Valvular Abnormalities and Cardiovascular Status Following Exposure to Dexfenfluramine or Phentermine/Fenfluramine FREE

Dexfenfluramine and fenfluramine are antiobesity agents that were indicated as adjunctive therapies to a weight loss regimen that included dietary restriction.^1- 2 Dexfenfluramine had been available in Europe since 1985 but was on the US market for less than 16 months. Fenfluramine was available both domestically and abroad for more than 20 years. Reports of valvular abnormalities in 1997 resulted in the voluntary withdrawals of these agents from the market. However, these observational reports lacked pretreatment echocardiographic data and nonanorexigen-treated control cohorts, and criteria for subject selection were unclear.

Since initial reports linking anorexigens and valvular abnormalities,^3- 7 additional studies have yielded disparate prevalence rates of echocardiographic findings, often without specific reference to the patients' overall cardiovascular health.^8- 11 These evaluations have studied different populations with differing or unspecified durations of exposure. This study was undertaken to evaluate the prevalence of valvular abnormalities and clinical signs and symptoms in large cohorts of dexfenfluramine and phentermine/fenfluramine-treated patients compared with an untreated group.

This was a reader-blinded, controlled, multicenter study evaluating cardiovascular clinical status and prevalence of echocardiographic valvular abnormalities in patients who had received dexfenfluramine or phentermine/fenfluramine for the treatment of obesity. This study was conducted according to a prospectively designed clinical protocol and was in full compliance with all federal, state, and local regulations pertaining to human research and with Good Clinical Practice guidelines. All study centers had prior approval from an appropriate institutional review board, and written informed consent was obtained from each patient prior to study entry.

Treated patients were required to have received at least 30 days of continuous therapy with dexfenfluramine or phentermine/fenfluramine and could not have discontinued therapy more than 14 months prior to enrolling in the study. Untreated subjects could not have been treated with anorexigens for 5 years prior to study entry (determined by patient interview and medical record review). Patients were excluded if they were younger than 18 years, had used other prescription or specified over-the-counter anorexigens or serotonergic migraine headache medications in the 5 years prior to study entry, or had a history of carcinoid tumor or syndrome. Because this was a general population study designed to represent those patients who took anorexigens, no patient was excluded for previously abnormal cardiovascular findings or medical history.

Enrollment of 1200 eligible participants was planned (400 in each group) to provide at least 80% power to detect a 10% difference in prevalence of valvular abnormalities between groups. Age, sex, and body mass index (BMI) have previously been shown to be independent predictors of cardiac valvular abnormality,^12- 18 and geographic region has been suggested as an independent predictor of valvular disease (eg, rheumatic disease). Therefore, the protocol plan was to match study patients by sex, age (within ± 5 years), BMI (within ±3 kg/m²), and geographic location (same site or geographically nearest). Investigators known to be frequent prescribers of dexfenfluramine and phentermine/fenfluramine in the United States were contacted. Those who were interested, qualified, and able to conduct the study and had adequate numbers of treated and untreated patients in their practices were invited to participate; no site meeting these requirements was excluded.

A dexfenfluramine-treated patient was recruited from site-specific randomly ordered lists of potentially eligible patients; then at the same site, a matched phentermine/fenfluramine-treated patient and control patient were enrolled (if matched patients could not be identified at the same site, the next nearest site was contacted). During patient recruitment, it became apparent that enrollment of such a large population matched with respect to 4 characteristics would not be feasible and the protocol was modified to allow all remaining eligible patients to participate. After enrollment was complete, the matching algorithm was reapplied to the evaluable patient list with untreated subjects as the reference group.

Each participant had a single clinical and echocardiographic evaluation consisting of a detailed medical history, a complete physical examination performed by the study physician (with emphasis on the cardiovascular system), and an echocardiogram (ECHO). Two-dimensional, M-mode, color-flow, pulsed, and continuous wave Doppler ECHOs were performed by trained sonographers blinded to all aspects of patient history including anorexigen use. A specified echocardiographic imaging protocol¹⁹ was used and echocardiographic equipment was standardized (Sonos 2000 or 2500 Imaging Systems; Hewlett-Packard, Andover, Mass).

