If one were to examine any recent edition of the Physicians' Desk Reference (PDR), virtually all drugs are recommended to be avoided in pregnancy, not because of known problems but because of a lack of information. Yet, data available from several surveys document a remarkably high number of drugs used in pregnancy. Although it is often assumed that only a small proportion of total birth defects are attributable to drugs, this remains to be documented. The proposed study will use COMPASS, an extremely large database (over 8 million subjects) containing Medicaid billing data, to address this question. The study consists of four parts: a descriptive study, a series of screening population-based case-control studies, a series of confirmatory population-based case-control studies, and a series of validation studies. The initial phase of this study will evaluate the patterns of drug use during pregnancy for this Medicaid population. We will evaluate use by state, by age group, by race, and by trimester. The experience of a total of 235,000 pregnancies will be used for this part of the study. The second part of the proposed study is a series of retrospective case-control investigations to screen for possible associations between the use of drugs in pregnancy and birth defects. In this study, cases with birth defects, in general, will be compared to an equal number of randomly selected normal infants, matched for state and maternal age, looking for differences in first trimester in utero drug exposure. Subsequent analyses will then look at specific defects. The third part of the study we will examine the associations which emerge from these screens as well as a series of a priori hypotheses, in more detail, using stratification and logistic regression to control for confounding and search for effect modification. Finally, a series of validation studies will be performed. We will request primary medical records from health care providers through each state's Medicaid program. Information will be sought regarding diagnoses, including complications of pregnancy, drugs, birth weight, parity, gestational age, drug abuse, smoking, and alcohol exposures. This study will enable us to quantify errors due to misclassification and due to under-recording in the Medicaid billing data. We will also be able to estimate the effect of potential confounders not evaluated with COMPASS data.
| Grisso, J A; Carson, J L; Feldman, H I et al. (1997) Epidemiological pitfalls using Medicaid data in reproductive health research. J Matern Fetal Med 6:230-6 |
| Strom, B L; Carson, J L (1990) Use of automated databases for pharmacoepidemiology research. Epidemiol Rev 12:87-107 |
| Strom, B L; Tugwell, P (1990) Pharmacoepidemiology: current status, prospects, and problems. Ann Intern Med 113:179-81 |
