Volume 87, Issue 10 p. 3649-3658
SYSTEMATIC REVIEW AND META-ANALYSIS
Free Access

Domperidone and the risks of sudden cardiac death and ventricular arrhythmia: A systematic review and meta-analysis of observational studies

Linda B. Ou

Linda B. Ou

Department of Pharmacy, Sunnybrook Health Sciences Centre, Toronto, ON, Canada

Centre of Clinical Epidemiology, Lady Davis Institute, Jewish General Hospital, McGill University, Montreal, QC, Canada

Department of Epidemiology, Biostatistics, and Occupational Health, McGill University, Montreal, QC, Canada

Search for more papers by this author
Carolina Moriello

Carolina Moriello

Centre of Clinical Epidemiology, Lady Davis Institute, Jewish General Hospital, McGill University, Montreal, QC, Canada

Search for more papers by this author
Antonios Douros

Antonios Douros

Centre of Clinical Epidemiology, Lady Davis Institute, Jewish General Hospital, McGill University, Montreal, QC, Canada

Department of Epidemiology, Biostatistics, and Occupational Health, McGill University, Montreal, QC, Canada

Institute of Clinical Pharmacology and Toxicology, Charité - Universitätsmedizin Berlin, Berlin, Germany

Department of Medicine, McGill University, Montreal, QC, Canada

Search for more papers by this author
Kristian B. Filion

Corresponding Author

Kristian B. Filion

Centre of Clinical Epidemiology, Lady Davis Institute, Jewish General Hospital, McGill University, Montreal, QC, Canada

Department of Epidemiology, Biostatistics, and Occupational Health, McGill University, Montreal, QC, Canada

Department of Medicine, McGill University, Montreal, QC, Canada

Correspondence

Kristian B. Filion, PhD, FAHA; Associate Professor and William Dawson Scholar, Departments of Medicine and of Epidemiology, Biostatistics, and Occupational Health, McGill University, 3755 Cote Ste-Catherine Road, Suite H410.1, Montreal, Quebec, Canada.

Email: [email protected]

Search for more papers by this author
First published: 13 January 2021
Citations: 17

L.B.O. was a Master of Science student at McGill University, Montreal, Canada at the time this study was conducted. She is now a clinical pharmacist at Sunnybrook Health Sciences Centre, Toronto, Canada. C.M. is an employee of the Canadian Network for Observational Drug Effects Studies (CNODES), a collaborating centre of the Drug Safety and Effectiveness Network (DSEN), funded by the Canadian Institutes of Health Research (Grant Number DSE-146021). A.D. is supported by a Junior I salary support award from the Fonds de recherche du Québec—Santé (FRQS). K.B.F. is supported by a Senior salary support award from the FRQS and a William Dawson Scholar award from McGill University.

Abstract

Aims

Concerns exist regarding the cardiovascular safety of domperidone. However, many of the previous studies addressing this issue had important limitations. We aimed to examine domperidone and the risks of sudden cardiac death and ventricular arrhythmia through a systematic review and meta-analysis of observational studies, including an in-depth methodological assessment.

Methods

We systematically searched MEDLINE, PubMed, EMBASE, Scopus and CINAHL Plus to identify observational studies examining the association of domperidone and sudden cardiac death and/or ventricular arrhythmia. We assessed study quality in duplicate using the ROBINS-I tool supplemented by an assessment of specific biases and the GRADE (Grading of Recommendations, Assessment, Development and Evaluations) approach. Data were pooled across studies using DerSimonian and Laird random-effects models.

Results

Six case–control studies, 1 case-crossover study and 1 retrospective cohort study were included (n = 480 395). Based on ROBINS-I, 3 studies had moderate risk of bias, 4 had serious risk, and 1 had critical risk. The overall GRADE rating is moderate. When data were pooled across nonoverlapping studies, domperidone was associated with an increased risk of composite endpoint of sudden cardiac death or ventricular arrhythmia compared to nonuse (adjusted odds ratio: 1.69; 95% confidence interval: 1.46, 1.95; I2: 0%; τ2: 0). This association persisted when restricted to higher-quality studies (odds ratio: 1.60; 95% confidence interval: 1.30, 1.97; I2: 0%; τ2: 0).

Conclusion

Domperidone is associated with an increased risk of sudden cardiac death and ventricular arrhythmia compared to nonuse. Further investigation comparing domperidone to an active comparator and in younger populations are warranted.

