Background: Use of pregnancy registries is a common method of postmarketing surveillance of pregnancy outcomes to identify potential teratogens. However, with the increase in electronic capture of healthcare data for administrative, audit and research purposes, data generated during routine clinical practice might be used to address questions similar to those explored using pregnancy registries.
Objectives: To establish how data from the UK General Practice Research Database (GPRD) compares with data from the UK Epilepsy and Pregnancy Register and to assess how it can contribute to postmarketing surveillance of pregnancy outcomes.
Methods: Pregnancy outcomes were identified from the GPRD for women aged 14-49 years with a diagnosis of epilepsy and supporting evidence. Outcomes with a major congenital malformation (MCM) were identified and the relative risks (RRs) of an MCM following a range of first-trimester antiepileptic drug (AED) exposures were calculated and compared with those reported by the UK Epilepsy and Pregnancy Register. In addition, we also evaluated whether the known association between valproate and spina bifida could be identified using data from the GPRD. The study period ran from 1 January 1990 until 31 December 2006.
Results: A total of 1766 live mother-baby pairs were identified, as well as 5:51 pregnancy terminations, 13 stillbirths and 1 neonatal death. Including those that resulted in a termination, there were 62 unique pregnancy outcomes with an MCM. An increased risk of spina bifida was identified using the GPRD following first-trimester monotherapy exposure to valproate when compared with those with no AED exposure (RR 8.02; 95% CI 1.5, 43.5). More generally, comparing the GPRD with the UK register, the GPRD ascertained a lower number of first-trimester AED exposures: monotherapy 711 versus 2468; polytherapy 156 versus 718. We reproduced the UK register results of an increased MCM risk following first-trimester polytherapy AED exposure compared with no AED exposure (RR 2.89; 95% CI 1.43, 5.84). Using the GPRD, we identified similar point estimates to the UK register following monotherapy and polytherapy exposures (4.1% vs 3.7% and 7.1% vs 6.0%, respectively) but we were unable to reproduce the level of statistical significance. For individual AEDs, the MCM rate following valproate exposure was 4.9% (11/225) in the GPRD compared with 6.2% (44/715) in the UK register.
Conclusions: The GPRD has potential for the identification of malformations and of a teratogenic association. For epilepsy, the GPRD does, however, identify fewer exposed pregnancies than a pregnancy registry. Therefore, in many circumstances pregnancy registries are likely to remain preferable as a method of surveillance. The GPRD may be better suited to monitoring medicines used in the treatment of more prevalent conditions, such as depression, or for monitoring medicines that have been on the market for a long time and for which no registry has been set up.