Statistics Seminar - 04/19/22

Abstract: Precision nutrition is an emerging research field in nutritional sciences. Being a major risk to children's neurobehavioral and cognitive development, excessive in utero exposure to lead for embryos would be detrimental if no intervention is in place. The calcium supplementation trial conducted by the ELEMENT team aims to study the effect of daily calcium supplement in reducing maternal lead exposure to infants during pregnancy. This article focuses on establishing an individualized treatment rule (ITR) that can guide pregnant women on taking daily calcium supplementation to maximize the reduction of maternal lead exposure to infants. In the analysis we present a novel method, termed synergistic self-learning (SS-learning), to address two major challenges in the derivation of ITR in the presence of multiple clinical outcomes, including heterogeneous multidimensional outcomes and complex missing data patterns. SS-learning can effectively synergize heterogeneous features of multiple training data sources in the derivation of ITR. Our analysis of the ELEMENT calcium supplementation trial identified several important predictors that can be used to form an ITR that would give a higher expected lead reduction should it be implemented to the whole study population. We also examined the sensitivity and stability of SS-learning by comprehensive simulation studies.