Dietary assessment in intermittent fasting: validation of a short food frequency questionnaire vs. food records in diurnal dry fasting and time-restricted eating

Abstract

Objectives: Food frequency questionnaire (FFQ) is a cost-effective method of dietary assessment in nutritional and clinical research. It can be easily adapted to different research questions or populations, but modified versions require careful validation. This study assessed the validity of a short 14-item semi-quantitative FFQ compared to weighted food records in a secondary analysis of an intermittent fasting trial.

Methods: Dietary assessment was conducted during the ParoFastin study, a controlled trial investigating the effects of religious Bahá'í fasting (19 days of diurnal dry fasting) and 16:8 time-restricted eating (TRE) on oral health and metabolic state compared to the habitual food intake. Daily consumption of meals, snacks, food groups, and overnight fasting time were assessed using both the short FFQ and food records. Food records were collected for 1 week at baseline and 19-21 days during the intervention and analyzed using PRODI^®, a professional dietary assessment software. The FFQ was completed once at baseline and twice during the intervention. Its validity was assessed using correlation and method agreement analysis, including Bland-Altman plots for continuous data. Energy and macronutrient intakes were quantified using food records only.

Results: Eight men and seven women, with a median age of 29 (27-34) years, were included in the validation analysis. Correlation coefficients ranged from 0.189 (tendency to snack) to 0.893 (meat consumption). Tendency to snack, frequency of snack consumption, and frequency of whole grain consumption showed insufficient agreement between the two methods. However, most questions of the short FFQ were found to be statistically valid in this population. According to food records, the energy, fat and carbohydrate intake were reduced during the Bahá'í fast and remained unchanged in the control and TRE groups compared to the baseline, while analysis of these parameters was not feasible based on the short FFQ.

Conclusion: Overall, good agreement for the methods was found, although data on snack tendency, frequency of snack consumption, and whole-grain consumption were unreliable, indicating a need for questionnaire modifications. In contrast to time-consuming food records, the short FFQ can be effectively used in clinical trials and medical practice for specific goals.

Clinical trial registration: https://drks.de/search/de/trial/DRKS00026701 German Clinical Trials Register (DRKS); identifier DRKS00026701.

Keywords: dietary assessment; food frequency questionnaire; intermittent fasting; religious fasting; time-restricted eating; validation.

FIGURE 1

Study design. After baseline, participants either observed the Bahá’í fast or followed a 16:8 TRE regimen for 19 days or did not change their dietary habits (control group). The 14-item FFQ data were collected at three visits: prior to intervention start (V1) and twice during the intervention (V2 and V3). Throughout the entire study period, participants completed paper-based weighted food records.

FIGURE 2

Bland–Altman plots as a measurement of method agreement. Difference between questionnaire and food record data plotted against the average value of the two methods. (A) Number of meals; (B) number of snacks; (C) overnight fasting time; (D) vegetable intake; (E) fruit intake; (F) sausage intake; (G) meat intake; (H) cheese intake; (I) confectionery (sweets) intake; (J) fast food intake; (K) sweet beverage intake; and (L) whole grain products. Notably, 95% limits of agreement (bias ±2*SD) as well as bias are shown as broken lines. The gray background indicates the 95% confidence interval of the bias. (G,J) Linear regression lines with significant slope (p < 0.05) are shown. (A,D–L) N = 42 data points = 15 participants in up to three recording periods. (B) N = 27 data points = 13 participants in up to three recording periods. (C) N = 40 data points = 15 participants in up to three recording periods.

FIGURE 3

Transformed Bland-Altman plots to correct for severe heteroscedasticity or proportional bias (N=42). Difference between log-transformed questionnaire and food record data plotted against the logarithmic average value of the two methods. Regression-based Bland-Altman analysis in case of significant slope of linear regression line (p < 0.05). (A) Sausage intake [log(Y+0.5)]. (B) Meat intake [log(Y+1)]. (C) Cheese intake [log(Y+1)]. (D) Fast food intake [log(Y+0.05)].

FIGURE 4

Average daily number of meals and snacks, overnight fasting time, and consumption of food groups according to FFQ and food records. (A) Number of meals; (B) number of snacks; (C) overnight fasting time; (D) vegetable intake; (E) fruit intake; (F) sausage intake; (G) meat intake; (H) cheese intake; (I) confectionery (sweets) intake; (J) fast food intake; (K) sweet beverage intake; and (L) whole grain products. Visualization of FFQ and food record data as Tukey boxplots. Circles represent outliers. Differences were statistically tested using a Wilcoxon signed-rank test. Stars indicate significance (*p < 0.05; **p < 0.01; ***p < 0.001). (A,D–L) N = 42 data points = 15 participants in up to three recording periods. (B) N = 27 data points = 13 participants in up to three recording periods. (C) N = 40 data points = 15 participants in up to three recording periods.

FIGURE 5

Daily consumption of energy (A) and the macronutrients proteins (B), fats (C), and carbohydrates (D) derived from food records. Bar diagrams illustrating mean daily intake for the three different groups, control (N = 4), Bahá’í (N = 5), and time-restricted eating (TRE; N = 3), during baseline and intervention. Error bars represent standard deviations. Three participants were excluded from this analysis because baseline data were not available. Comparison of the nutrient intake during baseline (V1) and intervention (V2 and V3) per group using paired-samples t-tests (*p < 0.05; **p < 0.01).