Development and validation of a food frequency questionnaire for Japanese athletes (FFQJA)

0
22

Development of FFQJA

Dataset

Pooled data from our previous studies of dietary records were used to develop the FFQJA. This dataset was collected from 440 Japanese athletes (220 men and 220 women) aged 15–28 years. In this study, athletes were defined as people who participated in training sessions to improve their performance and regularly took part in competitions. The median and interquartile ranges for height, body mass, and Body Mass Index (BMI) were 175.0 (171.0, 179.0) cm, 72.4 (67.2, 83.0) kg, and 23.8 (22.3, 26.1) kg/m2, respectively, for men, and 162.3 (157.7, 166.9) cm, 57.9 (52.8, 66.1) kg and 22.2 (20.8, 24.3) kg/m2, respectively, for women. Participants were engaged in track and field, swimming, soccer, sailing, volleyball, rugby, American football, basketball, handball, baseball, softball, badminton, lacrosse, bicycle racing, judo, gymnastics, rhythmic gymnastics, weightlifting, boxing, fencing, canoeing and cheerleading. This dataset was already de-identified. The ethics committee approved the use of this dataset without personal informed consent. The dataset included 3–7 days of dietary records for each participant, but we randomly selected 1 day of dietary records for each. Medians and interquartile ranges of intake of energy, protein, fat, and carbohydrate were 2834 (2178, 3553) kcal, 94.5 (74.2, 118.8) g, 85.7 (65.0, 112.0) g, and 383.9 (294.9, 508.2) g, respectively.

Dietary records

Participants were given a dietary log, a ruler, a scale, and a digital camera. Trained registered dietitians explained the dietary record method, and participants were instructed to record food and drink names, weight, and/or size in detail. Participants were asked to take photographs of all foods and drinks consumed. When processed food/drinks were consumed, participants kept the food labels. Trained registered dietitians confirmed the logs with participants. The registered dietitians then calculated the nutrient intake using nutrient calculation software (WELLNESS 21 ver. 2.84; Top Business System, Okayama, Japan) with the Standard Tables of Food Composition in Japan, 2015 [26]. The nutrient content of processed foods and foods eaten outside the home or purchased to go were calculated using the nutrient databases of the food producers, or divided into ingredients using photographs and logs. Before we used this dataset, one research nutritionist (KO) checked all data to assess whether the calculations had been conducted in a similar way.

Concept of FFQJA

To assess changes in dietary intake against the training schedule, we designed the FFQJA to calculate the average intake over a 2-week period. Eleven nutrients were selected: energy, protein, fat, carbohydrate, calcium (Ca), iron (Fe), retinol activity equivalents (RAE), vitamin B1, vitamin B2, vitamin C, and dietary fiber. These nutrients were selected based on the sufficiency of the nutritive data for all Standard Tables of Food Composition in Japan 2015 [26], food labels of supplements and processed foods, and nutrition recommendations for athletes.

Selection of food groups

Participants recorded 1220 foods in 440 dietary logs (Fig. 1). First, all reported foods, drinks, and supplements were grouped into 103 items based on their conceptual similarity. We defined the percentage contribution of nutrient k by food i as the arithmetic mean of the individual percentage contribution of nutrient k by food (IPCjik), which was estimated by ( {IPC}_{ji k}=frac{Q_{ji}{D}_{ik}}{sum limits_{i=1}^{103}{Q}_{ji}{D}_{ik}}times 100, ) if ( sum limits_{i=1}^{103}{Q}_{ji}{D}_{ik}=0 ) then IPCj = 0 was assumed. The percentage contribution of nutrient k by food i was calculated as ( sum limits_{j=1}^{440}{IPC}_{jik}/440, ) where j = 1…440 subjects, i = 1…103 food items, k = 1…11 nutrients, Q = grams of foods consumed and D = nutrient content per gram of food. We applied stepwise multiple regression analysis with the overall intake of each nutrient as dependent variables, and 103 grouped items as independent variables. This stepwise regression analysis was used to identify the food items that contributed at least 90% of the between-person variability. Finally, we chose 66 items both by a cumulative percentage contribution of 90% and 0.90 cumulative multiple regression coefficient from the multiple regression analysis with total energy or nutrient intake as dependent variables, and energy or nutrient intake from each item as independent variables. After these selections, we excluded sugar, sauces, soy sauce, stock, and other seasonings (sweet sake, ketchup, and spices) because these items were difficult to report for athletes who did not cook for themselves. Sugar was mainly used as a seasoning during cooking, and not added to drinks (i.e., tea and coffee). Most of the seasoning  intake was included as part of the intake of miso soup, curry or stew, fried dishes, and stir-fried dishes. In addition to the remaining 59 items, we added nuts, honey or maple syrup, bread filled with meat or vegetables, sponge cake or steamed bread, rice crackers, chips, and alcoholic drinks to the questionnaire because these items are often mentioned in nutritional counseling for athletes in Japan [27] (total number of food items = 66). From these 66 items, we presented supplements, minor cereals, sesame, and soymilk as supplemental questions, because these items were thought to be difficult to answer using the same type of question used for the other foods/food groups, and it was difficult to establish portion sizes and/or nutrient databases for the FFQJA. Minor cereals and soy milk were included within the cumulative 90% contribution to Fe and carbohydrate intakes. However, fewer than 30 participants consumed these foods. Sesame contributed to Ca and Fe intake, but the average consumption was only 2.5 g. Establishing average nutrient intake as a nutrient database for the FFQJA for supplements from one portion is difficult because of the wide variation in supplements. Thus, we presented these four items as supplemental questions to prompt participants to list supplements and foods eaten more than once a week, directly reporting the frequency and amount of these foods. As a result, the basic FFQ questionnaire included 62 food items (The food groups were shown in Appendix).

