Clinical and Echocardiographic Follow-up of Patients Previously Treated With Dexfenfluramine or Phentermine/Fenfluramine FREE

Limited information is available on the natural history of valvular regurgitation after discontinuation of anorexigen therapy.^1- 3 The only large, randomized controlled trial did not show progression or development of new valvular regurgitation approximately 1 year after discontinuation of dexfenfluramine,⁴ but only included patients treated for up to 3 months. We previously reported in a multicenter, prospective controlled study a statistically significant increase in prevalence of echocardiographic aortic regurgitation (AR) meeting Food and Drug Administration (FDA) criteria of mild or more severe regurgitation among patients previously treated with dexfenfluramine (8.9%) and phentermine/fenfluramine (13.7%) compared with untreated controls (4.1%) (P<.001).⁵ Among patients who took anorexigens for 3 months or less, there was no increase in AR prevalence by FDA criteria.

This was a 1-year follow-up study of 1466 patients who had an evaluable initial echocardiogram and clinical assessment.⁵ At baseline, treated patients had received at least 30 days of continuous anorexigen therapy and had not discontinued therapy more than 14 months previpously; untreated controls had not received anorexigens for 5 years.

One-year assessment included a physical examination and an echocardiogram. Interim medical and cardiovascular histories were obtained from patient interviews and chart reviews. Echocardiograms were performed at 25 clinical sites using a previously described protocol.⁵ Local sonographers remained blinded to the patients' history and treatment group.

Six board-certified cardiologists at 2 central laboratories evaluated paired echocardiograms for change in valvular regurgitation, aortic and mitral leaflet thickening and mobility, and left ventricular ejection fraction. Cardiologist-readers were blinded to patient information, treatment, and sequence of echocardiogram acquisition. Initial and follow-up echocardiograms were electronically masked and randomized for side-by-side review in videocassette recorders. Cardiologists noted whether the paired recordings were comparable and, if so, whether there was a difference of 1 grade or higher. After all pairs were evaluated, a biostatistician broke the blind.

Aortic and mitral regurgitation were evaluated using color Doppler criteria.^5- 7 For aortic and mitral valve thickness and mobility, the 2 studies were assessed for change. The left ventricular ejection fraction was assessed for a change of 1 grade or higher.⁵ Laboratory sonographers measured tricuspid regurgitation jet velocity using continuous wave Doppler. Pulmonary artery systolic pressure (PASP) was estimated as 4V² + 10 mm Hg, where V equals peak velocity (meters per second).

For echocardiogram pairs scored as noncomparable for valvular regurgitation, aortic and mitral leaflet thickening or mobility, a second blinded reading (n = 187 [16.4 %]) was performed. Consensus readings were used to arbitrate reader disagreement regarding comparability. A 5% random sample of tapes (n = 64) was recirculated to assess interreader agreement; a random sample of 20 echocardiograms was reread to assess intrareader agreement.

Primary end points were change in AR or mitral regurgitation (MR) grade and new cardiovascular symptoms and physical findings. One-way analysis of variance was used to compare means, and χ² or Fisher exact test was used to compare proportions among study groups. McNemar test for correlated proportions was used to compare within-group changes. Echocardiographic parameters judged to be noncomparable were not included in statistical analyses. McNemar test demonstrated 80% power to detect a change in any echocardiographic parameter for 300 pairs if 10% of pairs were discordant (ie, showed a change in grade) and the difference in proportions was 5%. Percentage exact agreement and unweighted κ coefficients with asymptotic SEs were calculated for intrareader and interreader agreement in assessing change in AR and MR. All statistical tests were 2-sided and P≤.05 was considered significant.

Multivariate stepwise logistic regression analyses were used to identify predictors of change for AR and MR. Independent variables tested were treatment group, demographic variables, change in weight, time between echocardiograms, medication use and change in medication use, comorbid factors (eg, diabetes, thyroid disease, previous myocardial infarction, murmur, and hypertension), and changes in comorbidity.

Of 1466 patients, 1142 (371 dexfenfluramine, 340 phentermine/fenfluramine, 431 controls) completed a 1-year echocardiogram (±1 month) with similar characteristics across all 3 groups (Table 1). Time between drug discontinuation and follow-up echocardiogram was approximately 1 month longer in the phentermine/fenfluramine group (P<.001). Mean (SD) duration of anorexigen use was 6.1 (3.3) months in the dexfenfluramine and 11.9 (10.4) months in the phentermine/fenfluramine group. There were more patients with hypertension (P<.001) in the dexfenfluramine group and fewer patients with diabetes (P = .02) in the phentermine/fenfluramine group.

The 331 participants in the original study who did not return for follow-up were evenly distributed across the 3 treatment groups. Reasons for nonparticipation included no interest (65 dexfenfluramine, 55 phentermine/fenfluramine, 55 controls), lost to follow-up (9 dexfenfluramine, 31 phentermine/fenfluramine, 28 controls), and other (33 dexfenfluramine, 28 phentermine/fenfluramine, 25 controls). One patient with diabetes in the phentermine/fenfluramine group, whose initial echocardiogram demonstrated no AR, MR, leaflet thickening, or mobility restriction reportedly died of a pulmonary embolism during the year. One patient in the dexfenfluramine group, who was lost to follow-up and whose initial echocardiogram demonstrated physiological MR, died of causes unknown to the investigators. Compared with participants, nonparticipants were younger (48.2 [11.2] vs 44.7 [11.6] years; P<.001) but not heavier (98.7 [23.2] kg vs 101.2 [25.7] kg; P = .11). There were no significant differences in valvular regurgitation on the initial echocardiogram between participants and nonparticipants.

