Baker Lindsay B, Heaton Lisa E, Stein Kimberly W, Nuccio Ryan P, Jeukendrup Asker E
Gatorade Sports Science Institute, 617 W, Main St,, Barrington, IL 60010, USA.
Nutr J. 2014 Apr 30;13:41. doi: 10.1186/1475-2891-13-41.
We developed a digital dietary analysis tool for athletes (DATA) using a modified 24-h recall method and an integrated, customized nutrient database. The purpose of this study was to assess DATA's validity and relative validity by measuring its agreement with registered dietitians' (RDs) direct observations (OBSERVATION) and 24-h dietary recall interviews using the USDA 5-step multiple-pass method (INTERVIEW), respectively.
Fifty-six athletes (14-20 y) completed DATA and INTERVIEW in randomized counter-balanced order. OBSERVATION (n = 26) consisted of RDs recording participants' food/drink intake in a 24-h period and were completed the day prior to DATA and INTERVIEW. Agreement among methods was estimated using a repeated measures t-test and Bland-Altman analysis.
The paired differences (with 95% confidence intervals) between DATA and OBSERVATION were not significant for carbohydrate (10.1%, -1.2-22.7%) and protein (14.1%, -3.2-34.5%) but was significant for energy (14.4%, 1.2-29.3%). There were no differences between DATA and INTERVIEW for energy (-1.1%, -9.1-7.7%), carbohydrate (0.2%, -7.1-8.0%) or protein (-2.7%, -11.3-6.7%). Bland-Altman analysis indicated significant positive correlations between absolute values of the differences and the means for OBSERVATION vs. DATA (r = 0.40 and r = 0.47 for energy and carbohydrate, respectively) and INTERVIEW vs. DATA (r = 0.52, r = 0.29, and r = 0.61 for energy, carbohydrate, and protein, respectively). There were also wide 95% limits of agreement (LOA) for most method comparisons. The mean bias ratio (with 95% LOA) for OBSERVATION vs. DATA was 0.874 (0.551-1.385) for energy, 0.906 (0.522-1.575) for carbohydrate, and 0.895(0.395-2.031) for protein. The mean bias ratio (with 95% LOA) for INTERVIEW vs. DATA was 1.016 (0.538-1.919) for energy, 0.995 (0.563-1.757) for carbohydrate, and 1.031 (0.514-2.068) for protein.
DATA has good relative validity for group-level comparisons in athletes. However, there are large variations in the relative validity of individuals' dietary intake estimates from DATA, particularly in athletes with higher energy and nutrient intakes. DATA can be a useful athlete-specific, digital alternative to conventional 24-h dietary recall methods at the group level. Further development and testing is needed to improve DATA's validity for estimations of individual dietary intakes.
我们使用改良的24小时回顾法和一个整合的、定制的营养数据库开发了一种针对运动员的数字饮食分析工具（DATA）。本研究的目的是通过分别测量其与注册营养师的直接观察（OBSERVATION）以及使用美国农业部五步多次回顾法的24小时饮食回顾访谈（INTERVIEW）的一致性，来评估DATA的效度和相对效度。
56名运动员（14 - 20岁）以随机交叉平衡顺序完成了DATA和INTERVIEW。OBSERVATION（n = 26）包括注册营养师记录参与者在24小时内的食物/饮料摄入量，且在进行DATA和INTERVIEW的前一天完成。使用重复测量t检验和Bland - Altman分析来估计方法之间的一致性。
DATA与OBSERVATION之间碳水化合物（10.1%，-1.2 - 22.7%）和蛋白质（14.1%，-3.2 - 34.5%）的配对差异不显著，但能量的配对差异显著（14.4%，1.2 - 29.3%）。DATA与INTERVIEW在能量（-1.1%，-9.1 - 7.7%）、碳水化合物（0.2%，-7.1 - 8.0%）或蛋白质（-2.7%，-11.3 - 6.7%）方面没有差异。Bland - Altman分析表明，OBSERVATION与DATA之间差异绝对值与均值的显著正相关（能量和碳水化合物的r分别为0.40和0.47），以及INTERVIEW与DATA之间差异绝对值与均值的显著正相关（能量、碳水化合物和蛋白质的r分别为0.52、0.29和0.61）。大多数方法比较的95%一致性界限（LOA）也很宽。OBSERVATION与DATA的平均偏差率（95% LOA），能量为0.874（0.551 - 1.385），碳水化合物为0.906（0.522 - 1.575），蛋白质为0.895（0.395 - 2.031）。INTERVIEW与DATA的平均偏差率（95% LOA），能量为1.016（0.538 - 1.919），碳水化合物为0.995（0.563 - 1.757），蛋白质为1.031（0.514 - 2.068）。
DATA在运动员群体水平比较中具有良好的相对效度。然而，从DATA估计个体饮食摄入量的相对效度存在很大差异，特别是在能量和营养素摄入量较高的运动员中。在群体水平上，DATA可以作为一种针对运动员的、有用的数字替代方法，替代传统的24小时饮食回顾方法。需要进一步开发和测试以提高DATA在估计个体饮食摄入量方面的效度。
