Objective Serotonergic appetite suppressants and ergot-derived dopamine agonists have been associated with drug-induced valvular heart disease. The purpose of this meta-analysis is to synthesise the current evidence of a link between several medications affecting sertonergic pathways and valvular heart disease.
Methods PubMed was searched to identify studies evaluating an association between medications with serotonergic activity and cardiac valvular pathology. Case reports, uncontrolled studies and in vitro studies were excluded. Relevant studies were assessed for quality and potential bias; those of adequate quality were included in a quantitative synthesis. Sensitivity analyses were conducted, and potential publication bias was examined.
Results There was a consistent, significant relationship between certain medications and heart valve disease, including serotonergic medications (OR 3.30, 95% CI 1.99 to 5.49) and dopaminergic medications (OR 2.56, 95% CI 1.68 to 3.91). Subanalyses, including analyses that limited exposure to a single medication or effects to a single heart valve were also consistently significant. Most studies were retrospective or observational in nature, with a higher risk of selection and presentation biases. There was significant heterogeneity and variability between studies, particularly when it came to dose and duration of exposure.
Conclusions There was a consistent, significant association between many medications that affect serotonergic pathways and valvular heart disease. Although many of these medications have been withdrawn from the market, some small studies suggest that recreational drug 3,4-methylenedioxymethamphetamine and widely prescribed selective serotonin reuptake inhibitors may affect similar pathways.
- valvular heart disease
Statistics from Altmetric.com
Valvular heart diseases (VHDs) are multifactorial disorders that disrupt the unidirectional blood flow during the cardiac cycle. VHDs may arise from acquired conditions or genetic predispositions and are the result of either pathological remodelling of the valve tissues or secondary consequences of a change in the anatomy and geometry of the cardiac structures.
There is a well-established association between certain medications and heart valve pathologies. One of the most prominent examples of this drug-induced valvular heart disease (DI-VHD) is the incidence of valvular insufficiency among patients taking serotonergic appetite suppressants such as fenfluramine-phentermine and benfluorex.1 A subset of the patients prescribed these medications for weight loss developed a pathological thickening and retraction of their cardiac valve leaflets, rendering the valves insufficient.2 Extensive reports of the correlation between these appetite suppressants and DI-VHD led to their eventual withdrawal from the market in the late 1990s. This was not the first time that medications were associated with cardiac valve pathologies; the early migraine treatments ergotamine and methylsergide had also been implicated in the fibrotic changes to the valves as early as the mid-1970s.3 A similar association between ergot-derived dopamine agonists (EDDAs) used to treat Parkinson’s disease (PD) and hyperprolactinaemia (HPL) and DI-VHD was also observed in the 2000s.4
The exact mechanism through which these medications induce pathological changes of the cardiac valves is not fully understood. Surgeons and pathologists observed, however, that there are similarities between valves affected by DI-VHD and valves affected by carcinoid syndromes.5 6 In patients with carcinoid syndromes, neuroendocrine tumours, primarily in the GI system, secrete large amounts of serotonin (5HT), resulting in increased plasma serotonin levels.7 For both carcinoid syndromes and DI-VHD, the valves have thickened cusps, few calcifications, a proliferation of extracellular matrix with glycosaminoglycans and collagen and frequently observed carcinoid plaque of dense, fibrotic tissue.2 In the case of mitral and tricuspid valves, there is also a shortening, thickening or occasional fusion of the chordae tendinae.6 These similarities led to a hypothesis that the transformation from a physiologically normal valve to one afflicted with DI-VHD may be initiated or modulated by a serotonergic agent.8
Although there have been previous meta-analyses looking at certain medications and their association with VHD, there has not been a broad analysis that includes, compares and contrasts multiple serotonergic medications and their association with VHD. In order to better understand the role of 5HT modulation in DI-HVD, we conducted a meta-analysis evaluating the clinical evidence of a relationship between DI-VHD and drugs that are either known to cause DI-VHD or those that are involved in either 5HT receptor 2B (5HTR2B) signalling or inhibition of the serotonin transporter.
We completed a comprehensive search of PubMed to identify articles relevant to DI-VHD. Medications were included if they had an established history of inducing DI-VHD; if there was hypothesised 5HTR2B agonist activity identified by assay9; or if the medications had known interactions with 5HT transporters.10 The search terms used are available in online supplemental table 1. An initial screening of titles was used to remove duplicate and irrelevant entries. Abstracts were reviewed to further refine the search results and identify articles that met the criteria for a full-text review. Articles that were human clinical studies reporting on DI-VHD were included, while the following were excluded: case report; small case series without control groups; or animal or in vitro studies. A hand search of the references of relevant review articles and meta-analyses was conducted to identify additional references.
