Exploring meal based dietary intake assessment: Development of statistical analysis strategies and development of a meal -based dietary intake assessment tool

There is limited investigation of meal patterns in nutrition research, that is, meal-based aspects of dietary intake including the timing or distribution of meals across a day, the combinations of foods consumed within meals, or the combinations of meals over a given period. Meal pattern research has the potential to further our understanding of the link between diet and disease and complement existing food-based dietary guidelines and inform personalised nutrition. Despite an increasing research focus on meal patterns in recent years, there is no research comparing the various statistical techniques used to identify meal patterns, limited research on the relationship between meal patterns and health, and an absence of a meal-based method of dietary intake assessment. The objectives of this thesis were therefore to review existing meal pattern research, refine the framework for meal pattern research, identify relationships between meal patterns and health, and develop a new meal-based method of dietary intake assessment. Analysis of data from the Irish National Adult Nutrition Survey (NANS) carried out for this thesis condensed the 27336 meals consumed by participants into 63 generic, or characteristic, meals that provide an accurate estimate of nutrient intakes. Using data from the National Health and Nutrition Examination Survey (NHANES) in the USA, this process was reproduced, and further analysis was conducted to identify the different combinations of generic meals consumed by the participants (meal patterns). This was carried out using three different statistical approaches for comparison: partitioning around the medoids clustering, principal component analysis (PCA), and latent class analysis (LCA). Differences arose among the methods with respect to the number of patterns identified, the identification of meal skipping in the patterns, and the different combinations of generic meal that were consumed. An exploratory analysis of the meal patterns arising from clustering identified differences among the meal patterns for diet quality but not for health variables. A novel meal-based dietary assessment tool, based on the generic meals identified, was found to have moderate agreement with 24-hour recalls in estimating nutrient intakes, with stronger agreement for some nutrients compared to others. In conclusion, a reproducible clustering approach can identify generic meals in dietary intake datasets. Meal patterns can be identified using different statistical methods; however, differences arise in the meal patterns identified by different methods, and their strengths and weaknesses should be considered when choosing a method. Some associations between meal patterns and diet quality were identified. A meal-based method of dietary intake assessment is a feasible method of collecting dietary intake data. Further research is required to further develop this tool and improve accuracy across a range of nutrients.