1 INTRODUCTION

Domperidone, a peripheral dopamine antagonist, is approved for a variety of pharmacological uses, including gastrointestinal motility disorders associated with gastritis and diabetic gastroparesis, as well as nausea and vomiting from the use of anti-Parkinson agents.1, 2 It is also frequently used off-label for postpartum prolactin stimulation.3, 4 Despite its frequent use in many jurisdictions, concerns exist regarding its safety. Since 2012, there have been warnings by multiple regulatory agencies (Health Canada, European Medicines Agency, Medicines and Healthcare Products Regulatory Agency in the UK, and Health Sciences Authority in Singapore) against the use of domperidone in daily doses >30 mg and in patients aged >60 years due to concerns of sudden cardiac death and ventricular arrhythmia.5-8 These recommendations were based on increased risks observed in 2 observational studies.9, 10 The findings may be explained by domperidone's propensity to prolong QT-interval and cardiac repolarization, which can trigger severe forms of ventricular arrhythmia (torsades de pointes) that can result in sudden cardiac death.11, 12 Despite these warnings, drug utilization studies show that there has not been substantial changes in its prescribing in several European countries.13, 14

In recent years, several additional observational studies have investigated the possible association between domperidone and the risk of sudden cardiac death and ventricular arrhythmia.15-18 These studies have provided widely divergent results, ranging from a small increase that did not reach statistical significance to a 4-fold increased risk of sudden cardiac death. Two systematic reviews on this potential adverse drug effect have been published.19, 20 However, they have not been updated with recently published studies,21, 22 and these previous reviews had important methodological limitations, including the use of the Newcastle–Ottawa Scale for quality assessment. This scale has been shown to have several important limitations,23 including high interuser variability24 and includes some concepts related to external validity as part of its quality assessment.25 In addition, an assessment of the overall body of evidence would be optimal to determine the need for future studies in this area. Given the heterogeneity of the existing literature, lack of reliable assessment on the risk of bias, the severity of purported adverse events, and unchanged prescribing patterns, there remains an urgent need to examine this potential association. This systematic review with meta-analysis aims to provide a comprehensive and rigorous assessment of the existing literature of observational studies regarding the association between domperidone and the risks of sudden cardiac death and ventricular arrhythmia. We also aim to explicitly explore the quality of studies in this area and potential heterogeneity in the literature through subgroup analyses.

2 METHODS

A prespecified protocol was followed in conducting this systematic review, and reporting is in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.26

2.1 Search strategy

We systematically searched MEDLINE, PubMed, EMBASE, Scopus and CINAHL Plus from inception to December 2019 to identify observational studies examining the association of domperidone with sudden cardiac death and/or ventricular arrhythmia. Search terms included Medical Subject Heading (MeSH) terms for PubMed, MEDLINE and CINAHL Plus, as well as Emtree terms for Embase. Keywords for domperidone and all types of ventricular arrhythmia and sudden cardiac death were added to the search strategy and used in all 5 databases (Table S1). No restrictions were applied on the type of study or the language of publication. References of previous systematic reviews and relevant articles were manually screened.

2.2 Inclusion and exclusion criteria

Observational studies (cohort, case–control, case–cohort and case-crossover studies) were included. Inclusion was further restricted to studies of adults aged 18 years or older and studies with sample sizes >1000 subjects to ensure adequate power to provide meaningful estimates of the risks of sudden cardiac death and ventricular arrhythmias due to their rare occurrence. Statistical adjustment or matching for age was considered a necessary inclusion criterion as it is a strong, independent risk factor for cardiovascular diseases. Cross-sectional studies, randomized controlled trials, case reports, case series, letters to the editor, commentaries and editorials, previous reviews and meta-analyses, animal studies, and basic science studies were excluded. Finally, unpublished data and conference abstracts were excluded as they usually contain insufficient information for quality assessment and their results can be preliminary.

Titles and abstracts were screened by 2 independent reviewers (L.B.O. and C.M.) and any publication deemed potentially relevant by either reviewer was carried forward to full-text review. Following full-text review, the final list of included studies was determined by the 2 reviewers, with any discrepancies to be resolved by consensus or, if necessary, by a third reviewer (K.B.F.).

