Medicines prescribed in pregnancy: assessing safety for mother and baby using routine NHS data

Research question Which medications prescribed in primary care are the most likely to be harmful to mother or baby when prescribed in pregnancy? Background Medication use in pregnancy is common and increasing.

Yet the safety to mother and baby of many medications used in pregnancy is unknown because pregnant women are often excluded from drug trials.

Routine health data have been used to look for evidence of harm and safety of specific medications, but not to systematically look across all prescribed medications in pregnancy.

Aims and Objectives Our aim is to systematically identify medications prescribed during pregnancy requiring further research into their potential harms to mother or baby.

The objectives are: To investigate trends in medications prescribed in pregnancy in primary care from 2003 to 2023, and how this varies by socio-economic status, ethnicity, age, co-morbidities and region.

To use Bayesian statistical signal detection methods on all medications prescribed in pregnancy in primary care to identify medications that are potentially harmful to the mother, fetus or infant.

To combine the results from 1 and 2 with published scientific evidence of harm or safety during pregnancy and to prioritise those medications requiring further research into their potential harms.

To perform cohort studies within the same dataset for three of the medications identified above to extend their safety evaluation by including detailed adjustment for potential confounders.

Methods We will analyse factors influencing the prevalence of prescriptions issued to pregnant women using data from pregnancies in the Clinical Practice Research Datalink (CPRD) from 2003 to 2023.

We will identify, clean and categorise maternal, fetal, and infant adverse outcomes reported in the CPRD which includes general practice and linked hospital data for mother and baby pairs.

Bayesian signal detection models will analyse all medication/outcome pairs to identify pairs where an outcome occurs significantly more frequently in mothers exposed to a specific medication than in mothers not exposed to it. Potential confounders will not be accounted for at this stage. Published evidence on the identified medication/outcome pairs will be reviewed.

For medications with insufficient data on safety, results from objective 1 and 2 including information on medications more commonly taken by underserved groups will be used by PPI and Study Advisory Board workshops to prioritise medications for further research.

For three medications cohort analyses will compare the occurrence of adverse outcomes in pregnancies exposed to the medication of interest to those in pregnancies not exposed to the medication of interest.

Maternal morbidity, co-medications and other confounders will be adjusted for using multivariable logistic regression with propensity scores, Lasso regression and multiple imputation.

Anticipated Impact and Dissemination Results will be discussed with UK Teratology Information Service (who will sit on our advisory board) and relevant findings will be added to bumps (best use of medicines in pregnancy) resources, which are widely used by clinicians and pregnant women.

This project will improve informed decision making for patients and clinicians to enhance the health and wellbeing of pregnant women and their babies.