School of Health and Sport Sciences, Faculty of Science, Health, Education and Engineering, University of the Sunshine Coast, Maroochydore, Queensland, Australia.
Eur J Clin Nutr. 2017 Nov;71(11):1353-1359. doi: 10.1038/ejcn.2017.23. Epub 2017 Mar 15.
BACKGROUND/OBJECTIVES: Nutrient profiling models classify the healthiness of foods based on their nutritional composition and provide the science that underlies nutrition signposting schemes. The two objectives were to examine the construct validity of the Health Star Rating (HSR) system by determining its diagnostic accuracy and to detect the optimal HSR cutoff points to define healthiness in packaged dairy foods. We hypothesised that ultra-processed dairy, defined by NOVA, would have less stars (less healthy) and non-ultra-processed dairy would have more stars (more healthy).
SUBJECTS/METHODS: The diagnostic accuracy of the HSR system used for 621 dairy foods for sale in an Australian regional supermarket was investigated. The healthiness of packaged dairy was measured using the NOVA food classification system.
The dairy beverages model was found to discriminate between healthy and less healthy dairy beverages as classified by NOVA (AUC: 0.653; 95% CI: 0.556-0.750; P=0.005). A receiver operating characteristic curve analysis for dairy beverages demonstrated that the optimal cutoff point corresponded to a rating of four stars. There was no discrimination power when using the HSR for predicting the health value of yoghurt and other dairy, or cheeses.
At the optimal cutoff point of four stars the HSR has a high sensitivity but a low specificity to correctly classify healthy packaged dairy beverages, as defined by NOVA. We provide evidence to support the construct validity of the HSR model for dairy beverages, but not for the models used for yoghurts and other dairy products, or cheeses.
背景/目的：营养成分分析模型根据食物的营养成分对其健康程度进行分类，并为营养标志计划提供科学依据。本研究的两个目的是通过确定健康星级评分（HSR）系统的诊断准确性来检验其结构有效性，并检测用于定义包装乳制品健康程度的最佳 HSR 切点。我们假设，超加工乳制品（由 NOVA 定义）的星级评分较低（健康程度较低），而非超加工乳制品的星级评分较高（健康程度较高）。
受试者/方法：对澳大利亚一家地区超市销售的 621 种乳制品的 HSR 系统的诊断准确性进行了研究。使用 NOVA 食品分类系统对包装乳制品的健康程度进行了测量。
研究发现，乳制品饮料模型能够区分 NOVA 分类的健康和不健康乳制品饮料（AUC：0.653；95%CI：0.556-0.750；P=0.005）。对乳制品饮料进行的受试者工作特征曲线分析表明，最佳切点对应四星评级。使用 HSR 预测酸奶和其他乳制品或奶酪的健康价值时，没有判别能力。
在最佳的四星切点，HSR 对 NOVA 定义的健康包装乳制品饮料的正确分类具有较高的敏感性，但特异性较低。我们提供的证据支持 HSR 模型对乳制品饮料的结构有效性，但不支持用于酸奶和其他乳制品产品或奶酪的模型。