The primary outcome of this systematic review was a composite endpoint of sudden cardiac death and ventricular arrhythmia. For studies that only investigated sudden cardiac death, the estimates for sudden cardiac death were included in the primary analysis as they are specific and often occur due to ventricular arrhythmia.27, 28

2.3 Data extraction

Information extracted independently by 2 reviewers included: (i) study characteristics (study design, sample size, patient population, intervention vs. comparator, study outcomes, country of study, data source, and study period); (ii) baseline patient characteristics (age, sex, smoking, history of cardiovascular diseases: cardiomyopathy, heart failure, valvular heart disease, hypertension, dyslipidaemia, myocardial infarction, ischaemic heart disease, arrhythmia, and cerebrovascular disease); (iii) concomitant medication use (QT-prolonging medications and cytochrome P450 3A4 inhibitors); (iv) point estimates and corresponding 95% confidence intervals (CIs) for measure of association (hazard ratio [HR], odds ratio [OR], or rate/risk ratio) with event counts. When the outcome measures were adjusted for or matched on covariates, these variables were recorded. Specific concomitant medications were assessed. Due to domperidone's effect on the QT-interval, other QT-interval prolonging medications could cause additive effects.12 As a CYP 3A4 substrate, inhibition of CYP 3A4 could increase the concentration of domperidone and lead to higher risk of adverse events.1

2.4 Quality assessment

The included studies underwent quality assessment by 2 independent reviewers using the ROBINS-I (Risk of Bias in Non-Randomized Studies - of Interventions) tool.25 In this tool, studies were compared to a perfect target study resembling a randomized controlled trial. The 7 domains of bias assessed include bias due to confounding, bias in selection of participants, bias in classification of interventions, bias due to deviations from intended interventions, bias due to missing data, bias in measurement of outcomes, and bias in the selection of reported results. Signalling questions helped guide assessment for each domain where the risk of bias was assessed as low, moderate, serious, critical or no information for each study.

Our ROBINS-I assessment was supplemented by the detailed assessment of specific biases, including prevalent user bias, persistent user bias (case-crossover studies only), time-window bias, misclassification of exposure, and residual confounding. Prevalent user bias is a selection bias in which chronic users were included in the study.29 This results in the depletion of susceptible subjects as patients who had the event during the initial exposure period will not survive to be included in the study. In case-crossover studies, persistent user bias occurs when nontransient exposures are assessed and usually biases the estimates upwards due to long-term medication use and right truncation.30 Time-window bias occurs due to differential duration of follow-up for cases and controls resulting in differential opportunity for exposure (i.e., the longer the follow-up time, the higher the probability of exposure).31 Misclassification of exposures can be either differential or nondifferential, depending on whether the probability of misclassification is different or comparable in the different groups. Lastly, residual confounding arises due to unmeasured or unknown confounding, inappropriate or lack of adjustment of certain confounding variables, or the absence of adjustment for time-varying confounding.

The body of evidence was then assessed using the GRADE (Grading of Recommendations, Assessment, Development and Evaluations) framework. GRADE rates the quality of evidence as very low, low, moderate or high for any particular outcome. It can be influenced by a number of factors such as bias, imprecision, inconsistency, indirectness, magnitude of effect, dose–response relationship and confounding.32

2.5 Statistical analysis

Meta-analysis of adjusted point estimates of studies with nonoverlapping databases was conducted using the DerSimonian–Laird random-effects models with inverse variance weighting. In subsequent analyses, studies were stratified by the overall risk of bias, dose, duration of use, and age. Heterogeneity of studies was assessed using I2 statistic to estimate the variation between studies that was due to heterogeneity as opposed to chance and the τ2, the between-study variance in the random-effects model. Publication bias was not assessed due to the limited number of included studies. An influence analysis was conducted using leave-one-out method where estimates were pooled omitting 1 study at a time. All statistical analyses were conducted using R version 3.4.3.

3 RESULTS

3.1 Search results

Our search yielded 914 potentially relevant articles after duplicates were removed (Figure 1). Following title/abstract screening, 16 full-text articles were retrieved and assessed for eligibility. Finally, 8 articles were included in the systematic review (6 case–control studies,9, 10, 15-17, 21 1 case-crossover study18 and 1 retrospective cohort study22).