Supplementary file 1
Data extraction and quality assessment
The full text of the remaining articles was independently reviewed for quality by two of the authors (JHF and BP) using the Cochrane Review’s Tool to Assess Risk of Bias in Cohort Studies.11 Each article was rated on a scale from 1 (high risk of bias) to 4 (low risk of bias) for eight criteria (online supplemental table 2). Under the category ‘Was selection of exposed and non-exposed cohorts drawn from the same population?’, examples of lower risk of bias would be patients from the same PD clinic during the same time period who were prescribed an EDDA or a non-EDDA medication. An example of a higher risk of bias in that category is the use of patients in a certain cardiology practice prescribed benfluorex as the exposed group, and medical staff, friends and family from the same practice for use as the unexposed group. In cases of different ratings between the two assessors, the authors came to a consensus, with the option of a third assessment (JG) if necessary. The authors chose a score of 24 as the cut-off score below which an article would be excluded, as articles with a score of at least 24 had a low risk of bias in most categories.
Supplementary file 2
Data were extracted from each article to be included in the quantitative synthesis, including information about the population, the medication used and event rates. In the case of dopaminergic agents, data were extracted in a way that separated patients being treated for PD versus HPL to account for differences in dosing regimens. The authors chose to use the Centers for Disease Control (CDC) and Food and Drug Administration (FDA) definition of DI-VHD, which is mild or greater aortic valve regurgitation or moderate or greater mitral or tricuspid valve regurgitation.12
The analysis was conducted in RevMan using a random effects model.13 A minimum of three studies contributing data was required for each analysis to calculate the OR. Where possible, analyses were conducted by medication type, valve affected and duration of exposure. A sensitivity analysis, testing whether outliers in the cohort might be affecting the results, was carried out. Each analysis was examined for heterogeneity using the I2 statistic, and the potential for publication bias was assessed using funnel plots Figure 1.
A total of 786 records were identified in our literature searches, which was reduced to 626 after duplicates were removed (Figure 2). A total of 461 articles were judged to be ineligible for inclusion based on the title. Of the 165 remaining articles, 3 were in vitro experiments, 2 did not have comparison groups, 7 looked at non-VHD outcomes, 42 described single cases or small case series and 46 were commentaries, reviews, letters to the editor, meta-analyses or other articles presenting no original data. This left 65 articles that were read and assessed for quality. Twenty-four articles were judged to be of adequate methodological quality with sufficiently low risk of bias to include in the meta-analysis.14–37 The assessments of each included article can be found in online supplemental table 2.
The relevant data were extracted and are presented in table 1. The articles included in the analysis were published between 1998 and 2016, with studies from the USA, Europe and Asia. The sample sizes ranged from n=62 to n=15 813. The doses of medications and durations of exposure were variable. The study designs included prospective and retrospective cohort studies, case–control studies, cross-sectional studies and randomised controlled trials (RCTs). The majority of the studies were observational studies assessing the association between medications and VHD. The included RCTs were initially devised to evaluate the efficacy of the medications for their FDA-approved purpose. For example, the RCT described by Weissman et al intended to compare dexfenfluramine to sustained-release dexfenfluramine to a placebo for the treatment of obesity, but the trial was amended to include echocardiographic assessment of patients enrolled in the trial after concerns about the potential valvulopathic effects of dexfenfluramine became widely known.34
Association between HVD and 5HTR2B-related serotonergic medications
There was a positive association between serotonergic medications and VHD (Figure 3A–C). This association was significant among the serotonergic agents as a group (OR 3.30, 95% CI 1.99 to 5.49), and when the analysis was limited specifically to dexfenfluramine (OR 5.90, 95% CI 1.99 to 17.55). When the analysis was limited to benfluorex, the positive association did not reach statistical significance (OR 3.26, 95% CI 0.44 to 24.07).
We analysed specific association with different heart valves. As shown in Figure 3D,E there was a significant association when the analysis was restricted to the aortic valve (OR 2.02, 95% CI 1.23 to 3.31) and the mitral valve (OR 2.11, 95% CI 1.17 to 3.82) but not for the tricuspid valve (OR 1.48, 95% CI 0.10 to 22.73). Insufficient data were reported to conduct analyses for the pulmonic valve.