Tapes were interpreted by an established central core laboratory (University of California, Irvine) staffed by trained sonographers and board-certified cardiologists specializing in echocardiography who were blinded to patient group and medical history. The following parameters of valvular function were evaluated: (1) aortic (AR), mitral (MR), tricuspid (TR), and pulmonic (PR) regurgitation, and (2) aortic, mitral, and tricuspid valve leaflet thickness and mobility.

Aortic regurgitation was visually graded based on modified Perry criteria²⁰ as follows: none, no regurgitant color flow in any view; trace, regurgitant jet diameter in the parasternal (or apical) long-axis view 5% or less of the outflow tract diameter; mild, regurgitant jet diameter more than 5% and less than 25% of the outflow tract diameter; moderate, regurgitant jet diameter 25% or more and less than 47% of the outflow tract diameter; moderately severe, regurgitant jet diameter 47% or more and less than 65% of outflow tract diameter; and severe, regurgitant jet diameter 65% or more of the outflow tract diameter.

Mitral and tricuspid regurgitation were graded based on modified criteria of Helmcke et al²¹ as follows: none, no regurgitant color flow; physiologic, nonsustained jet within 1 cm behind the annular plane with a maximal jet 5% or more of the left atrial (for MR) or right atrial (for TR) area; mild, sustained color flow jet with a maximal jet area more than 5% and less than 20% of atrial area; moderate, maximal jet area 20% or more and 40% or less of atrial area; and severe, maximal jet area more than 40% or a color flow jet reaching the back of the atrium. Eccentric jets of the mitral and tricuspid valves were upgraded 1 grade if jet impingement on a chamber wall precluded development of the full jet area.

Valve leaflets were considered abnormally thickened if mitral or tricuspid leaflets measured more than 4 mm during diastole, or if echodensities of the aortic valve were detected in more than 1 view. Leaflet mobility was coded abnormal if there was restriction of motion due to morphologic changes with maximal cusp separation of 1.5 cm or less (aortic valve) or impaired excursion of leaflets by more than 50% (mitral valve). Pulmonary artery systolic pressure (PAP) was calculated using the modified Bernoulli equation.²²

A second, independent, blinded reading of an ECHO was performed when abnormal valvular regurgitation, leaflet thickness, or mobility was detected on the initial reading. Consensus readings were performed to arbitrate reader disagreement. Data from the first reading were used in the final analysis except where a consensus reading was performed. In addition, a random sample of tapes was reread to assess interreader and intrareader variability.

Primary echocardiographic end points included mild or greater AR and moderate or greater MR based on US Food and Drug Administration (FDA) criteria.^23- 24 These criteria were used in original estimates of regurgitation prevalence by the FDA and in subsequent publications evaluating prevalence of AR and MR in anorexigen-treated patients. However, because this definition excludes patients with milder degrees of regurgitation, secondary echocardiographic end points included valvular regurgitation by grade, as well as aortic, mitral, and tricuspid valve leaflet mobility and thickness. Other study end points included serious cardiovascular outcomes, symptoms, and signs on physical examination.

Data reported describe the total evaluable study population, unless otherwise noted. Means and SDs are presented for continuous variables, and percentages are presented for categorical and ordinal data. Participant characteristics and demographics were compared among the 3 treatment groups using a 1-way analysis of variance for continuous variables and the Fisher exact or χ² test for categorical variables.

Data from the echocardiographic assessments were compared using the Kruskal-Wallis nonparametric test for regurgitation grade and the Fisher exact test for dichotomous outcomes. Because the majority of patients took the anorexigen for less than 90 days while the drugs were on the market, an a priori decision was made to also specifically analyze prevalence among patients who took the drugs for 3 months or less.

Relative risks (RRs) with 95% confidence intervals compared with the untreated group were computed for AR and MR—both by FDA criteria and for all grades. Cohen κ statistic was used to assess interreader and intrareader variability of ordinal and categorical echocardiographic parameters. Logistic regression analyses were performed to identify predictors of valvular regurgitation (FDA criteria). Independent variables included in the model were treatment group, demographic parameters (eg, sex, BMI, age, race, and geographic location), selective serotonin reuptake inhibitor use, and comorbid factors (eg, diabetes, thyroid disease, previous myocardial infarction, and hypertension). Treatment and geographic location were included in each model. Interactions among variables were also tested. Comparisons among treatment groups were considered statistically significant at P≤.05 (all statistical tests were 2-tailed). All analyses were performed using SAS (version 6.12, SAS Corp, Cary, NC) and Stat Exact statistical software (Cytel, Cambridge, Mass).

