Domperidone for insufficient lactation in England 2002‐2015: A drug utilization study with interrupted time series analysis

Our aim was to describe trends in the prescription of domperidone for insufficient lactation in England, the characteristics of women prescribed it postpartum, and the impact of a 2014 European Medicines Agency (EMA) recommendation to restrict its use due to a potential increased risk of sudden cardiac death associated with its use.


| INTRODUCTION
Domperidone is a dopamine antagonist used for nausea, vomiting, and dyspepsia related to motility disorders. 1 Although not licensed for lactation induction in any country, domperidone is the only medication currently listed by the United Kingdom's National Institute for Health and Care Excellence to address breastfeeding problems 2 and is increasingly being used internationally to induce lactation among women experiencing low milk supply. 3,4 However, a growing number of studies have implicated domperidone in cardiac deaths. 1,[5][6][7][8] In addition, a recent Canadian study reported a potentially increased risk of ventricular arrhythmia associated with postpartum domperidone use (hazard ratio = 2.25; 95% confidence interval [CI], 0.84-6.01), with the increased risk driven by use among women with a history of ventricular arrhythmias. 9 To reduce cardiac risks, regulatory agencies in Canada and the European Union have advised that domperidone doses not exceed 30 mg per day, that domperidone not be used in conjunction with medications that intensify its cardiac effects, and to avoid use among patients with risk factors such as existing heart problems or severe hepatic impairment. 10,11 The primary objective of this study was to describe the characteristics of women prescribed domperidone for insufficient lactation in England in terms of their demographic, obstetric, and medical profile and to describe trends over time as well as the dosage, timing of prescriptions, and coprescription of medications with arrhythmogenic properties. Our secondary objectives were to determine whether prescribing of domperidone for insufficient lactation decreased following a European Medicines Agency (EMA) recommendation to restrict its use and to describe cardiac outcomes and all-cause mortality among women prescribed domperidone postpartum.

| Data source
We constructed a population-based cohort study using data from the Clinical Practice Research Datalink (CPRD) and Hospital Episodes Statistics (HES). The CPRD contains detailed clinical information on persons in the United Kingdom seen at over 700 general practitioner practices and includes routinely collected demographic, behavioral (eg, smoking), and diagnostic information using the Read coding system as well as prescriptions written. 12 Approximately 75% of English CPRD patients are linkable to HES, which contains hospitalization records for National Health Service hospitals in England, including a maternity file with specific information about the birth. Hospital Episodes Statistics contains diagnoses recorded using the International Classification of Diseases tenth revision coding system and procedures recorded using the OPCS Classification of Interventions and Procedures version 4. Both the CPRD and HES undergo ongoing quality checks and have been validated extensively. [13][14][15][16][17] We also linked to the Index of Multiple Deprivation to obtain deprivation scores (a proxy of socioeconomic status) and to the Office for National Statistics vital statistics data.
The study protocol was approved by the Independent Scientific Advisory Committee of the CPRD (protocol number 16_178) and the research ethics board of the Jewish General Hospital (protocol number 16_255), Montreal, Canada.

| Study population
Our cohort consisted of all women aged 15 to 45 years with a delivery between May 1, 2002, and March 31, 2015, who were linkable to HES and had a minimum of 1 year of observation time in the CPRD prior to delivery, in order to ensure accurate assessment of medical and obstetric history. Cohort entry was defined in the HES by the end of a delivery episode, and in the CPRD by Read codes indicating a delivery.
As our study was assessing domperidone use for insufficient lactation, we restricted inclusion to women with livebirths and excluded those with nonlactation indications for domperidone use. Nonlactation indications included a diagnosis of Parkinson's disease or use of antiparkinsonian agents, 18 a recorded diagnosis of gastroparesis, two or more prepregnancy proton pump inhibitor prescriptions in the year prior to conception (indicative of pre-existing gastrointestinal conditions, except where a woman was pregnant in the year prior to conception), or prepregnancy domperidone use other than use in the year following a previous birth. The study included repeat live births by the same woman. Cohort exit was defined by end of follow-up (6 months after the delivery), death, onset of Parkinson's disease or gastroparesis, another birth, end of the study period, or end of registration at the CPRD practice, whichever occurred first.

| Exposure assessment
Domperidone use was defined as having been prescribed domperidone orally in the 6 months following childbirth. Patients were stratified based on a first prescription dose of domperidone of ≤30 and >30 mg/d. 1,5,10,11 A small number of women received more than one prescription for domperidone at the time of their first prescription (ie, on the same day); to provide conservative estimates, we assumed that they started on the lowest dose among prescriptions received.
We further assessed the average dose and postpartum day of first prescription and the average number of domperidone prescriptions given in the 6 months postpartum. Information on duration of prescription was not available. with risk factors such as existing heart problems or severe hepatic impairment or those taking other medications that enhance domperidone's effects or reduce its breakdown. Table 1 describes the sequence of key restrictions and recommendations issued on domperidone in Europe, the United Kingdom, and internationally.

