University of Hartford, West Hartford, CT 06117, USA.
University of Wyoming, Laramie, WY, USA.
Nutrients. 2020 Apr 30;12(5):1276. doi: 10.3390/nu12051276.
We investigated the impact of nutrient intake on hydration biomarkers in cyclists before and after a 161 km ride, including one hour after a 650 mL water bolus consumed post-ride. To control for multicollinearity, we chose a clustering-based, machine learning statistical approach. Five hydration biomarkers (urine color, urine specific gravity, plasma osmolality, plasma copeptin, and body mass change) were configured as raw- and percent change. Linear regressions were used to test for associations between hydration markers and eight predictor terms derived from 19 nutrients merged into a reduced-dimensionality dataset through serial k-means clustering. Most predictor groups showed significant association with at least one hydration biomarker: 1) Glycemic Load + Carbohydrates + Sodium, 2) Protein + Fat + Zinc, 3) Magnesium + Calcium, 4) Pinitol, 5) Caffeine, 6) Fiber + Betaine, and 7) Water; potassium + three polyols, and mannitol + sorbitol showed no significant associations with any hydration biomarker. All five hydration biomarkers were associated with at least one nutrient predictor in at least one configuration. We conclude that in a real-life scenario, some nutrients may serve as mediators of body water, and urine-specific hydration biomarkers may be more responsive to nutrient intake than measures derived from plasma or body mass.
我们研究了 161 公里骑行前后骑行者营养摄入对水合生物标志物的影响，包括骑行后 650 毫升水冲击后 1 小时。为了控制多重共线性，我们选择了基于聚类的机器学习统计方法。将五种水合生物标志物（尿液颜色、尿液比重、血浆渗透压、血浆 copeptin 和体重变化）配置为原始和百分比变化。线性回归用于测试水合标志物与通过串行 k-均值聚类合并到降维数据集的 19 种营养素衍生的 8 个预测因子之间的关联。大多数预测因子组与至少一种水合生物标志物存在显著关联：1）血糖负荷+碳水化合物+钠，2）蛋白质+脂肪+锌，3）镁+钙，4）肌醇，5）咖啡因，6）纤维+甜菜碱，7）水；钾+三种多元醇，甘露醇+山梨醇与任何水合生物标志物均无显著关联。所有五种水合生物标志物在至少一种配置中均与至少一种营养预测因子相关。我们的结论是，在现实生活场景中，一些营养素可能作为身体水分的介质，尿液特异性水合生物标志物可能比来自血浆或体重的测量更能响应营养摄入。