A total of 1473 evaluable participants provided 89% power to detect differences between the smaller treated group (phentermine/fenfluramine, n = 455) and untreated subjects (n = 539), assuming an untreated group prevalence of 6% and a difference of at least 6%. A total of 1640 study participants were enrolled at 25 clinical sites. Treated patients were matched to the untreated group by age, sex, BMI, and geographic location yielding a matched population of 1012 (dexfenfluramine, n = 305; phentermine/fenfluramine, n = 303; untreated, n = 404), which provided 75% power to detect differences between study groups using the same assumptions as above. Participant enrollment and matching are shown in Table 1. Data shown describe the total evaluable population of 1473 patients, unless otherwise specified.

In all 3 study groups, participants were predominantly white, female, obese, and in their fifth decade. The mean (SD) age was 47.4 (11.4) years; mean BMI was 35.0 (7.4) kg/m²; and 74% were women (Table 2). Median maximum daily doses taken by participants were equivalent to the recommended daily doses of dexfenfluramine (30 mg), phentermine (30 mg), and fenfluramine (60 mg); however, daily doses as high as 90 mg of dexfenfluramine, 112 mg of phentermine, and 120 mg of fenfluramine were reported. Mean duration of treatment with phentermine/fenfluramine was 11.9 (10.4) months and with dexfenfluramine was 6.0 (3.3) months.

Cardiovascular symptoms (including chest pain, chest pounding, tachycardia, syncope, lightheadedness, dizziness, and dyspnea on exertion and at rest) were similar across the 3 study groups. There were no statistically significant clinical differences in cardiovascular history or physical findings (including heart murmur, edema, jugular venous distention, and rales) across the 3 study groups among patients with AR, or MR, or both (FDA criteria). No patient had endocarditis or a valve replacement (other than 2 dexfenfluramine-treated patients who had a prior history of valve replacement as a result of rheumatic heart disease at least 4 years before receiving anorexigen treatment). Prevalence of serious cardiac events (including myocardial infarction, congestive heart failure, ventricular arrhythmia, and endocarditis) at any point was not statistically greater in treated than untreated subjects (dexfenfluramine, 9.0%; phentermine/fenfluramine, 4.0%; untreated, 8.4%). Serious cardiac events following anorexigen therapy occurred in 2.3% of dexfenfluramine-treated and 2.4% of phentermine/fenfluramine-treated patients. To compare prevalence of serious cardiac events between treated and control patients during the postanorexigen period, the median start date of anorexigen use among treated patients was used as the reference date for controls. Using this method, 3.3% of the control population experienced serious cardiac events in a similar period (all treated vs controls, P = .32).

Although there were no statistically significant intergroup differences, dexfenfluramine-treated patients with FDA-positive valvular regurgitation were more likely to have had a history of heart murmur, rheumatic fever, myocardial infarction, ventricular arrhythmia, or other cardiovascular history or to have had an ECHO or cardiac catheterization prior to anorexigen therapy.

ECHO Interpretation. Interreader concordance among the 3 cardiologist readers was assessed using 367 ECHOs that were reread for abnormalities and quality control purposes. κ Coefficients (and percent exact agreement) for interreader concordance (using FDA criteria for AR and MR and moderate or greater for TR and PR) were: AR, 0.68 (85%); MR, 0.81 (95%); TR, 0.78 (96%); and PR, 0.55 (99%). Interreader consistency was addressed by a consensus adjudication process. κ Coefficients for intrareader agreement (n = 53 ECHOs) were: AR, 0.84 (94%); MR, 0.74 (94%); TR, 0.69 (94%); and PR, 1.0 (100%).

