PDF
Citation
Print
INTRODUCTION
Appropriate vitamin D status improves bone health and chronic diseases [1234]; moreover, it has shown benefits during the coronavirus disease 2019 pandemic [56]. Therefore, assessing vitamin D status is crucial, especially in Korea, which has a population with vulnerable vitamin D nutritional status [7]. In Korea, despite the potential for vitamin D synthesis through ultraviolet-B (UV-B) exposure from sunlight, factors like ozone depletion and air pollution have curtailed outdoor activities and increased the use of UV-blocking tools, reducing adequate sunlight exposure [89]. Additionally, the UV-B radiation in winter is substantially lower than in summer, resulting in decreased vitamin D levels, highlighting the need for dietary vitamin D intake during this period [10].
Research on vitamin D intake among Koreans began recently after establishing a vitamin D content database for common Korean foods and re-analyzing surveys like the Korea National Health and Nutrition Examination Survey (KNHANES) [111213]. Previous studies within the KNHANES used the 24-hour recall (24RC) method. The KNHANES employs a food frequency questionnaire (FFQ) and assesses all nutrients; however, it is less suitable for specific nutrients like vitamin D [14]. Compared with other methods, the 24RC method and dietary records (DRs) provide more accurate evaluations and are considered the gold standard in FFQ development; however, they are burdensome for year-round monitoring in large populations [141516]. Many of the primary sources of vitamin D in the Korean diet are seasonal, necessitating time-consuming and costly annual surveys. Developing an FFQ to simplify the survey of vitamin D intake in Koreans could facilitate research on vitamin D nutritional status and lead to the development of necessary nutritional policies.
To the best of our knowledge, no tools have been developed to assess vitamin D intake exclusively among Koreans. The British Columbia Dairy Association created a calcium calculator that was the basis for the Canadian Calcium Assessment Tool [17]. Park et al. [18] adapted this tool to create the Korean Calcium Assessment Tool (KCAT), which includes 45 types of foods that are major sources of calcium and vitamin D according to the KNHANES data. Participants completed the KCAT and the KNHANES FFQ. While calcium intake assessments were consistent, vitamin D intake was overestimated. Jin et al. [19] updated the KCAT to reflect the major food sources of calcium and vitamin D, as shown in the KNHANES VI and VII (2013–2018). The updated KCAT showed consistent calcium assessments; however, vitamin D intake remained escalated. Because the KCAT included various major Korean vitamin D sources, such as fish, mushrooms, and fortified milk, the assessed vitamin D intake was about twice as high as that using the KNHANES FFQ, failing to validate its appropriateness. Therefore, a dedicated vitamin D assessment tool remains to be developed and tested against the 24RC reference method.
The FFQs for assessing vitamin D intake have been undertaken in various countries, including Ireland [20], Serbia [21], Libya [22], Qatar [23], England [24], Croatia [25], the United States [2627], Finland [16], and Poland [28]. Vitamin D FFQs have been developed in various countries, reflecting local dietary patterns and policies. In Europe, vitamin D is primarily sourced from oily fish and fortified foods, with some countries mandating fortification in products like margarine and dairy [82529]. FFQs from Poland and England illustrate regional differences, including various food groups such as animal products, cereals, and fats [2428]. Unlike these countries, in Korea, the primary sources of vitamin D contributing to 99% of its intake are fish, eggs, dairy and milk products, meat, mushrooms, and cereals [12]. Cheese and butter consumption are not as prevalent in Korea; however, various fish and mushrooms are significant vitamin D sources. Moreover, Korea implements a voluntary fortification policy for vitamin D in foods, making fortified vitamin D products almost exclusively available in some dairy products, soymilk, and cereals. Owing to these distinct dietary patterns and policies, a vitamin D FFQ tailored for Koreans is necessary.
This study aimed to develop a semi-quantitative FFQ for assessing vitamin D intake, considering the dietary habits of Koreans, and initially focused on adult women to evaluate its appropriateness. The FFQ for vitamin D intake assessment was conducted in the same sample group twice, approximately 9–11 months apart, to evaluate its reproducibility. Furthermore, its validity was tested by comparing the results with those obtained from 12 days of DRs (12-day DRs) across the 4 seasons.
