Moskal Aurelie, Pisa Pedro T, Ferrari Pietro, Byrnes Graham, Freisling Heinz, Boutron-Ruault Marie-Christine, Cadeau Claire, Nailler Laura, Wendt Andrea, Kühn Tilman, Boeing Heiner, Buijsse Brian, Tjønneland Anne, Halkjær Jytte, Dahm Christina C, Chiuve Stephanie E, Quirós Jose R, Buckland Genevieve, Molina-Montes Esther, Amiano Pilar, Huerta Castaño José M, Gurrea Aurelio Barricarte, Khaw Kay-Tee, Lentjes Marleen A, Key Timothy J, Romaguera Dora, Vergnaud Anne-Claire, Trichopoulou Antonia, Bamia Christina, Orfanos Philippos, Palli Domenico, Pala Valeria, Tumino Rosario, Sacerdote Carlotta, de Magistris Maria Santucci, Bueno-de-Mesquita H Bas, Ocké Marga C, Beulens Joline W J, Ericson Ulrika, Drake Isabel, Nilsson Lena M, Winkvist Anna, Weiderpass Elisabete, Hjartåker Anette, Riboli Elio, Slimani Nadia
Dietary Exposure Assessment Group, International Agency for Research on Cancer, Lyon, France.
Nutritional Epidemiology Group, International Agency for Research on Cancer, Lyon, France.
PLoS One. 2014 Jun 5;9(6):e98647. doi: 10.1371/journal.pone.0098647. eCollection 2014.
Compared to food patterns, nutrient patterns have been rarely used particularly at international level. We studied, in the context of a multi-center study with heterogeneous data, the methodological challenges regarding pattern analyses.
METHODOLOGY/PRINCIPAL FINDINGS: We identified nutrient patterns from food frequency questionnaires (FFQ) in the European Prospective Investigation into Cancer and Nutrition (EPIC) Study and used 24-hour dietary recall (24-HDR) data to validate and describe the nutrient patterns and their related food sources. Associations between lifestyle factors and the nutrient patterns were also examined. Principal component analysis (PCA) was applied on 23 nutrients derived from country-specific FFQ combining data from all EPIC centers (N = 477,312). Harmonized 24-HDRs available for a representative sample of the EPIC populations (N = 34,436) provided accurate mean group estimates of nutrients and foods by quintiles of pattern scores, presented graphically. An overall PCA combining all data captured a good proportion of the variance explained in each EPIC center. Four nutrient patterns were identified explaining 67% of the total variance: Principle component (PC) 1 was characterized by a high contribution of nutrients from plant food sources and a low contribution of nutrients from animal food sources; PC2 by a high contribution of micro-nutrients and proteins; PC3 was characterized by polyunsaturated fatty acids and vitamin D; PC4 was characterized by calcium, proteins, riboflavin, and phosphorus. The nutrients with high loadings on a particular pattern as derived from country-specific FFQ also showed high deviations in their mean EPIC intakes by quintiles of pattern scores when estimated from 24-HDR. Center and energy intake explained most of the variability in pattern scores.
CONCLUSION/SIGNIFICANCE: The use of 24-HDR enabled internal validation and facilitated the interpretation of the nutrient patterns derived from FFQs in term of food sources. These outcomes open research opportunities and perspectives of using nutrient patterns in future studies particularly at international level.
与食物模式相比，营养模式很少被使用，尤其是在国际层面。在一项包含异质数据的多中心研究背景下，我们研究了模式分析方面的方法学挑战。
方法/主要发现：我们从欧洲癌症与营养前瞻性调查（EPIC）研究中的食物频率问卷（FFQ）中识别出营养模式，并使用24小时膳食回顾（24-HDR）数据来验证和描述营养模式及其相关食物来源。还研究了生活方式因素与营养模式之间的关联。主成分分析（PCA）应用于来自特定国家FFQ的23种营养素，这些数据来自所有EPIC中心（N = 477,312）。EPIC人群代表性样本（N = 34,436）可获得的统一24-HDR，通过模式得分五分位数以图形方式呈现了营养素和食物的准确平均组估计值。将所有数据合并的总体PCA捕获了每个EPIC中心解释的很大一部分方差。识别出四种营养模式，解释了总方差的67%：主成分（PC）1的特征是植物性食物来源的营养素贡献高，动物性食物来源的营养素贡献低；PC2的特征是微量营养素和蛋白质的贡献高；PC3的特征是多不饱和脂肪酸和维生素D；PC4的特征是钙、蛋白质、核黄素和磷。从特定国家FFQ得出的在特定模式上具有高负荷的营养素，当从24-HDR估计时，其在EPIC平均摄入量中按模式得分五分位数也显示出高偏差。中心和能量摄入解释了模式得分中大部分的变异性。
结论/意义：24-HDR的使用实现了内部验证，并有助于从食物来源角度解释从FFQ得出的营养模式。这些结果为未来研究，尤其是国际层面的研究，开启了使用营养模式的研究机会和前景。
