Modeling Differential Effects of Maternal Dietary Patterns across Severity Levels of Preterm Birth Using a Partial Proportional Odds Model

Preterm birth is a common cause of death worldwide of children under the age of five years. This condition is linked with short and long term neonatal morbidity and mortality. Maternal nutrition during pregnancy has a profound effect on fetal growth and development and subsequently also on the incidence of preterm birth. The aim of this study was to assess the differential effect of dietary patterns of pregnant women across ordered levels of preterm birth. Dietary assessments were performed using a food frequency questionnaire, presented to 687 pregnant women, in the “Mother and Child in the Environment” birth cohort during the period of 2013 to 2017. Each pregnancy resulted in a live birth. Eight dietary patterns were extracted, using exploratory factor analysis. The partial proportional odds model was employed to model severity levels of preterm birth. The partial proportional odds model has been recognized to be a flexible approach since it allows the effect of predictor variables to vary across categories of the ordinal response variable of interest. Women with increased consumption of vegetable-rich foods showed a reduced risk of very to moderately preterm birth incidence (AOR = 0.73, 95% CI = (0.531, 0.981), p = 0.036). Lower odds of very/moderately preterm birth compared to late preterm or term birth were observed for women following “nuts and rice foods” dietary pattern (AOR = 0.25, 95% CI = (0.099, 0.621), p = 0.003). High dietary consumption of starch foods dietary pattern (AOR = 2.09, 95% CI = (1.158, 3.769), p = 0.014) was associated with the most severe level of preterm birth outcome incidence, i.e. very/moderately preterm birth. The partial proportional odds modeling allowed the description of the effect of maternal dietary patterns across the different severity levels of preterm birth.