(P14-048-23) Reproducibility and Validity of Low-Carbohydrate Diet Indices Derived From Food-Frequency Questionnaires and 24-Hour Recalls

Objectives: Low-carbohydrate diet (LCD) indices are frequently used to study diet-disease associations, but the reproducibility and validity of these indices derived from food-frequency questionnaires (FFQs) and 24-hour dietary recalls (HDR24s) remain unknown. We investigated the reproducibility and validity of 5 LCD indices including the overall, plant-based, animal-based, healthful, and unhealthful LCDs.

Methods: Of 620 participants in Anhui Lifestyle Validation Study in China, 482 completed a 141-item FFQ at baseline and again 1 year later, and 4 quarterly 3 consecutive HDR24s (2 weekdays and 1 weekend, n = 482) over this same year; 138 participants completed 2 7-day dietary records (DDR7s) 1 month apart between FFQ assessments. The reproducibility of FFQ-derived LCD indices was assessed by comparing the 2 FFQs using intraclass correlation coefficients (ICCs). The validity of LCD indices derived from FFQs and 3 commonly used HDR24s including a random HDR24, 2 random HDR24s, and 3 consecutive HDR24s was evaluated by comparing with those from the average of 2 DDR7s and 4 quarterly 3 consecutive HDR24s, respectively.

Results: The FFQ-derived LCD indices showed moderate to good reproducibility (energy-adjusted ICCs: 0.57-0.62) and fair to moderate validity [energy-adjusted, deattenuated Spearman rank correlation coefficients (rs): 0.32-0.55]. The LCD indices derived from 2 random HDR24s (r_s: 0.74-0.81) and 3 consecutive HDR24s (r_s: 0.67-0.76) generally showed higher validity than those derived from a random HDR24 (r_s: 0.41-0.55).

Conclusions: The FFQ-derived LCDs demonstrated acceptable reproducibility and validity. Two random days or 3 consecutive days of dietary recalls seemed superior to one-day dietary recall in measuring LCDs in epidemiological studies.

Funding Sources: This work was supported by the National Natural Science Foundation of China.