2016 Theses Doctoral

The natural history of pregnancy loss

Sapra, Katherine Jane

Pregnancy loss, the demise of a pregnancy at any time between implantation and delivery, is a common event in women’s lives, affecting approximately one in three pregnancies. Pregnancy loss often causes profound psychological distress to women, their partners, and their families. However, despite its frequency and troubling nature, relatively little is known about the natural history of pregnancy loss, especially the multitude of signs and symptoms that precede a loss and distinguish it from an ongoing healthy pregnancy. One of the challenges in describing the natural history of pregnancy loss is that most losses occur very early, before entry to clinical care, necessitating the use of preconception cohort studies. Few such studies have ever been conducted worldwide. This dissertation aimed to describe the natural history of early pregnancy loss at <20 weeks gestation for the first time using a unique preconception cohort with daily prospective follow-up from the start of the pregnancy attempt through seven weeks post-conception.
To accomplish this goal, three specific aims were undertaken. First, a systematic literature review was conducted to synthesize the existing literature on the relationships between the signs and symptoms and pregnancy loss. Two analytic aims were then undertaken to delineate thoroughly the relationships between prospectively ascertained signs and symptoms—namely, vaginal bleeding, lower abdominal cramping, nausea and vomiting (hereafter referred to as “signs and symptoms”)—and subsequent early pregnancy loss. The first analytic aim used a fixed covariate and fixed effect survival analytic approach to estimate the cumulative incidence of early pregnancy loss by the presence of individual, combinations, and patterns of signs and symptoms and the associations between signs and symptoms and the cumulative incidence of pregnancy loss. The second analytic aim used a time-varying covariate and time-varying effect survival analytic approach to estimate the weekly associations between signs and symptoms and pregnancy loss to determine if these relationships were consistent or divergent across gestational ages. The results of the first and second analytic aims were then compared to gain a more complete understanding of the natural history of early pregnancy loss.
The literature review revealed a dearth of studies on the signs and symptoms of pregnancy loss. Two preconception and 16 pregnancy cohort studies were identified. The literature suggested that vaginal bleeding, particularly heavy vaginal bleeding, was associated with an increased risk of pregnancy loss while vomiting, and in some studies nausea, was associated with a decreased risk of pregnancy loss. However, reliance on care-seeking cohorts, maternal retrospective reports of signs and symptoms after pregnancy loss, and retrospective recall of signs and symptoms over long periods (e.g., entire trimesters) may have biased the observed associations between signs and symptoms and pregnancy loss leading to incorrect inferences regarding the relationships between signs and symptoms and pregnancy loss.
The two analytic aims addressed the data gaps identified in the literature review. The preconception cohort design with prospective daily follow-up from the beginning of the pregnancy attempt facilitated the ascertainment of pregnancies at the earliest stages of gestation and losses prior to clinical care entry through the use of urine-based home pregnancy testing. The daily reporting of multiple signs and symptoms in the first five weeks after a positive home pregnancy test, or approximately two to seven weeks post-conception, allowed for a full description of the relationships between signs and symptoms of pregnancy loss without recall bias.
Data for the two analytic aims come from the Longitudinal Investigation of Fertility and the Environment (LIFE) Study, a population-based cohort with preconception recruitment of couples in 16 counties in Michigan and Texas followed for 12 months of trying for pregnancy and then through pregnancy loss or delivery for couples achieving an hCG pregnancy. 501 couples entered the study, and 347 achieved a pregnancy during the study period. Three hundred forty-one singleton pregnancies comprise the study population for the two analytic aims in this dissertation. Overall, 95 (28%) pregnancies in the study population ended in a pregnancy loss. Lower abdominal cramping, nausea, and vomiting were often reported during the early pregnancy period; vaginal bleeding was less common. The results of the fixed covariate fixed effect survival analysis from the first analytic aim demonstrated that vaginal bleeding, particularly heavy bleeding and bleeding accompanied by lower abdominal cramping, was associated with an increased risk of pregnancy loss. In contrast, the presence of vomiting, but not nausea alone, during the early pregnancy period was associated with a lower risk of loss. Analyses in the second analytic aim using weekly time-varying covariates and time-varying effects of signs and symptoms on pregnancy loss revealed some new findings. The first week after a positive pregnancy test appeared to be a vulnerable period. Vaginal bleeding and lower abdominal cramping were associated with an increased risk of loss in the first week but not in later weeks; conversely, nausea and/or vomiting were associated with lower risk of pregnancy loss but only after the first week.
The observed weekly variations in the signs and symptoms of pregnancy loss may reflect changes in maternal adaptation to pregnancy across gestation. Overall, relatively little is known about the biological processes underlying healthy and unhealthy adaption to pregnancy as well as how embryo quality may affect these adaptive processes. More work is required from basic scientists, clinicians and epidemiologists to better understand the causes of signs and symptoms and their relationships to pregnancy loss, including genetic and environmental factors and their interactions. In the meantime, prognostic models developed from data in this dissertation using time-varying signs and symptoms may be useful to women and their health care providers for identifying pregnancies at increased risk for pregnancy loss. These models could prompt women to seek medical care when concerning patterns of signs and symptoms arise.