Association between HVD and dopaminergic medications
There was also a positive association between dopamine agonists that are known to have 5HTR2B affinity and VHD (OR 2.56, 95% CI 1.68 to 3.91) (Figure 4A). Pergolide (OR 2.34, 95% CI 1.29 to 4.25) and cabergoline (OR 4.62, 95% CI 1.93 to 11.09) were significantly associated with VHD when the analysis was restricted to these specific medications, but there was no significant association for bromocriptine (OR 1.97, 95% CI 0.91 to 4.30) (Figure 4B–D). For all dopamine agonists, there was a significant association with aortic (OR 4.10, 95% CI 1.77 to 9.50), mitral (OR 2.32, 95% CI 1.27 to 4.22) and tricuspid valve pathologies (OR 2.62, 95% CI 1.22 to 5.63) (Figure 4E–G).
Dopamine agonist in PD and HVD
When the analysis was limited to patients taking dopamine agonists to treat PD, there was a significant association (OR 2.36, 95% CI 1.66 to 3.36). However, when the analysis was restricted to patients taking dopamine agonists to treat conditions other than PD, there was no significant association with VHD (OR 1.45, 95% CI 0.22 to 9.65) (Figure 5A-B).
Sensitivity and heterogeneity analysis
In the sensitivity analysis, exclusion of studies with results that were significantly different from the overall cohort did not change the direction or significance of the results for any of the analyses. There was significant heterogeneity in the analysis. This was most pronounced in the overall analyses that pooled studies of serotonergic agents (I2=82%) and dopaminergic agents (I2=52%) and was less significant in the subanalyses, such as the analysis looking at DAs prescribed for patients with PD (I2=26%).
Publication bias was assessed separately for both the serotonergic medications as well as the dopamine agonists with funnel plots (Figure 5). For both types of medications, there was an indication that there may be smaller studies and studies that did not disprove the null hypothesis that were unpublished or unavailable.
In this work, we conducted a meta-analysis to bring together the current information on the associations between DI-VHD and agents known to have serotonergic or dopaminergic activities. Our analysis supports the idea that certain medications are associated with an increased risk of cardiac valve pathologies. The consistency of the association between medications that are possible agonists of 5HTR2B, SERT inhibitors or DI-VHD also supports the overall idea that 5HT signalling is associated with direct functional role on heart valve pathological remodelling.
With respect to our analysis on serotonergic medications, there was a strong and consistent association between the medications and clinically significant VHD. The greatest depth of literature, as expected, was in the area of serotonergic appetite suppressants, but we identified two studies that looked for links between DI-VHD and selective serotonin reuptake inhibitors (SSRIs) and 3,4-methylenedioxymethamphetamine (MDMA or ecstasy), respectively. Although these were two small case–control studies, they are reminders that other medications that are 5HT receptor agonists have not been as thoroughly studied and may have similar effects.
Regarding the dopamine agonists, there was a consistent relationship between the medications and clinically significant VHD. When stratified by indication for the DA, however, the relationship was only significant when the DAs were prescribed to treat PD. This may be due to differences in doses by indication; patients taking dopamine agonists for PD are prescribed doses that are often 10 times greater than those prescribed for HPL. Our results here are consistent with some studies we identified in the literature.38 39 The FDA’s decision to withdraw cabergoline and pergolide’s marketing approval for the treatment of PD but not for HPL is also consistent with these findings.
Several of our subanalyses did not reach statistical significance, including the subanalyses of benfluorex and bromocriptine and the analysis of tricuspid disease among patients taking serotonergic agents. These analyses often had relatively fewer, smaller studies contributing data, so it is possible that they were underpowered to detect a treatment effect. For the analysis of patients taking dopaminergic agents for non-PD, which we discussed previously, there was a relatively robust sample size of over 13 000 patients, and we feel that this null result is more likely due to the relatively low doses given to patients with HPL.