Aortic Valvular Regurgitation. The prevalence of AR by FDA criteria was significantly increased in anorexigen-treated patients: 8.9%, 13.7%, and 4.1% in the dexfenfluramine, phentermine/fenfluramine, and untreated groups, respectively (P<.001) (Table 3). Among patients who took anorexigens for 3 months or less (phentermine/fenfluramine, n = 48; dexfenfluramine, n = 92), there was no statistically significant difference in AR prevalence by FDA criteria compared with untreated subjects (dexfenfluramine, 0%; phentermine/fenfluramine, 4.2%; untreated, 4.1%). There was also no statistical difference in AR prevalence between patients treated for 3 months or less and controls when AR was assessed using all grades (P = .18). However, because of the small numbers of patients enrolled in this study who received treatment for less than 3 months, the power to detect a difference in this group was less than 50%. Among patients who took anorexigens for more than 3 months, there was a statistically significant increase in the prevalence of AR by FDA criteria (dexfenfluramine, 11.1%; phentermine/fenfluramine, 14.9%; untreated, 4.1%; P<.001) (Table 3 and Figure 1).

Moderate and severe AR were uncommon (<1% of the total population) even in patients treated for the longest duration, and prevalence in treated patients was not statistically different from untreated subjects. Furthermore, there was no relationship between regurgitation grade and length of therapy in patients treated for more than 3 and up to 24 months. Four of 48 subjects treated for more than 24 months with phentermine/fenfluramine had moderate AR; 3 of the 4 had a history of heart murmur predating anorexigen exposure. There were no patients in this group with severe regurgitation.

Mitral Valvular Regurgitation. The prevalence of MR by FDA criteria was not statistically different across the 3 groups (dexfenfluramine, 4.9%; phentermine/fenfluramine, 5.1%; untreated, 3.2% (P = .24) [Table 3]) and there was no duration effect. However, when all grades of MR were evaluated by the Kruskal-Wallis test, there was a statistically significant increase in MR grade in dexfenfluramine and phentermine/fenfluramine groups compared with the untreated group, primarily due to an increase in mild MR (P = .02). In subjects treated for 3 months or less, there was a statistically significant (P = .04) increase in MR in the phentermine/fenfluramine group, but not in the dexfenfluramine group, when all grades were assessed.

Other Echocardiographic Parameters. There was no statistically significant difference in the prevalence of moderate or greater TR (3.1% of dexfenfluramine patients, 4.2% of phentermine/fenfluramine patients, and 2.2% of untreated patients [P = .22]). Furthermore, there was no statistically significant difference in other echocardiographic parameters across the 3 groups, including PR, calculated PAP (comparing either mean PAP or patients with PAP estimates ≥40 mm Hg), and aortic, mitral, and tricuspid leaflet thickening and mobility. Left ventricular ejection fraction was normal (defined as >0.50) in more than 98% of patients and left ventricular end-diastolic, left ventricular end-systolic, and left atrial dimensions did not differ across the 3 study groups.

Analyses of the matched vs the total evaluable population were performed, and demographic characteristics were similar (Table 2). The prevalence of AR (FDA criteria) in the matched population also was significantly increased in treated patients: 7.3% in the dexfenfluramine group (n = 305), 14.0% in the phentermine/fenfluramine group (n = 303), and 3.7% in the untreated group (n = 404), P<.001. Prevalence of MR (FDA criteria) was 5.0% in the dexfenfluramine group, 5.0% in the phentermine/fenfluramine group, and 3.2% in the untreated group (P = .40). With the exception of aortic valve leaflet thickening, which was more prevalent among matched anorexigen-treated patients (6.7% of dexfenfluramine-treated patients, 6.4% of phentermine/fenfluramine-treated patients compared with 3.0% of untreated subjects, P = .04), all other results including cardiovascular clinical findings were comparable in the matched and total evaluable populations.

We also performed a post hoc analysis of the matched population (n = 846) from which 166 patients with echocardiographic evidence of valvular abnormalities characteristic of other valvular pathology (eg, rheumatic heart disease, congenitally bicuspid aortic valve, and mitral valve prolapse) and those who had an ECHO prior to anorexigen use were excluded. Prevalence of AR in this population was also significantly increased among anorexigen-treated patients (dexfenfluramine, 6.3%; phentermine/fenfluramine, 13.9%; untreated, 3.4%; P<.001) and was comparable to the results obtained from the analyses of the total evaluable population.

A multivariate statistical model using stepwise logistic regression indicated that increased age (P<.001), history of heart murmur (P<.001), lower BMI (P = .004), and female sex (P = .01) were associated with AR (FDA criteria) independent of treatment (Table 4). History of hypertension was also associated with AR (P = .009) and there was a statistically significant treatment-hypertension interaction for both treatment groups.²⁵ The logistic regression model indicated that the prevalence of AR increased approximately 60% in all 3 study groups for each 10-year increase in age. Variables tested in the model, but not found to be statistically significant, included race, diabetes, previous myocardial infarction, prior or current alcohol use, selective serotonin reuptake inhibitor use, thyroid hormone replacement, smoking history, mitral valve prolapse, other aortic valvular pathology, and use of angiotensin-converting enzyme inhibitors.

