Obesity Abstracts

Background: In the UK, there are few brief, validated dietary assessment tools available. The purpose of this study was to use dietary patterns methods to identify food groups and sample characteristics that were associated with diet quality to inform the development of a brief, diet quality assessment tool.

Methods: Diet quality was explored using a priori and a posteriori dietary patterns methods in adults from the UK National Diet and Nutrition Survey (n=2083). Principal Component Analysis of 60 foods was used to derive empirical dietary patterns that were analysed for their associations with sample characteristics and nutrient biomarkers. Backwards elimination regression was conducted with 60 foods and sample characteristics to identify models that were independently predictive of a theory driven, validated, Nutrient-based Diet Quality Score (NDQS). Confirmatory analysis and further exploratory analysis identified the most parsimonious models of diet quality.

Results: Four a posteriori dietary patterns, subjectively labelled as ‘fruit, vegetables, oily fish, ‘snacks, fast food, fizzy drinks’, ‘meat, potatoes, beer’ and ‘sugary food, dairy’ explained 13.6% of the dietary variance in the data. Three patterns were clearly associated with nutrient biomarkers in the expected direction. Fourteen foods characterising these patterns were included in a ‘diet quality model’. Backwards elimination regression identified a second model of 12 foods and a third model of 9 foods, age and smoking status. Confirmatory analysis showed all models were moderately associated with the NDQS (R²=0.29, 0.33 and 0.33 respectively). Further analyses revealed a 5-item tool of ‘fruit’ (B=0.04, P<0.001), ‘vegetables’ (B=0.03, P<0.001), ‘sugary drinks’ (B=−0.004, P=0.01), ‘coated chicken’ (B=−0.05, P=0.03) and ‘wholemeal bread’ (B=0.04, P<0.001) was moderately associated with the NDQS (R²=0.26). When ‘coated chicken’ was replaced by ‘beer, lager and cider’ the model fit was improved (R²=0.28). A tool of ‘fruit’, ‘vegetables’ and ‘smoking status’ was similarly predictive (R²=0.26).

Conclusion: Brief tools can be used in the UK population to assess and monitor broad patterns of diet quality and may be of particular help in pragmatic evaluations with time and resource limitations.

Keywords: Diet Quality, dietary assessment, dietary patterns

Disclosures: Work was funded by MRC Scholarship. First author currently consults for Slimming World.