Get access

α-Priori and α-Posterior Dietary Pattern Analyses Have Similar Estimating and Discriminating Ability in Predicting 5-Y Incidence of Cardiovascular Disease: Methodological Issues in Nutrition Assessment

Authors

  • Demosthenes B. Panagiotakos,

    1. Author Panagiotakos is with Dept. of Dietetics-Nutrition, Harokopio Univ., Athens, Greece. Authors Pitsavos and Stefanadis are with First Cardiology Clinic, School of Medicine, Univ. of Athens, Greece. Direct inquiries to author Panagiotakos (E-mail: d.b.panagiotakos@usa.net).
    Search for more papers by this author
  • Christos Pitsavos,

    1. Author Panagiotakos is with Dept. of Dietetics-Nutrition, Harokopio Univ., Athens, Greece. Authors Pitsavos and Stefanadis are with First Cardiology Clinic, School of Medicine, Univ. of Athens, Greece. Direct inquiries to author Panagiotakos (E-mail: d.b.panagiotakos@usa.net).
    Search for more papers by this author
  • Christodoulos Stefanadis

    1. Author Panagiotakos is with Dept. of Dietetics-Nutrition, Harokopio Univ., Athens, Greece. Authors Pitsavos and Stefanadis are with First Cardiology Clinic, School of Medicine, Univ. of Athens, Greece. Direct inquiries to author Panagiotakos (E-mail: d.b.panagiotakos@usa.net).
    Search for more papers by this author

Abstract

ABSTRACT:  The 5-y incidence of cardiovascular disease (CVD) in relation to dietary habits was evaluated, using 2 methodological approaches, an α-priori and an α-posterior diet pattern analyses. The aim of this study was to compare these methods in predicting CVD events. From May 2001 to December 2002, 3042 men and women (>18 y) without any clinical evidence of CVD were enrolled in the ATTICA study. In 2006, the 5-y follow-up was performed (941 of the 3042 [31%] participants were lost to follow-up). Development of CVD (coronary heart disease, acute coronary syndromes, stroke, or other CVD) during the follow-up period was defined according to WHO-ICD-10 criteria. An α-posterior statistical method, principal component analysis (PCA), was applied to extract dietary patterns from various foods or food groups. Moreover, a special diet score that incorporates inherent characteristics of the Mediterranean diet (the MedDietScore) was also used to assess level of adherence to this diet (α-priori approach). Using PCA 10 components were extracted that explained the 57% of the total variation in intake. The model that also included various potential confounders together with the extracted components showed adequate goodness of fit and very good discriminating ability (C-statistic = 0.85). The other model that included the same potential confounders together with the MedDietScore also showed adequate goodness of fit and very good discriminating ability, too (C-statistic = 0.83). α-priori and α-posterior dietary assessment showed similar estimating and discriminating ability in predicting CVD, in our sample.

Ancillary