SUBJECTS AND METHODS
Development of an FFQ for assessing vitamin D intake
To develop an FFQ for evaluating vitamin D intake, this study reviewed vitamin D source foods consumed by adults aged 19 years and older from the 7th KNHANES (2016–2018) using 24RC results. The vitamin D content database was updated using previous research [12]. The FFQ food items were selected based on frequency and total intake, covering up to 90.99% of the cumulative intake (Supplementary Table 1). ‘Bread’ was excluded to prevent confusion owing to its typical inclusion of other vitamin D sources, such as meat or eggs. Fish roe and imitation crab meat were removed because of their minimal contribution (0.2%) to vitamin D intake. This resulted in identifying 25 frequently consumed foods, with eel, salmon, cereal, Spanish mackerel, quail eggs, and beef added, creating an FFQ with 30 items. Serving sizes for mushrooms, milk products, mayonnaise, and cereals were based on the 2020 Dietary Reference Intakes for Koreans (KDRIs) [30]. For fish, eggs, meat, and soymilk, serving sizes were set closer to those in the 2016 KNHANES FFQ [31]. For clarity, pictures of each serving size were shown. Food intake frequency was divided into 9 categories: ‘almost never,’ ‘once a day,’ ‘2 times a day,’ ‘3 times a day,’ ‘1–2 times a week,’ ‘3–4 times a week,’ ‘5–6 times a week,’ ‘once a month,’ and ‘2–3 times a month.’ The amount consumed per occasion was divided into 50%, 100%, and 200% of one serving.
In June 2020, a pilot study was conducted using the newly developed FFQ and a 3-day DR on 11 women in their twenties. Participants recorded their food and beverage intake over 3 non-consecutive days (2 weekdays and 1 weekend day). The comparison aimed to assess whether the items and portion sizes included in the FFQ accurately captured the participants' diet. The review of the 3-day DR and the FFQ indicated that soymilk, which was initially excluded, significantly contributed to vitamin D intake. Finally, the new FFQ was completed with 31 food items across 8 food groups (Table 1).
Table 1
List of food items included in the vitamin D FFQ
FFQ, food frequency questionnaire.
1)One serving size was defined from the 2020 Dietary Reference Intakes for Koreans for mushrooms, milk and its products, mayonnaise, and cereals and from the 2016 Korea National Health and Nutrition Examination Survey FFQ for fish, eggs, meat and its products, and soybean milk.
2)The vitamin D content (µg) per serving, utilized by the vitamin D database, was modified and supplemented in this laboratory based on previous research.
Study participants
Women in Daejeon were recruited through online social networks in 2 stages: the first included women aged between 19 and 29 years, and the second included women aged between 45 and 69 years. Participants met the inclusion criteria if they were able to participate throughout the year, had a body mass index (BMI) between 18 and 30 kg/m2, had no chronic diseases, were not taking continuous medications, were not pregnant or breastfeeding, and consumed at least 2 regular meals per day. The first recruitment in September 2020 involved 135 participants, and the second in September 2021 involved 148 participants, totaling 283. After excluding 32 participants who did not meet the criteria and 84 dropouts, 152 remained (67 from the first and 85 from the second recruitment). To ensure study reliability, 15 participants were excluded due to a more than 5-fold difference in vitamin D intake between FFQ1, FFQ2, and the 12-day DR records. The study was conducted from September 2020 to August 2021 for the first cohort and from September 2021 to August 2022 for the second cohort. This study was approved by the Chungnam National University Institutional Review Board (IRB) (approval numbers: 202005-BR-047-01, 202107-SB-130-01). Informed consent was confirmed by the IRB. The study was conducted in accordance with the Declaration of Helsinki and its later amendments.
Food intake survey
Participants recorded their dietary intake over 3 days in each season (Spring: March–May; Summer: June–August; Autumn: September–November; Winter: December–February), including 2 weekdays and 1 weekend day, totaling 12 days (12-day DR). The participants also completed the FFQ at the start of the research in autumn, the first season (FFQ1), and at the end in summer, the final season (FFQ2), in order to assess its reproducibility after a long interval. At the beginning of the study, participants were provided with an instructional video on how to complete the DRs and FFQ for self-recording. For the DRs, participants detailed the menu, ingredients, and amounts consumed and took photos of their meals before and after eating using a provided ruler. Experienced researchers collected all DRs and diet photos, and the nutrients consumed were calculated using a computer-aided nutritional analysis program (CAN pro, web version 5.0) [32] from the Korean Society of Nutrition. A vitamin D database was developed in a laboratory [12]. For the FFQ, this database was used to calculate food intake by multiplying the frequency of intake by the amount consumed.
Reproducibility and validity of the FFQ
The newly developed FFQ was validated for reproducibility and validity using Willett’s method [14]. Reproducibility was verified by comparing the results of FFQ1 and FFQ2, administered approximately 9–11 months apart. Validity was assessed by comparing FFQ2 to the 12-day DR, as the FFQ2 included the period covered by the 12-day DR.