Details are in the caption following the image
PRISMA flow diagram of studies identified on domperidone and the risks of sudden cardiac death and ventricular arrhythmia

3.2 Study and patient characteristics

A total of 480 395 subjects (case–control studies, 10 900 cases and 123 788 controls; case-crossover study, 25 356 subjects; cohort study, 320 351 subjects) were included in this systematic review (Table 1). Two studies15, 16 investigated the risk of sudden cardiac death with various QT-prolonging medications including domperidone, while the other 6 studies9, 10, 17, 18, 21, 22 focused solely on domperidone. Study populations consisted of mostly older subjects with multiple comorbidities. Two studies21, 22 restricted inclusion to Parkinson's patients and postpartum women, respectively. In the included case–control studies, exposures were assessed based on the index date (date of event) for current and past exposure, although the duration of days used to assess exposure varied between studies. Study periods ranged from 1990 to 2012, and outcomes included sudden cardiac death or the composite outcome of sudden cardiac death/ventricular arrhythmia.

TABLE 1. Characteristics of observational studies examining association of domperidone with sudden cardiac death and/or ventricular arrhythmia
Study Study design Study sample size Patient population Exposure vs. reference group Study outcome(s) Study location Data source Study period
Straus15 2005 Population based case–control n = 775 cases, 6297 controls

Subjects 18 y and older

Age: 71 y (cases), 69 y (controls)

Sex: 40% females

Domperidone vs. nonuse Sudden cardiac death Netherlands Integrated primary care information database 1995–2003
Jolly16 2009 Population based case–control n = 1010 cases, 3030 controls

Adults aged 20–85 y who had died in the community

Age: 67.6 y

Sex: 32.6% female

Domperidone vs. nonuse Sudden cardiac death UK Public health mortality files 2003–2007
Van Noord9 2010 Population based case–control n = 1366 cases, 14 114 controls

Patients 18 y and older

Age: sudden cardiac death: 72.5 ± 14.1 y (cases), 66.3 ± 13.9 y (controls)

Sex: 40% female

Domperidone vs. nonuse Sudden cardiac death, nonfatal ventricular arrhythmia Netherlands Integrated primary care information database 1996–2007
Johannes10 2010 Nested case–control n = 1608 cases, 6428 controls

Eligible patients enrolled in health database

Age: 79.4 y

Sex: 53% female

Domperidone vs. nonuse, PPI Serious ventricular arrhythmia/sudden cardiac death Canada Saskatchewan health database 1990–2005
Arana17 2015 Nested case–control n = 3239 cases, 12 572 controls

Subjects at least 2 y of age

Age: 55 ± 19 y

Sex: 56.9% female

Domperidone vs. nonuse, metoclopramide, PPI Sudden cardiac death UK Clinical practice research datalink 2005–2011
Chen18 2015 Case-crossover a n = 25 356

Patients 18 y or older

Age: 61 ± 19 y

Sex: Unknown

Domperidone vs. nonuse, metoclopramide, PPI Ventricular arrhythmia, sudden cardiac death Taiwan Taiwan's longitudinal health insurance database (LHID) 2000–2011
Renoux21 2016 Nested case–control n = 2902 cases, 81 347 controls

Patients aged 50 or older with first diagnosis of Parkinson's disease or first prescription for anti-Parkinson agent during study period

Age: 74.4 y

Sex: 52.6% female

Domperidone vs. nonuse Ventricular tachyarrhythmia/sudden cardiac death Canada, UK Canadian provincial database, clinical practice research datalink 1990–2012
Smolina22 2016 Retrospective cohort n = 320 351 (21 events)

All women with live birth

Age: Majority from 25–39 y

Sex: 100% female

Domperidone vs. nonuse Hospitalization for ventricular arrhythmia or cardiac arrest Canada British Columbia health database 2002–2011
  • PPI = proton-pump inhibitor.
  • a Case period: 1–30 days before ventricular arrhythmia, control period: 91–120 days before ventricular arrhythmia.