Fig. 1
figure1

Flow chart of the selection of the foods/food groups list in the food frequency questionnaire for Japanese athletes

Frequency and portion size

The consumption frequencies for 40 items that are usually eaten as meals were reported separately for breakfast, lunch, dinner, and supplement consumption. The consumption frequency for drinks and sweets in a day was also reported. The frequency was classified into four categories: almost every day, three to four times per week, once or twice a week, and never or rarely. The FFQJA assessed food intake over 2 weeks, and “never or rarely” was defined as less than once a week.

The standard portion size was calculated for foods from the dietary records using typical values and/or natural units from the literature. The amount of food consumed was reported via open-ended responses regarding how many times the standard portion size was consumed.

Nutrient database for the FFQJA

A nutrient database for the FFQJA was constructed using dietary record data. Using the foods classified into each item in the FFQJA, a weighted nutrient profile of 1–11 foods for each food item was calculated using the Standard Tables of Food Composition in Japan 2015 [26]. The food used in the calculation was selected using the appearance rate or weight contribution for each food item in the dietary record, covering approximately 70% of the total intake of each nutrient from each food group. When participants answered supplemental questions, nutrient content was obtained from the Standard Tables of Food Composition in Japan 2015 [26] or the nutrient databases of the food producers.

Validation of FFQJA

Protocol

The FFQJA was conducted on the first day of the study. A self-administered dietary record measure was completed over the following 2 weeks (Fig. 2).

Fig. 2
figure2

Protocol for development and reproducibility studies

Participants

We distributed handbills about the study at one high school and one university. We followed this with explanatory meetings for athletes who were interested in this study. A total of 92 athletes aged 15–21 years participated in the validation study. Of these athletes, 14 individuals were excluded before all analyses, of whom four did not complete the 14-day dietary record, one changed his usual diet because he started weight reduction before a competition, and nine were outliers (more or less than three standard deviations from the mean) for energy, protein, fat or carbohydrate intake assessed by the dietary record. This left 78 athletes (41 men and 37 women) in the validation study. Median and interquartile values for height, body mass and BMI were 177.5 (170.2, 182.8) cm, 70.4 (66.7, 75.6) kg, and 22.7 (21.6, 24.0) kg/m2, respectively, for men, and 160.0 (157.6, 164.1) cm, 56.7 (52.7, 59.9) kg, and 21.9 (20.7, 23.9) kg/m2, respectively, for women. These athletes participated in track and field, ice hockey, weightlifting, bicycle racing, judo, swimming, skiing, skating, table tennis, volleyball, basketball, baseball, lacrosse, and wrestling. All participants and/or their guardians (for athletes younger than 20 years old) were fully informed about all aspects of the study and provided informed consent upon entering the study. This study was approved by the ethics committee of Waseda University (No. 2015–320, 22nd Jan 2015).

Dietary records

Dietary records were kept over 14 days, to match the assessment period of the FFQJA. The same method and database used to develop the FFQJA were used in the validation study. However, we did not include the nutrient intakes of supplements. To calculate whole nutrient intake from both dietary records and the FFQJA, we added the nutrient contents from supplements taken by each athlete because the FFQJA did not include average nutrient values for supplements.

Reproducibility of FFQJA

Protocol

The second FFQJA was taken within a week of the 14-day dietary record completion (Fig. 2).

Reproducibility of FFQJA

Only one school from the validation study joined the reproducibility study. All 36 athletes from one school completed the second FFQJA.

Statistical analysis

Data were expressed as median and interquartile range because most variables were not normally distributed.

In the validation study, nutrient intakes and intakes from each food group calculated from dietary records and the FFQJA were compared using crude Spearman’s correlation coefficients (CCs), energy-adjusted CCs, and intraclass CCs (ICCs). The residual method proposed by Willett et al. was used to calculate the energy-adjusted variables [28, 29]. Differences between energy and nutrient intake assessed by dietary record and the FFQJA were compared using the Wilcoxon rank sum test. To examine the effectiveness of ranking, we divided the participants in the validation study into four categories by the intake of each nutrient. We then compared the proportion of athletes who were classified into the same, adjacent, ± 2, and ± 3 categories. Kappa statistics were also calculated. Degree of agreement for the weighted kappa values (Kw) were characterized as poor (Kw <  0.20), acceptable (Kw = 0.20–0.59), and good agreement (Kw ≥ 0.60) [30].

In the reproducibility study, Spearman’s CCs, energy-adjusted CCs, and ICCs were calculated between the first and second FFQJA. The residual method proposed by Willett et al. was used to calculate the energy-adjusted variables [28, 29].

P values less than 0.05 were considered statistically significant. All statistical analyses were conducted using IBM SPSS Statistics ver. 22 (IBM Japan, Ltd., Tokyo, Japan).