In this clinical study, we documented an increase in the prevalence of AR (FDA criteria) among patients who were treated with dexfenfluramine or the combination of phentermine/fenfluramine compared with untreated subjects. There was no increase in moderate or severe AR in treated patients, and no difference in the prevalence of MR (FDA criteria) between the untreated and treated groups irrespective of duration of therapy. There was, however, a statistically significant increase in MR prevalence in both anorexigen-treated groups when all regurgitation grades were evaluated, primarily due to an increase in mild MR.

Morphologic valve leaflet abnormalities have previously been described among anorexigen-treated patients^3,11 and have been linked to anorexigen treatment. We found no significant differences in the prevalence of valve leaflet thickening or impaired mobility, with the exception of aortic leaflet thickening, which was more prevalent in anorexigen-treated subjects in the matched group analysis (P = .04), but not in the total evaluable population (P = .22). Therefore, the clinical significance of this finding is unclear. Moreover, in patients with AR or MR (FDA criteria), we found no statistically significant differences among treatment groups in cardiovascular signs or symptoms or in serious cardiovascular outcomes, and more than 98% of patients in our study in all groups had normal left ventricular ejection fraction.

Initial reports suggested a higher prevalence of cardiac valvular abnormalities among patients treated with anorexigens.^3,23 These reports involved small groups of patients and lacked baseline data, blinded ECHO readings, and an untreated cohort. Three large clinical studies have provided additional information. In a prospective, double-blind clinical trial of 1072 participants treated with dexfenfluramine, sustained-release dexfenfluramine, or placebo for a median duration of 78 days, Weissman and colleagues¹¹ found no statistically significant difference in the prevalence of AR or MR (FDA criteria). Similarly, our study showed no increase in FDA-defined valvular regurgitation in the relatively small group of patients who took dexfenfluramine for 3 months or less. Additionally, the control populations in the 2 studies had similar AR prevalence rates, 3.6% and 4.1%, higher than were previously thought to occur in the general population.²⁶ Khan and colleagues⁹ reported that obese subjects who took anorexigens (primarily phentermine concomitantly with fenfluramine) had a significantly higher prevalence of cardiac valvular insufficiency compared with untreated patients. Mean (SD) duration of therapy was 20.5 (12) months, significantly longer compared with the study by Weissman et al¹¹ or our study. Like ours, the findings from the study by Khan et al consisted primarily of AR, with no significant increase in FDA-defined MR. Jick et al⁸ reported a population-based follow-up evaluation and nested case-control analysis of 8903 subjects who received dexfenfluramine or fenfluramine. Prevalence rates reported in that study suggested that clinically relevant anorexigen-associated valvular abnormalities occurred rarely (0.1%) and predominantly in patients treated for 4 months or longer.

Our study had several limitations inherent to its design. First, because dexfenfluramine and fenfluramine were withdrawn from the market, a randomized prospective study could not be performed. Second, although investigators made every effort to randomly select patients and were given instructions to include all treated patients on their randomization lists, enrollment bias was possible. Third, complete matching of the 3 study groups could not be achieved, and as a result, dexfenfluramine patients tended to be older, more obese, more hypertensive, have more history of cardiovascular disease prior to anorexigen treatment, and were more frequently males and smokers. This may reflect a potential source of bias as investigators may have been more apt to enroll treated subjects with preexisting cardiovascular abnormalities into this study. Fourth, available anorexigen dosing data may not have been entirely accurate as they were often based on outpatient report of maximum daily dose rather than medication records. Finally, the study was not specifically designed or adequately powered to evaluate specific categories of anorexigen therapy duration.

In summary, use of dexfenfluramine and phentermine/fenfluramine was associated with a significantly increased prevalence of AR (FDA criteria) compared with untreated patients. However, there were no significant differences in the prevalence of moderate and severe AR, in clinical cardiovascular status, or serious cardiac outcomes, between anorexigen-treated and untreated patients.