3.3 Risk of sudden cardiac death/ventricular arrhythmia

Domperidone was either compared to nonuse or to an active comparator (proton-pump inhibitors and/or metoclopramide). Five studies9, 16, 17, 21, 22 reported treatment effects that suggested an increased risk that did not reach statistical significance (Table S3). The reported harm ranged from an adjusted OR of 1.22 (95% CI: 0.99, 1.50) to 3.80 (95% CI: 1.50, 9.70). Straus et al.15 and Renoux et al.21 were excluded from the meta-analyses as their data overlapped with those of other included studies. Following meta-analysis (Figure 2), domperidone was associated with an increased risk of sudden cardiac death and ventricular arrhythmia (adjusted OR: 1.69; 95% CI 1.46, 1.95; I2: 0%; τ2: 0). Two studies10, 18 provided 87.9% of the weight of the overall analysis. Figure 3 shows that this association persisted when analyses were restricted to higher quality studies (adjusted OR: 1.60; 95% CI: 1.30, 1.97; I2: 0%; τ2: 0). Among the 3 studies10, 17, 18 that compared domperidone to proton-pump inhibitors, 2 studies10, 18 found an increased risk while the other17 showed no difference (Table S4); the adjusted OR ranged from 1.26 to 1.83. When domperidone was compared to metoclopramide for the outcome of sudden cardiac death, domperidone was associated with a decreased risk (OR: 0.40; 95% CI: 0.17, 0.94).17

Details are in the caption following the image
Forest plot of the association of domperidone and the risks of sudden cardiac death and ventricular arrhythmia. NA: nonapplicable
Details are in the caption following the image
Forest plot of the association of domperidone and the risks of sudden cardiac death and ventricular arrhythmia stratified by overall study quality based on the ROBINS-I tool. NA: nonapplicable

3.4 Subgroup analyses

3.4.1 Dose and duration of use

Due to differences in the cut-offs for dose and duration of use of domperidone in the studies, pooled analysis was only conducted for domperidone dose >30 mg (Table S5). The risk appears to be higher with >30 mg compared to nonuse (adjusted OR: 3.32; 95% CI: 1.38, 7.96; I2: 32%; τ2: 0.22). Only 2 studies explored a duration–response association. One study17 showed that there were higher odds of sudden cardiac death in patients exposed to domperidone for <16 days compared to ≥16 days. In contrast, another study21 showed no difference in harm based on the duration of exposure (Table S6).

3.4.2 Age

Two studies10, 17 conducted subgroup analyses stratified by age, and 1 study22 only included postpartum women. When the study population was stratified by age, subjects aged >60 years had adjusted ORs of 1.64 (95% CI: 1.31, 2.05) in 1 study and 1.65 (95% CI: 0.89, 3.07) in another when domperidone was compared to nonuse. The evidence was less clear in patients aged ≤60 years due to limited number of exposed events (Table S7).

3.5 Sensitivity analysis

In our influence analysis, there was no evidence that any 1 study had a substantial impact on the overall results (Figure S1).

3.6 Overall quality assessment

The overall quality assessment for each study was determined by the domain with the highest bias rating in ROBINS-I. Three studies had a moderate risk of bias, four studies had serious risk and one study had critical risk (Table 2). Four studies had the rating of serious in the confounding domain due to lack of adjustment of 1 or more important covariates, most often cardiomyopathy, valvular heart disease, or concomitant medications (Table S8). Four studies9, 15, 16, 22 had the rating of serious in the selection of participants domain due to presence of prevalent user bias. One study (the case-crossover study)18 had the rating of critical due to the presence of persistent user bias and substantial differences in baseline characteristics. The body of evidence was rated to be moderate by the GRADE framework.

TABLE 2. Quality assessment using the ROBINS-I tool, supplemented by the assessment of specific biases present in studies of the association between domperidone and the risk of sudden cardiac death
Study ROBINS-I rating Prevalent user bias Time-window bias Persistent user bias a Misclassification of exposure Important residual confounding
Straus 2005 Serious x x x x
Jolly 2009 Serious x x x x
van Noord 2010 Serious x x x x
Johannes 2010 Moderate
Arana 2015 Moderate x
Chen 2015 Critical x x
Renoux 2016 Moderate
Smolina 2016 Serious x x x
  • a Present in case-crossover studies.

3.7 Prevalent user bias

The inclusion of prevalent users was observed in multiple studies9, 15, 16, 18, 22 where inclusion was not restricted to new users (Table 2). Only 3 studies10, 17, 21 restricted their study population to new users with cohort entry as first exposure to any study drug to avoid depletion of susceptibles.33 This bias may underestimate the risk of outcome as the subjects potentially at the greatest risk may have been excluded.

