Using Food Network Analysis to Understand Meal Patterns in Pregnant Women with High and Low Diet Quality

Abstract

Abstract Objectives Our aim was to qualitatively assess networks describing interrelations among foods consumed at meals in pregnant women with high and low overall diet quality as defined by the Healthy Eating Index-2015 (HEI). Methods Multiple 24-hour dietary recalls (mean = 3.4) from 365 women in the Pregnancy Eating Attributes Study (PEAS) were collected across pregnancy. Food networks using intakes of 40 food groups were derived for each meal (breakfast, lunch, dinner) for women in the highest (n = 122) and lowest (n = 121) HEI tertiles using semiparametric gaussian copula graphical models (SGCGM). Networks are based on the resulting sparse matrix of partial correlations (edges) between food groups (nodes); communities within the networks were detected using the Louvain method (LM). Results In breakfast networks, 3 communities were detected in the low and 4 communities in the high group. In both groups, breakfast community (BC) 1 included white bread, eggs, cured meat and cheese; whole grain bread was also part of BC1 only in the high group. BC2 included milk and breakfast cereals in both groups; fried potatoes and sugar-sweetened beverages (SSB) were part of BC2 in the low group only. BC3, composed of cooked cereals and legumes/nuts/seeds in the high group, replaces a community composed of pancakes/other breads and sugars/sweets in the low group. In lunch networks, 5 communities were detected in both the low and high groups. Fried potatoes and SSBs were consumed with sandwiches in the low group but not in the high group. Vegetables were consumed more often (green P < .0001, red/orange P < .0001, other p 0.004) and in combinations with other foods in the high group. In dinner networks, 6 communities were detected in both groups. Sandwiches, fried potatoes and SSBs were consumed together and more often in the low group (5%, 12%, 23% of dinners, respectively) but separately in the high group (2%, 6%, 10% of dinners, respectively). Conclusions Food network analysis may provide a better understanding of the interrelationships of food intake at meals than traditional methods of meal pattern analysis. Differences observed in networks among pregnant women with low and with high diet quality may serve as a guide for meal-specific recommendations and interventions aimed at improving diet quality during pregnancy. Funding Sources This research was supported by the NICHD Intramural Research Program.