Development of a Portfolio Diet Score and Its Concurrent and Predictive Validity Assessed by a Food Frequency Questionnaire

• Published in: Nutrients 2021-08, Vol.13 (8), p.2850
• Main Authors: Glenn, Andrea J, Boucher, Beatrice A, Kavcic, Chloe C, Khan, Tauseef A, Paquette, Melanie, Kendall, Cyril W. C, Hanley, Anthony J, Jenkins, David J. A, Sievenpiper, John L
• Format: Article
• Language: English
• Subjects: Antilipemic agents; Barley; Body weight; Cardiovascular diseases; Cholesterol; Confidence intervals; Correlation coefficient; Correlation coefficients; Diet; diet index; Dietary fiber; dietary pattern; FDA approval; Grocery industry; Health care; Health risks; Households; Intervention; Low density lipoprotein; Low density lipoproteins; Nutrients; Nutrition research; Nuts; Population; portfolio diet; Proteins; Questionnaires; Risk analysis; Risk factors; Risk management; Statistical analysis; Surveys; validity
• ISSN: 2072-6643; 2072-6643
• DOI: 10.3390/nu13082850

The Portfolio Diet, a plant-based portfolio of cholesterol-lowering foods, has been shown to reduce low-density lipoprotein cholesterol (LDL-C), and other cardiovascular risk factors, in randomized controlled trials (RCTs). It is not known if these beneficial effects translate to a lower incidence of cardiovascular disease (CVD). To support examinations between Portfolio Diet adherence and disease, a Portfolio Diet score (PDS) was developed and its predictive and concurrent validity was assessed within the Toronto Healthy Diet Study, a six-month RCT in overweight adults. Predictive validity was assessed using change in the PDS measured by food frequency questionnaire (FFQ) and concomitant change in LDL-C from baseline to six months using multiple linear regression, adjusted for potential confounders (n = 652). Concurrent validity was assessed in a subset of participants (n = 50) who completed the FFQ and a 7-day diet record (7DDR) at baseline. The PDS determined from each diet assessment method was used to derive correlation coefficients and Bland–Altman plots to assess the between-method agreement. The change in PDS was inversely associated with change in LDL-C (β coefficients: −0.01 mmol/L (95% confidence intervals (CIs): −0.02, −0.002; p = 0.02). The correlation between the PDS from the FFQ and 7DDR was 0.69 (95% CIs: 0.48, 0.85). The Bland–Altman plot showed reasonable agreement between the score from the FFQ and 7DDR. These findings indicate predictive validity of the PDS with lower LDL-C, and reasonable concurrent validity of the PDS as assessed by an FFQ against a 7DDR.