3.8 Persistent-user bias

Domperidone can be taken as needed for symptomatic relief of gastrointestinal disorders, but in most cases, patients use it on a regular basis. Therefore, its use is often not transient, and it may not be suitable for study using the cross-over design, where only discordant exposure pairs are included for the analysis. When medications are taken persistently, the most likely scenario involving discordant pairs is where case periods are exposed and control periods are unexposed due to underlying patterns of drug utilization. The opposite scenario of case period as unexposed and control period as exposed is unlikely.30 This would bias the estimates upwards, which is consistent with the findings of Chen et al.,18 where there was a consistent harmful effect when domperidone was compared to nonuse.

3.9 Time-window bias

Time-window bias could be present in 3 population-based case–control studies.9, 15, 16 Since these case–control studies did not match on follow-up time when selecting controls, it is possible that follow-up time was not equal among cases and controls. Matching on duration of follow-up is necessary to ensure that cases and controls have the same opportunity for exposure.

3.10 Misclassification of exposure

Nondifferential misclassification of exposure likely occurred in the studies that used European databases, as domperidone is available over the counter in many parts of Europe. In 2014, domperidone was switched to prescription only in the UK.34 Due to the nature of the databases used, information on over-the-counter medication use was probably not available. This would potentially bias the estimates towards the null, resulting in an underestimated risk. Misclassification also occurred in Smolina et al.22 when subjects were considered exposed until 30 days after the end of the prescription. The grace period may be too long given the purported mechanism.

3.11 Residual confounding

Residual confounding is present in all observational studies due to their nonrandomized treatment allocation. With nonuse as the comparison group for most included studies,9, 15, 16, 21, 22 confounding by indication is likely as those prescribed domperidone may be intrinsically different from nonusers. Although studies adjusted for some covariates to address this issue, unmeasured or unknown confounding remains a potential limitation. The use of active comparator could attenuate this type of confounding.35 However, confounding by indication or contraindication remains possible; for example, the protective association observed when comparing domperidone vs. metoclopramide in the study by Arana et al.17 may be the result of patients with risk factors for ventricular arrhythmias being preferentially prescribed metoclopramide given the safety concerns associated with domperidone. Important confounding variables requiring adjustment to warrant a bias assessment of moderate were age, sex, cardiomyopathy, heart failure, history of myocardial infarction, history of arrhythmia, diabetes mellitus, valvular heart disease and concomitant medications.36, 37 There were also concerns when variables were measured until the index date in case–control studies (date of event) as oppose to at baseline as this could result in adjustment for the consequence of exposure. Time-varying confounding was also not considered in any of the included studies.

4 DISCUSSION

Our systematic review was designed to synthesize the available evidence regarding the real-world effect of domperidone on the risks of sudden cardiac death and ventricular arrhythmia. We identified 8 observational studies of varying quality that suggest domperidone is associated with a 60% increased risk of sudden cardiac death and ventricular arrhythmia compared to nonuse. This risk was especially evident with higher doses and in elderly individuals. There is, however, a lack of evidence in patients younger than 60 years due to the small number of exposed events in the studies. The results were conflicting when domperidone was compared to active comparators. Our detailed analysis of specific biases showed that 5 studies included prevalent users, 3 studies did not consider equal follow-up time for cases and controls, 5 studies could be affected by exposure misclassification, and 4 studies could have important residual confounding.

Two previous systematic reviews19, 20 published on this topic had important limitations as study quality was assessed using the outdated Newcastle–Ottawa quality assessment scale for evaluation of nonrandomized studies. Reliability between reviewers assessing the quality of the studies has shown to be a limitation of the Newcastle–Ottawa quality assessment scale due to lack of clear instructions and ambiguity of undefined factors.23, 24 In our review, we used the more contemporary ROBINS-I quality assessment tool for each individual study and thorough assessment of specific biases in addition to the GRADE framework to assess the overall body of evidence. ROBINS-I offers a more comprehensive assessment of different biases and includes extensive instructions with signalling questions to minimize interuser variability. Furthermore, we included 2 recently published observational studies. Our updated review and utilization of ROBINS-I allowed us to conduct pooled analyses stratified by study quality, which showed that some studies with higher risk of bias may overestimate the risk of sudden cardiac death and ventricular arrhythmia compared to those studies rated moderate. These results demonstrated that there is a consistent increased risk of sudden cardiac death and ventricular arrhythmia associated with domperidone compared to nonuse, but the effect may not be as large as some studies suggest. Moreover, we conducted multiple subgroup analyses not present in previous reviews, including analyses stratifying by active comparators, domperidone dose, duration of use and age.

