Antenatal Exposure and Diseases in the Offspring: The Role of Big Data
In the current issue of Drug Safety, Charlton et al. have published the results of a study that evaluated the association between epilepsy and in utero exposure to antiepileptic drugs and estimated the prevalence of neurodevelopmental disorders (NDD) in children born to epileptic mothers . The main focus of the study was to compare the results between two separate data sources rather than to answer the question of whether epilepsy or prenatal exposure to antiepileptic drugs is associated with the risk of NDD since this association has been previously demonstrated. Charlton et al. aimed to determine whether data from the Clinical Practice Research Datalink (CPRD), a large, high-quality UK database that is based on the de-identified electronic records of approximately 15 million patients from general practitioners (GPs) [2, 3], produced similar risk estimates to a prospective UK longitudinal face-to-face study by the ‘Liverpool and Manchester Neurodevelopmental Group’ in which pregnant mothers, with or without epilepsy, were interviewed to assess lifestyle and drug use and were prospectively followed for 6 years to identify NDDs in the offspring . The CPRD has been the source of over 1000 publications over the past 20 years, many of them in the area of drug safety, including studies of potential teratogenic drug effects [5, 6, 7]. A particular strength of this database is that it is possible to link mothers to their babies and thus identify exposures during pregnancy in the mother and health outcomes in the offspring. Based on their analyses, the authors concluded that the sensitivity to detect NDD in the CPRD was slightly lower than in the prospective study. This result is to be expected since the Liverpool and Manchester study was specifically designed to follow women during and after pregnancy and to learn about possible adverse effects of antenatal exposures on the fetus in utero, while the CPRD is a comprehensive medical record system designed to study all manner of health-related questions; it encompasses comprehensive data collection and information on routine daily clinical patient details for hundreds of GP offices all over the UK. While designed differently, the results across the two resources were quite similar. Thus, we should certainly not conclude that the CPRD is not suitable for studies that require linked information on mothers and their babies. On the contrary, the CPRD is a valuable and efficient resource for these important studies, where the data are collected as part of routine GP practice, resulting in a high quality, cost effective, existing source of data on mothers and babies.
A very relevant and important statement was made by the authors in the discussion of the paper, where they attempted to explain why the CPRD may have yielded lower NDD prevalence and risk estimates compared with the prospectively collected Liverpool and Manchester study data. While the authors primarily searched the CPRD data for coded NDD diagnoses, they realized that they were able to capture several additional cases of NDD by sending for original medical records and by investigating free text fields in addition to the coded diagnoses. In other words, important information was available to augment the electronically coded data, in the form of medical documents and GP notes.
The ability to send for original medical records such as hospital discharge or consultant letters afforded access to a rich source of information important for the evaluation of outcomes that require detailed case validation or assessment. This feature made the CPRD particularly valuable to the medical research community. Unfortunately, this practice is no longer possible due to data access restrictions implemented because of fear among the public and some politicians that this database, of crucial value to the UK and global society, may be misused. It is important to clarify that only vetted researchers have access to the anonymized data (a crucial point) and that they must agree to maintain patient confidentiality and to not publish information that might potentially lead to patient identification, thus the risk of confidentiality breeches is extremely low. While there has never been a known case of breach of confidentiality, not even a known attempt, this fear of potential misuse has resulted in increased administrative hurdles and restrictions on the use of data for research purposes to the detriment of the public health. This important loss of information affects the quality of research for investigators, the medical community, and, ultimately, for patients. Research on databases such as the CPRD, for example in the area of drug safety in pregnancy, has led to important advances in medical innovation which by far outweigh the theoretical risk of hypothetical confidentiality breaches. The medical community and politicians must ensure that rich data sources can be used for medical research in the future, and that all relevant and recorded data can be accessed (in anonymized form) by researchers to enable them to conduct important medical research that will benefit patients, doctors, other healthcare providers, industry, and additional stakeholders.
For us, the crucial message from the Charlton et al. study is that high-quality, anonymous medical records must remain available to qualified researchers. Using existing data sources is the best and most efficient way to study the safety of antenatal drug exposures if the data quality is kept at a high standard and access is not unnecessarily restricted. Randomized prospective trials for many drug safety issues, especially in pregnancy, are not ethical and therefore not feasible. Prospective approaches such as the one taken by the initiators of the ‘Liverpool and Manchester Neurodevelopmental Group’ have to be applauded, but they are not feasible in all situations and settings, they are labor intensive and expensive, and they take a long time to produce results. Thus, even though the results of this analysis suggest that the CPRD may not be perfect for studying some outcomes in relation to maternal drug exposures, we are convinced that the CPRD could continue to be a unique source and the gold standard for research on adverse effects of antenatal drug exposure if researchers again had access to all patient data. This is a political decision and not a question of technical feasibility.
Compliance with Ethical Standards
No sources of funding were used to assist in the preparation of this commentary.
Conflict of interest
Christoph R. Meier and Susan S. Jick have no conflicts of interest to declare that are directly relevant to the content of this commentary.
- 1.Charlton RA, McGrogan A, Snowball J, Yates LM, Wood A, et al. Sensitivity of the UK Clinical Practice Research Datalink to detect neurodevelopmental effects of medicine exposure in utero: comparative analysis of an antiepileptic drug-exposed cohort. Drug Saf. doi:10.1007/s40264-017-0506-5.