedilla et al. 1994), both dietary intakes and biochemical measurements varied with age and sex, but these were not uniform among the fat-soluble vitamins. That older adults had significantly higher plasma phylloquinone concentrations is consistent with other studies (Ferland et al. 1993, Sadowski et al. 1989), although Sokoll and Sadowski (1996) only noted an age effect among women. The dietary intake data did not explain the higher plasma phylloquinone concentrations among the older group in our study.

Whereas plasma retinol, -tocopherol and 25(OH)vitamin D concentrations had within subject variances that were less than between subject variances, phylloquinone had a within-to-between subject variance ratio of 5.4:1. Kohlmeier et al. (1995) reported a within-to-between subject variance ratio of 0.17:1 for fasting plasma phylloquinone concentrations among hemodialysis patients. However, differences in the two study populations and study designs preclude comparison of these variance ratios. The within-to-between subject variance ratio for dietary phylloquinone intakes was also the highest of all the fat-soluble vitamins. Collectively these data have important implications for population studies which often rely on a single 24-h recall and/or 2-d diet record, or a single plasma or serum sample.

Of the food sources contributing to the dietary intakes of the individual fat-soluble vitamins, fortified breakfast cereals were the principal food items for retinol equivalents, vitamin D, and -tocopherol equivalents. Breakfast cereals have already been identified as important food sources of retinol equivalents (Block et al. 1985) and -tocopherol equivalents (Block et al. 1994, Murphy et al. 1990), but it was surprising that they contributed more than a fifth of the dietary intakes in this study group. The important contribution of fortified cereals to the dietary intakes of retinol equivalents, -tocopherol equivalents, and vitamin D emphasizes the need for accurate food composition data that is current with products available on the market at the time of the collection of both dietary data and blood samples. Although nearly 50% of the dietary intake of vitamin D came from milk, this estimate may not be representative of the general population as one of the criteria for study eligibility was a daily tolerance of 250 g of milk. Margarine was another common dietary source of all the fat-soluble vitamins. Phylloquinone was primarily obtained from green vegetables and vegetable oils, which is a consistent finding with national food consumption data (Booth et al. 1996). To date, foods have not been fortified with phylloquinone. Menaquinones, another potential form of dietary vitamin K, do not have available food composition data for estimating usual intakes. Menaquinones are present in fermented foods and are not assumed to have an important role in the American diet (Suttie 1995). However this is an area of vitamin K nutriture that remains to be studied.

In summary, both dietary intakes and biochemical variables were affected by age and sex, although these effects were not uniform among the fat-soluble vitamins A, D, E and K. If not controlled for, such nondietary influences can bias the magnitude of the diet-plasma correlation. In this study group, phylloquinone was the single fat-soluble vitamin that had a significant correlation between dietary intakes and fasting plasma concentrations when adjusted for age and sex. The magnitude of this correlation improved with an increase in both the number of plasma measures and the number of diet records, although a training effect was associated with the first visit. Of the fat-soluble vitamins measured in this study, phylloquinone had the greatest within subject variance in both dietary intakes and fasting plasma concentrations. Collectively these data have important implications for epidemiologic studies which usually rely on a single dietary or plasma measure.