Statistical analyses
The age and anthropometrics of the study participants are presented as means ± SDs. Means, SDs, medians, and minimum and maximum values for the 12-day DR, FFQ1, and FFQ2 are provided, along with the percentage of participants meeting adequate intake levels in each survey. Differences among the 12-day DR, FFQ1, and FFQ2 groups were analyzed using the analysis of variance. The adequate vitamin D intake levels, according to the 2020 KDRIs, were used as the reference: 10 µg/day for those aged between 19 and 64 years and 15 µg/day for those aged 65 years and older [30].
The FFQ was validated using reproducibility assessment (FFQ1 vs. FFQ2) and validity assessment (FFQ2 vs. 12-day DR).
1) Because the vitamin D intake data from the FFQ1, FFQ2, and 12-day DRs did not follow a normal distribution, a non-parametric Spearman’s rank correlation analysis was performed. The intra-class correlation coefficient (ICC) was also used to analyze the correlation and bias between the measurements, using Bland & Altman’s reliability criteria [3334]. To compare the vitamin D intake from each food group, as calculated from the FFQ2 and 12-day DRs, means and SDs are presented, and differences were analyzed using the Wilcoxon rank test. The correlation between the 2 variables was analyzed using Spearman’s rank correlation coefficients.
2) Vitamin D intake from the 2 methods was categorized into quartiles, and the weighted kappa coefficient was calculated to evaluate concordance between the methods. The weighted kappa coefficient assigns weights based on the degree of disagreement [35]. According to Landis and Koch [36], values less than 0.20 indicate slight agreement, 0.21–0.40 fair agreement, 0.41–0.60 moderate agreement, 0.61–0.80 substantial agreement, and 0.81–1.0 almost perfect agreement [36].
3) To verify the agreement between the 2 survey methods, a Bland–Altman plot analysis was conducted using the means and differences in vitamin D intake measurements. The clustering of differences around the mean difference indicated a high level of agreement. The limits of agreement (LOA) were set at ± 1.96 times the SD of the differences (95% confidence level), and the percentage of values outside this LOA was calculated as the Bland–Altman index [37]. Values up to a maximum of 5% were considered positive indicators [34]. Bland and Altman recommend a logarithmic transformation if larger values result in greater variability between measurements [3334]. In this study, a logarithmic transformation was applied owing to the spreading pattern of mean values.
A P-value less than 0.05 was considered statistically significant. Statistical analyses were performed using IBM SPSS Statistics for Windows, version 28.0 (IBM Corp., Armonk, NY, USA) and MedCalc Statistical Software version 20.216 (MedCalc Software Ltd., Ostend, Belgium; https://www.medcalc.org; 2023).
RESULTS
Participant's age and anthropometrics
The study included 152 participants with an average age of 42.0 ± 17.4 years. The average weight was 57.8 ± 9.0 kg, and the average height was 160.1 ± 5.5 cm. The average BMI was 22.5 ± 3.3 kg/m2 (Table 2).
Comparison of the vitamin D intakes assessed by the FFQs and the 12-day DR
The average daily vitamin D intake calculated from the FFQs and 12-day DR are presented in Table 3. The average daily vitamin D intake was higher in the FFQ1 at 4.90 ± 3.46 µg/day than in the FFQ2 at 4.58 ± 3.75 µg/day. The intake in the 12-day DR was the lowest at 4.07 ± 2.78 µg/day. Including vitamin D supplements, the average vitamin D intakes were 13.5 ± 17.5 µg/day for the FFQ1, 14.3 ± 19.7 µg/day for the FFQ2, and 12.6 ± 14.1 µg/day for the 12-day DR. The percentage of participants with vitamin D intake from only food sources below the adequate intake recommended by the 2020 KDRIs was 94.1% for the FFQ1, 97.4% for the FFQ2, and 96.1% for the 12-day DR. The percentage with total vitamin D intake, including supplements below the 2020 KDRIs, was 64.5% for the FFQ1, 61.8% for the FFQ2, and 61.8% for the 12-day DR.
Table 3
Comparison of the vitamin D intakes assessed by FFQs and 12-day DR
FFQ, food frequency questionnaire; DR, dietary record; AI, adequate intake.
1)P-values were derived by the analysis of variance.
2)According to the 2020 Dietary Reference Intakes for Koreans, the AI of vitamin D is 10 µg/day for the 19–64 years age group and 15 µg/day for the ≥ 65 years age group.