| Covariates
Maternal characteristics of interest were age, body mass index (BMI; <18.5, 18.5-24.9, 25-30, ≥30 kg/m 2 , and missing), smoking status, alcohol-related disorders, and socioeconomic status as measured by the Index of Multiple Deprivation. We further assessed obstetric characteristics such as parity, preterm birth (defined as <37 weeks of gestation), cesarean delivery, multifetal gestation (ie, twin or higher order pregnancies), diabetes or gestational diabetes mellitus, and gestational or pre-existing hypertension. We assessed pre-existing (prior to childbirth) cerebrovascular and cardiac conditions such as circulatory disorders complicating pregnancy, cardiomyopathy, cerebrovascular disease, ventricular tachycardia/fibrillation, ischemic heart disease, heart failure, pulmonary heart disease, valvular heart disease, history of pacemaker/defibrillator use, or use of anti-arrhythmic agents. We assessed whether women experienced renal impairment, asthma, schizophrenia, liver disease, proton pump inhibitor use, and cancer diagnosis prior to childbirth.  (Table S1). 23,24 Coprescribing was defined as receiving a prescription of a potentially arrhythmogenic drug in the 30 days before or after receiving a domperidone prescription. We assessed cardiac outcomes based on recorded diagnoses in CPRD, HES, or Office for National Statistics data.

| Statistical analysis
Standardized differences were used to compare patient characteristics among women prescribed and not prescribed domperidone in the 6 months following childbirth, with a standardized difference of 0.1 considered important. 25  births that occurred in the final 6 months of the study period were excluded to allow adequate follow-up time. For analyses of cardiac outcomes, event-specific follow-up was censored at the time of event, and exposure was defined by ever use in a time-dependent manner. Calculation of the upper limit of the 95% CIs for outcomes with zero events incorporated the rate characteristics of the population. 26 We used an interrupted time series analysis to estimate changes in domperidone prescribing after the EMA recommendations, assessing monthly changes in the trend and level of prescribing in the 6 months postpartum. 27 The interrupted time series used trends before the recommendation (May 2002 to March 2014) to predict the trends had the recommendations not occurred and controlled for baseline trends. 28 Generalized least-squares models that allowed for a first-order autoregressive structure were used to account for correlation between consecutive months. Alternative models with variations in the autocorrelation structure (or no autocorrelation structure) were compared using likelihood ratio tests.
In sensitivity analyses, we modeled the interrupted time series assuming the effect would occur 1 year prior to the recommendation, when the notification of the domperidone safety assessment was published by the EMA. We further modeled the prerecommendation period as starting March 2011 to account for a steeper increase prior to the recommendation. We further tested for seasonality in prescribing based on the hypothesis that insufficient lactation may be more common in the winter months. An additional sensitivity analysis restricted the cohort to first pregnancies to determine whether the analysis would be affected by women who had previously delivered.
Analyses were performed using SAS version 9.4, the R statistical package, and OpenEpi (Version 3.01, http://www.openepi.com).  domperidone were less likely to be aged <25 years, while they were more likely to have BMI ≥30 kg/m 2 or a missing BMI, to have had a preterm births, multiple births, or to have had a cesarean section ( Table 2). They were also more likely to have diabetes or gestational diabetes and to have used proton pump inhibitors prior to childbirth.
Having a history of circulatory or cardiac disorders was very rare among women prescribed domperidone (<5 among women prescribed domperidone, exact number suppressed due to CPRD privacy policy), as was having a history of liver disease (0.35% of domperidone users).

| Prescription rates and prescribing characteristics
The

| Cardiac events
Among women prescribed domperidone in the 6 months postpartum, no ventricular arrhythmias, cardiac arrests, or sudden cardiac deaths were observed (

| Sensitivity analyses
Alternative models with variations in the autocorrelation structure (or no autocorrelation structure) led to similar estimates and       Figure S2). Analyses restricting the cohort to each woman's first delivery resulted in a stronger decrease in prescribing ( Figure S3). No seasonal patterns in prescribing were observed.

| DISCUSSION
In this population-based cohort study, we found a 3. Exposure to domperidone postpartum was low throughout the study period in the United Kingdom, and we observed no cardiac morbidity among women exposed to domperidone postpartum and low use among women with cardiac risk factors. This finding is reassuring, given that in the previous Canadian study, the increased risk of ventricular arrhythmia among domperidone users postpartum was driven by use among women with a prior history of ventricular arrhythmia. 9 Similar to others, we found that domperidone use was associated with a high BMI; having a chronic condition (hypertensive disorder or diabetes in our study); and having had a preterm birth, multiple birth, or cesarean delivery. 3 Although all cause of death was higher among women prescribed domperidone, the outcome was too rare to conclude that the crude risk was truly higher among domperidone users. to 19% for term births and 17% to 32% for preterm births. 3 The comparable rates in England (in the 6 months postpartum per live birth) were much lower, from approximately 0.28% to 1.05% among all deliveries and 1.2% to 5.2% among preterm deliveries between The proportion of preterm births among women prescribed domperidone increased in the study, which supports the possibility that physicians may have been less likely to change practices for women with preterm infants, due to the advantages of breastmilk for these infants.
Our study also has several potential limitations. We only cap-