Maternal health exposures and pregnancy outcome: examining symptoms of nausea and vomiting in early pregnancy, maternal caffeine consumption, and spontaneous abortion utilizing regression and propensity score methodologies

This dissertation addressed two important issues concerning pregnant women: caffeine consumption and nausea and vomiting in early pregnancy (NVP). Factors influencing NVP have not been well established and methodological limitations in previous research merit closer evaluation on the caffeine consumption-spontaneous abortion (SAB) and NVP-SAB relationships. This study examined potential risk factors or markers for NVP, and the associations between NVP, caffeine exposures, and SAB using refined classifications of NVP and traditional and novel analytic methodologies. Our analyses included 2,430 newly pregnant women, who participated in a prospective cohort epidemiologic study from 2000 to 2004. Detailed NVP and caffeine consumption data and other health and pregnancy outcome information were collected through interviews, ultrasound assessments, and medical and vital records. Modified Poisson regression with robust error variance models generated risk ratios for potential risk factors to NVP characteristics. Discrete-time continuation ratio logistic survival models were used to estimate week-specific pregnancy loss associated with NVP and caffeine exposures. Additionally, we introduced propensity score methods to evaluate the effects of caffeine exposures on SAB. Most maternal characteristics were not associated with having NVP, except for plurality. Increased risk for delayed onset was found with increasing maternal age, among non-Hispanic Blacks, Hispanics, or age at menarche between 12 and 13 years old. Additionally, older women and non-Hispanic Blacks were less likely to experience symptoms beyond first trimester. Absence of NVP is associated with an increased risk for SAB, compared to having any symptoms. Furthermore, age was found to modify the association between symptom severity and duration and SAB. Reduced risks for SAB was found across all age groups for longer duration, but the effects were much pronounced in the oldest age groups. There was little indication of potential harmful effects of caffeine on SAB; our results showed no overall meaningful differences in estimates from traditional covariate adjustment versus propensity score models. Using improved exposure assessments and analytic methodologies, we identified characteristics that are associated with subclasses of NVP, in addition to providing better understanding in NVP-SAB and caffeine-SAB relationships. Finally, our findings suggest that propensity score methodology is an important addition in studying pregnancy health