Reproducibility and validity of the FFQ
The FFQ reproducibility and validity assessment results are presented in Table 4. The Spearman’s rank correlation coefficient for reproducibility was 0.592 (P < 0.001), and for validity, it was 0.460 (P < 0.001). The ICC analysis showed that the reproducibility was 0.599, and the validity was 0.543, indicating moderate agreement. Analysis of quartiles for reproducibility showed that 38.82% of the participants were classified into exact agreement, 96.71% into the sum of exact and adjacent agreement, and 3.29% as the opposite. For validity, the analysis of quartiles showed that 37.50% of the participants were classified into exact agreement, 97.37% into the sum of exact and adjacent agreement, and 2.63% into opposite agreement. The weighted kappa coefficients for reproducibility and validity were 0.379 and 0.284, respectively, indicating fair agreement.
Table 4
Validation of the FFQ, including assessment of reproducibility and validity
FFQ, food frequency questionnaire; DR, dietary record; ICC, Intra-class correlation coefficient.
1)ICC of 0.4–0.6 indicates moderate agreement.
2)Weighted kappa coefficient of 0.21–0.40 indicates fair agreement.
***P < 0.001.
Reproducibility through Bland–Altman plots of the difference in vitamin D intake
Reproducibility was validated using a Bland–Altman plot (Fig. 1). The mean difference of the log-transformed values between the FFQ1 and FFQ2 was 0.07 µg, and the LOA ranged from −1.05 to 1.20. The Bland–Altman index was 3.95%.
Validity through Bland–Altman plots of the difference in vitamin D intake
Validity was assessed using a Bland–Altman plot (Fig. 2). The mean difference of the log-transformed values between the FFQ2 and 12-day DR was −0.11 µg, and the LOA ranged from −1.34 to 1.12. The Bland–Altman index was 3.95%.
Correlation between the FFQ2 and the 12-day DR in the major food groups for vitamin D
Table 5 presents the vitamin D intake values obtained from the FFQ2 and 12-day DR for the major vitamin D food groups. The food group with the highest vitamin D intake from both survey methods was the fish group (2.50 ± 3.51 µg/day and 2.27 ± 2.37 µg/day, respectively), and no significant differences were observed. Among other food groups, excluding fish, only the vitamin D intake from eggs was significantly higher in the 12-day DR results compared with the FFQ2 (0.87 ± 0.43 µg/day vs. 0.65 ± 0.53 µg/day, respectively; P < 0.001). The results from the FFQ2 were significantly lower for all other food groups than those for the 12-day DR. A significant positive correlation was observed between the 2 survey methods across all food groups. The highest Spearman’s rank correlation coefficient was observed for the fish group (r = 0.498; P < 0.001), followed by milk and dairy products (r = 0.413; P < 0.001), pulses and legumes (r = 0.335; P < 0.001), and cereals (r = 0.316; P < 0.001).
Table 5
Correlation between the FFQ2 and 12-day DR in the major food groups for vitamin D
FFQ, food frequency questionnaire; DR, dietary record.
1)P-values were derived from the Wilcoxon-signed rank test.
2)Spearman’s rank correlation.
*P < 0.05; ***P < 0.001.
DISCUSSION
Vitamin D is vital for managing immune-related disorders [38], cardiovascular diseases [39], and chronic degenerative diseases [1], highlighting its importance in nutritional status. Serum vitamin D levels are remarkably low among Koreans [7], and previous studies show that vitamin D intake is insufficient [1112]. Because vitamin D intake from foods or supplements is a major factor that can influence serum vitamin D levels, developing a simple and accurate FFQ to assess vitamin D intake among Koreans is crucial.
In Korea, approximately 60% of the total vitamin D intake comes from fish [12], causing considerable day-to-day variation based on fish consumption. Moreover, 99% of daily vitamin D intake in Koreans is derived from 6 food groups: fish, eggs, meat, dairy, cereals, and mushrooms [12]. Serum vitamin D concentrations are lower in spring and winter when UV exposure is low; hence, the impact of vitamin D intake from food is essential during these periods [12]. Therefore, the FFQ method, which can assess annual food intake, is more appropriate for capturing nutrient intake such as vitamin D. The primary purpose of a FFQ is to assess long-term dietary patterns, making it suitable for obtaining data on the habitual intake of large populations over extended periods [14]. Because FFQs can be developed to reflect the dietary habits and cultural lifestyles of a specific population, they must be validated within the appropriate cohort [1440].