As Rothman and Baumann noted in their review of DI-VHD, ‘[given] the histopathological similarities between the heart valves from carcinoid patients and those from fenfluramine-treated patients, it seemed logical to assume that elevations in plasma 5HT could be responsible for fenfluramine-associated VHD’.40 However, the 5HT hypothesis is not as simple as a direct link between plasma levels of serotonin and valvular pathology. A more current iteration of the hypothesis is that the HVD is induced through one or more 5HT-associated pathways. One potential mechanism is through the activation of serotonin receptors. These receptors, encoded by 13 distinct genes and classified into seven families including the Gq/11-coupled 5HT receptors (5HTR2), are understood to mediate both the contractile and proliferative effects of 5HT on smooth muscle cells, fibroblasts and valve interstitial cells.10 41 These molecules are presently the most investigated regulators of 5HT signalling. Many of the drugs commonly associated with DI-VHD have demonstrated high affinity to the 5HTR2B through in vitro receptor binding assays, although evidence suggests that other receptors may also be important for 5HT signalling and regulation in certain HVDs.9 42 It has also been suggested 5HT-related signalling pathways that are altered through mechanisms involving the serotonin transporter (SERT) may contribute to DI-VHD; studies have demonstrated aortic and mitral valvulopathy in animal models with SERT deletions43; changes to 5HT metabolism and leaflet function observed as a result of altered mechanotransductions responses44; and hypotheses that genetic variations of SERT may alter 5HT elimination and thus affect susceptibility to the development of VHD.45
Several studies have reported a role for 5HT signalling in the remodelling of heart valve in vivo and in vitro. The valvular interstitial cells (VICs) found in the leaflets of diseased valves undergo significant changes during pathological remodelling but were generally believed to be quiescent in the mature adult healthy valves. Recent research has suggested, however, that these cells can become activated as a result of certain stimuli. The activated VICs can then undergo cellular transformations changing them to a more synthetic phenotype. These include biological (ie, TGFb1, BMP4 and angiotensin) and mechanical (deformation) stimuli or a combination of these.46 47 This plasticity is modulated by multiple drugs. Recent findings shows that a 5HTR2B antagonist (LY272015) is controlling mitigates mitral VIC activation in vitro and leaflets’ remodelling in vivo.48
The breadth of our analysis is intentional. The available literature supports a wide view of the medications with the potential to induce DI-VHD. Many of these medications have been removed from the market by competent authorities in the various jurisdictions in which they were marketed, as is the case with the serotonergic appetite suppressants and EDDAs. Our analysis should serve as a reminder, however, that many potential inducers of DI-VHD have been inadequately studied. For example, our analysis and others have suggested that the use of MDMA may be associated with an increased risk of DI-VHD. MDMA use has been increasing among young adults, with many reporting a perception that the drug presents few risks to their health.49 In the 2014 National Survey on Drug Use and Health, more than 17 million young adults in the USA reported taking MDMA at least once.50 Although the evidence of an association between SSRIs and DI-VHD is less well studied, the small study by De Backer and colleagues strongly suggests that this is an area of the literature that deserves further research.16
This study is limited by the observational, non-randomised nature of most of the included studies. Many of the included studies were cohort trials, often retrospective in nature and subject to the biases of this study methodology. The heterogeneity of the included studies is also a limitation; the included studies did not enrol homogenous populations, and there may be other differences that contributed to the results. There was also significant variability in the duration of treatment with each medication, the doses of various medications and the length of time between the start of treatment and when echocardiographic assessment was done; each of these factors would contribute to the heterogeneity observed in the pooled analyses. Given the strength and consistency of the results, however, as well as the numerous publications that have come to similar conclusions, we are confident in the association between these medications and DI-VHD.
Our meta-analysis found a consistent association between DI-VHD and both serotonergic appetite suppressants and EDDAs. Although some of these drugs have been withdrawn from the market, signals from small studies associating DI-VHD with the recreational drug MDMA, and SSRIs currently used to treat mood disorders suggest that vigilance and population-scale surveillance of these medications may be indicated.
What is already known on this subject?
The link between certain medications and valvular pathologies has been well established in the literature. Evidence from the basic and translational sciences has indicated that these medications may exert their effect through serotonergic pathways.
What might this study add?
This study takes a broad approach to drug-induced valvular heart disease including a variety of medications associated with valvular pathologies. We found significant associations between valvular heart disease and serotonergic medications (OR 3.30) and dopaminergic medications (OR 2.56).
How might this impact on clinical practice?
Our results suggest that regulators, clinicians and pathologists must be aware of the potential for certain medications to induce changes to the cardiac valves and consider enhanced surveillance for patients taking medications that are known to activate serotonin pathways.
Contributors The research question was developed by authors GF, BP, RJL and JG. The literature search and assessment of articles was completed by JHF and BP, with JG acting as a third assessor in case of a disagreement between authors. The analysis was conducted by JHF and RES. The interpretation was conducted by JHF, RES, RJL, GF and JG. The manuscript was drafted by JHF and JG, with significant intellectual revisions and input from BP, RES, RJL and GF.
Funding This work was supported by the following research grants and funds: National Institutes of Health (R01-HL131872 to GF and RJL), The Kibel Fund for Aortic Valve Research (to GF and RJL), The Valley Hospital Foundation ’Marjorie C Bunnel’ charitable fund (to GF and JG), and both Erin’s Fund and the William J Rashkind Endowment of the Children’s Hospital of Philadelphia (to RJL).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The full set of data used in this analysis, including search terms and results, full set of records reviewed, quality assessments of articles, data extracted from articles and analyses conducted, will be maintained by the authors. They can be made available upon requested to the corresponding author.
Patient consent for publication Not required.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.