Dietary assessment traditionally relies on self-reported data, which are often inaccurate and may result in erroneous diet–disease risk associations. We illustrate how urinary metabolic phenotyping can be used as an alternative approach to obtain information on dietary patterns. We used two multipass 24 h dietary recalls, obtained on two occasions on average 3 weeks apart, paired with two 24 h urine collections from 1,848 US individuals; 67 nutrients influenced the urinary metabotype (metabolic phenotype) of 46 structurally identified metabolites characterized by ¹H NMR spectroscopy. We investigated the stability of each metabolite over time and showed that the urinary metabolic profile is more stable within individuals than reported dietary patterns. The 46 metabolites accurately predicted healthy and unhealthy dietary patterns in a free-living US cohort, and these predictions were replicated in an independent UK cohort. We mapped these metabolites into a host-microbial metabolic network to identify key pathways and functions related to diet. These data can be used in future studies to evaluate how this set of diet-derived, stable, measurable bioanalytical markers is associated with disease risk. This knowledge may give new insights into biological pathways that characterize the shift from a healthy to an unhealthy metabolic phenotype and hence indicate entry points for prevention and intervention strategies.

饮食评估传统上依赖于自我报告的数据，这些数据通常不准确，并且可能导致错误的饮食-疾病风险关联。我们说明了如何将尿液代谢表型作为获得饮食模式信息的替代方法。我们使用了两次多次全天候24小时饮食召回，这两次间隔平均3周，两次分别与来自1848名美国个体的两次24小时尿液收集配对；67种营养素影响了46种结构鉴定的代谢物的尿代谢型（代谢表型），其特征为¹1 H NMR光谱。我们调查了每种代谢物随时间的稳定性，结果表明，与报告的饮食模式相比，个体内的尿代谢状况更稳定。这46种代谢物准确地预测了美国自由人群的健康和不健康饮食模式，并且这些预测在独立的英国人群中得到了重复。我们将这些代谢物映射到宿主-微生物代谢网络中，以识别与饮食相关的关键途径和功能。这些数据可用于将来的研究中，以评估这组饮食来源的，稳定的，可测量的生物分析标记物与疾病风险之间的关系。