In some jurisdictions, alternatives to domperidone such as proton-pump inhibitors or metoclopramide for diabetic gastroparesis are available.38 However, prescriptions for domperidone in special populations remain prominent given the lack of therapeutic alternatives. In Parkinson's patients, domperidone remains the first choice for nausea and vomiting as an antiemetic and prokinetic agent.39 The use of metoclopramide is not recommended as it could lead to drug-induced parkinsonism and increased severity of disease.40 As part of the Canadian Network for Observational Drug Effect Studies (CNODES), Renoux et al.21 demonstrated that a clinically significant increased risk remains possible in this patient population, although the results did not reach statistical significance. Similarly, domperidone was prescribed to 1 in 5 Canadian postpartum women for breastfeeding in 2011, often in doses >30 mg.41 A systematic review of randomized controlled trials on domperidone for breast milk production concluded that there were no reported cases of prolonged QT syndrome or sudden cardiac death with short-term benefit in breast milk volume.4 However, the sample sizes of included trials are often too small to detect rare adverse events. The only observational study we identified in postpartum women22 did not identify a definitive association and had important limitations, including a small number of events, exposure and outcome misclassification, and important residual confounding. Risks in this population remain uncertain as patients are usually young with minimal comorbidities and comedication use. Through our subgroup analyses, we identified that risks may be increased with doses >30 mg. Nonetheless, the effect from duration of exposure is less clear.

Our systematic review has several important strengths. To our knowledge, this is the first review to thoroughly assess study quality and its impact on overall results. Second, our search strategy was broad with no restrictions on the type and language of publication. Third, we defined our inclusion criteria to only include quality studies with adequate power to detect the outcome. Fourth, we followed a prespecified protocol.

Our study also has some potential limitations. We did not conduct a search of the grey literature or include any unpublished results in our review as we felt the results would be incomplete. Due to the scarcity of high-quality evidence, inclusion was restricted to observational studies; while these studies provide key information regarding real-world effects, confounding by indication and by other variables remains possible. Although ventricular arrhythmia and sudden cardiac death often occur with no prodromal symptoms, protopathic bias cannot be ruled out. As is true with all knowledge syntheses, our study may have been affected by publication bias. ROBINS-I is designed for target trials and may not perform optimally in detecting the extent of bias in case–control studies; however, it is still the most comprehensive and rigorous risk assessment tool for nonrandomized studies. We also did not include any studies conducted in the paediatric populations as the focus was on the adult population. Lastly, we treated estimates of OR and HR interchangeably and pooled them together. In nested case–control studies with risk-set sampling, the OR provides an unbiased estimator of the hazard ratio.42 In addition, in traditional case–control studies, the OR estimates relative risk under the rare disease assumption. With HRs estimating relative risks under the assumption of constant hazards, we are confident this assumption did not considerably affect our results.

5 CONCLUSIONS

Domperidone is associated with an increased risk of sudden cardiac death and ventricular arrhythmia compared to nonuse. The strength of evidence was stronger in older patients using daily doses >30 mg. While many of the included studies had important methodological limitations, similar results were observed when restricting to higher quality studies. Most of the included studies were conducted in older individuals with large comorbidity burden. While 1 recent study was identified in postpartum women, the amount of evidence that is available for this population remains limited.

NOMENCLATURE OF TARGETS AND LIGANDS

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, and are permanently archived in the Concise Guide to PHARMACOLOGY 2019/20 (Alexander et al., 2019 a,b).43-45

ACKNOWLEDGEMENTS

Dr Douros holds a Chercheur Boursier Junior I award from the Fonds de recherche du Québec–Santé (FRQS; Quebec Foundation for Health Research). Dr Filion holds a Chercheur Boursier Senior award from the FRQS and a William Dawson Scholar award from McGill University. Mrs Moriello is an employee of the Canadian Network for Observational Drug Effects Studies (CNODES), a collaborating centre of the Drug Safety and Effectiveness Network (DSEN), funded by the Canadian Institutes of Health Research (Grant Number DSE-146021). The authors received no specific funding associated with this work.

    COMPETING INTERESTS

    All authors declare no conflicts of interest.

    DATA AVAILABILITY STATEMENT

    The data are available in the article and in its online supplementary material. No additional data are available.