National Research Council/Institute of Biomedical Technologies, Milan, Italy.
J Nutr Sci. 2021 Nov 15;10:e97. doi: 10.1017/jns.2021.93. eCollection 2021.
When evaluating the impact of macronutrient intakes on health outcomes, researchers in nutritional epidemiology are mostly interested in two types of information: the relative importance of the individual macronutrients and the absolute effect of total energy intake. However, the usual substitution models do not allow these separate effects to be disentangled. Dietary data are typical examples of compositional data, which convey relative information and are, therefore, meaningfully expressed in the form of ratios. Various formulations of log-ratios have been proposed as a means of analysing compositional data, and their interrelationships when they are used as predictors in regression models have been previously reported. This note describes the application of distinct log-ratio transformations to the composition of dietary macronutrients and discusses the interpretative implications of using them as explanatory variables in regression models together with a term for the total composition (total energy intake). It also provides examples that consider serum glucose levels as the health outcome and are based on data coming from an Italian population-based study. The log-ratio transformation of dietary data has both numerical and conceptual advantages, and overcomes the drawbacks of traditional substitution models.
在评估宏量营养素摄入量对健康结果的影响时，营养流行病学研究人员主要关注两类信息：个体宏量营养素的相对重要性和总能量摄入的绝对影响。然而，常用的替代模型不允许分离这些单独的影响。饮食数据是组成数据的典型例子，它们传达相对信息，因此以比例的形式有意义地表达。已经提出了各种对数比的公式化方法，作为分析组成数据的一种手段，并且之前已经报道了它们在回归模型中用作预测变量时的相互关系。本说明描述了将不同的对数比变换应用于饮食中宏量营养素的组成，并讨论了将它们与总组成（总能量摄入）一起用作回归模型中的解释变量的解释意义。它还提供了一些示例，这些示例将血清葡萄糖水平作为健康结果，并基于来自意大利基于人群的研究的数据。饮食数据的对数比变换具有数值和概念上的优势，并克服了传统替代模型的缺点。