In this study, we first verified the reproducibility of the newly developed FFQ by conducting 2 FFQ assessments approximately 9–11 months apart. Głąbska et al. [25] explored the applicability of an FFQ for assessing vitamin D intake among women in their 20s and 30s in Croatia. An FFQ was administered at 6-week intervals, which revealed a Spearman correlation of 0.80, ICC of 0.81, and weighted kappa of 0.62. Short intervals between measures can cause biased reproducibility. Differences between Głąbska’s study and this study likely result from survey interval differences [14].
While most validation studies of existing vitamin D FFQs compared results with 3–4 days of DR, this study used 12 days of DR across 4 seasons. This approach may have increased variability in food intake; however, it was necessary to account for seasonal production changes of vitamin D source foods. A U.S. study with 123 athletes comparing 7-day DR and FFQ showed an insignificant Spearman’s rank correlation coefficient (0.12), and the Bland–Altman plot results indicated the FFQ did not reflect DR well [27]. However, better validity of the FFQ developed in this study may be because of the limited variety of vitamin D source foods and fewer fortified foods consumed by Koreans, reducing individual variability [12]. Generally, nutrient intake assessments using FFQs tend to be higher than those using DRs, a phenomenon observed not only in vitamin D FFQ validation studies [202324252728] but also in FFQ validations for various other foods and nutrients [414243]. When participants recall the intake frequency of a limited number of foods through an FFQ, intake tends to be overestimated [14].
The newly developed FFQ showed a significant correlation with the 12-day DR for most food groups, especially fish and dairy, key sources of vitamin D in the Korean diet. This supports the FFQ’s accuracy in estimating vitamin D intake from these food groups. While the FFQ accurately measured vitamin D intake from fish, it tended to overestimate it in other groups, which minimally impacted total vitamin D levels. The high correlation and consistency in fish group assessments highlight the FFQ’s strengths given the predominant vitamin D intake from fish in Korean diets. Adjustments may be needed for food groups like ‘meat and its products’ and ‘mushrooms,’ where correlation was low. For eggs, higher vitamin D intake recorded in the 12-day DR than in the FFQ may be due to its unrecognized presence in processed foods like omelets, fried rice, toast, bread, and pastries. Therefore, explanations for processed foods containing eggs in the FFQ should be improved.
Based on this study, it was possible to further refine the food items in the previously developed FFQ to enhance its completeness. Among the food items included in the FFQ, those with very low consumption were shrimp, mayonnaise, and processed meats. Additionally, the number of food items for mushrooms could have led to an overestimation compared with the total vitamin D intake in this group (Supplementary Table 2). Therefore, the following 3 steps were proposed to reduce the number of vitamin D FFQ items and minimize the potential for overestimation: shrimp and mayonnaise should be removed from the food items; processed meats can be combined with pork into a single item; and for mushrooms, Juda’s ear mushroom, which contains a high level of 8.8 µg of vitamin D per 100 g, and other mushrooms, which range from 0.3–0.9 µg/100g in vitamin D content, can be divided into 2 separate items (Juda’s ear mushroom and mushrooms). As modified, the revised vitamin D FFQ now comprises 25 food items across 7 food groups (Supplementary Table 3). When the responses from our study were reanalyzed statistically in this way, the FFQ was validated, showing a Spearman’s rank correlation coefficient of 0.445 (P < 0.001), ICC of 0.539, quartile opposite of 2.63%, weighted kappa of 0.274, and a Bland–Altman plot index of 4.0%, indicating reliability (data not presented).
The FFQ analysis showed fish significantly contributed to vitamin D intake in postmenopausal women, while younger women consumed a more diverse range of vitamin D sources (Supplementary Fig. 1). Despite these differences, the developed FFQ may be suitable for assessing vitamin D intake among Korean women of all ages.
This study has a few limitations. First, the study was conducted only among women; hence, additional research involving men is necessary. Previous studies suggest no significant difference between men and women regarding vitamin D sources, indicating the FFQ could also apply to men [12]. However, further validation in men is needed. Second, the 12-day DR results showed middle-aged women consumed significant amounts of vitamin D from supplements and beverages. Thus, it is essential to include an investigation on supplements and nutritional drinks when researching FFQs in relation to vitamin D biomarkers.
Despite these limitations, this study also has several strengths. It developed an FFQ specifically tailored to the dietary habits of Koreans, verified by comparison with DRs. The study used a 12-day DR over a year, considering seasonal variations, providing a more accurate picture of annual vitamin D intake patterns.
XML Download