Age-Specific Total Fecundability and Fetal Loss in Bangladeshi Women

Demographers and human biologists have long been interested in behavioral and biological components of fertility and how those components change with age. One of the most difficult issues has been the estimation of age-specific fecundability (the monthly or cycle-wise probability of conception) and pregnancy loss. These two components are confounded by an inability to detect pregnancies early in gestation, so that an ovarian cycle in which no pregnancy is observed might result from either no conception or a conception followed by an unrecognized loss of the pregnancy. We developed a parametric model of age-specific pregnancy loss and age-specific fecundability, based on the presumed underlying biology of pregnancy loss. Pregnancy detection is probabilistically determined by the sensitivity and specificity of the pregnancy assay. Simulated data were used to examine the ability of the model to recover parameters for the entire distribution of pregnancy loss. New data were collected in a nine-month prospective study conducted in Matlab, Bangladesh. About 500 female subjects provided twice-weekly interviews and urine samples over 1,561 menstrual cycles. The samples were assayed for human chorionic gonadotropin to detect early pregnancies. Pregnancies were diagnosed for 329 subjects; 151 ended in live birth, 93 ended in fetal loss, and 85 were right-censored. Total fecundability was estimated to be high and constant until age 41, and then declined sharply. The probability of pregnancy loss was about 45% per pregnancy at age 18 and rose to about 92% by age 38. These results suggest that the apparent decline in fecundability through the early 40's results from a dramatic increase in early pregnancy losses rather than a true change in fecundability.