Aortic valve recordings were comparable in 1114 pairs (97.6%) of echocardiograms. Among untreated controls, AR increased in 1 (0.2%) and decreased in 7 (1.7%) (P = .07; Table 2). In the dexfenfluramine group, AR increased in 6 (1.7%) and decreased in 23 (6.4%) (P<.001 vs controls). The AR grade increased from none to trace in 4 patients; none to mild in one patient; and trace to mild in another. For the phentermine/fenfluramine group, there were no increases in AR; decreases were observed in 15 patients (4.5%) (P = .03 vs controls). Within-treatment group changes were significant only for the treated groups. Time elapsed between discontinuation of anorexigen treatment and first echocardiogram was significantly shorter among patients with vs those without a decrease in AR (133.4 [77.8] vs 182.3 [93.2] days; P = .002).

Mitral valve images were comparable in 1106 echocardiogram pairs (96.8%). Among untreated controls, MR increased in 9/419 (2.1%), and decreased in 25/419 (6.0%). In the dexfenfluramine group, MR increased in 13/356 (3.7%) and decreased in 31/356 (8.7%) (P = .14 vs controls). The MR grade increased from none to physiological in 3 patients, physiological to mild in 5, mild to moderate in 3, and physiological to moderate in 2. In the phentermine/fenfluramine group, MR increased in 10/331 (3.0%) and decreased in 21/331 (6.3%) (P = .71 vs controls). The MR grade increased from none to physiological in 1 patient, none to mild in 1, physiological to mild in 5, mild to moderate in 2, and moderate to severe in 1 patient. Within-group changes were significant for dexfenfluramine (P = .01) and controls (P = .009), but not for phentermine/fenfluramine (P = .07). All changes in patients initially demonstrating moderate or severe MR (n = 50) were decreases, except for a 2-grade increase in 1 phentermine/fenfluramine patient with calcified mitral annulus.

There were no changes in aortic or mitral leaflet mobility or aortic leaflet thickening in any group. Among controls, 1 patient had decreased mitral leaflet thickening and 1 had increased left ventricular ejection fraction. The PASP was measurable in 150/1142 patients (equally measurable among the 3 groups). Follow-up mean (SD) PASP (32.8 [5.0]; 33.0 [4.5]; 32.5 [5.9] mm Hg) and changes in PASP (−0.76 [7.2]; 1.05 [7.3]; 0.82 [8.1] mm Hg) were not different among dexfenfluramine, phentermine/fenfluramine, and control groups, respectively.

Interreader agreement for change in AR and MR, assessed using 251 reading pairs, was 87.4% (κ, 0.63 [0.06]) for AR and 57.1% (κ, 0.32 [0.04]) for MR. Intrareader agreement (n = 120) was 96.5% (κ, 0.32 [0.24]) for AR and 86.8% (κ, 0.30 [0.13]) for MR.

There were no significant differences between treated patients and controls in the reporting of new cardiovascular symptoms, physical findings, or events. There were no cardiac valve surgeries. Among patients with changes in AR or MR, there were no significant differences between anorexigen-treated patients vs controls in the reporting of new cardiovascular symptoms or events, or physical findings. For the 6 treated patients and 1 control who had increased AR, there were no new cardiovascular symptoms, events, or positive cardiovascular physical findings (data not shown). Among patients who had an increase in MR, pulmonary hypertension and heart failure were not reported.

In logistic regression models including predictors of decreased regurgitation, there were no significant predictors of decrease in MR. There were 2 independent predictors of decrease in AR: treatment status (received anorexigen treatment vs no treatment; adjusted odds ratio [AOR], 2.75 [95% confidence interval {CI}, 1.20-6.32]; P = .02), and initial AR severity using FDA criteria (mild to more severe vs less than mild; AOR, 6.81 [95% CI, 3.57-12.99]; P<.001). There were no significant changes in systolic or diastolic blood pressure between initial and follow-up visits among or within groups.

These data demonstrate a lack of progression, and for some patients, a decrease in AR grade for those treated with dexfenfluramine and phentermine/fenfluramine compared with controls.

Change in AR appears unlikely to be related to the "regression to the mean" statistical artifact. For regression to the mean to exist, similar patterns of change should be evident in treated and control groups, associated with a low rate of specificity for change. Among controls, there were only 8/423 (1.9%) changes in AR among echocardiogram pairs (98.1% specificity). In addition, reliability was high, with 85.4% agreement among readers regarding change. We believe that the high rate of specificity, absence of a significant bias in the direction of change among controls, and randomized and blinded side-by-side reading protocol make regression to the mean highly unlikely.

Hypertension has been significantly associated with higher grades of AR,^5,8- 11 which was also observed in our study. Unlike previous studies, our study had power to detect a significant relationship of shorter time elapsed between discontinuation of anorexigen treatment and initial echocardiogram to decreased AR between echocardiograms.

Several serial echocardiographic studies have assessed the natural history of valvular abnormalities in anorexigen-treated patients, but 2 had small numbers of patients.^3,12 A large, longitudinal study reported a significant decrease in AR (P<.001) among dexfenfluramine-treated patients and no change among controls (mean follow-up of 11 months).⁴ The current study extends these findings by including patients who took anorexigens for more than 3 months (≤6.6 years).

This investigation has several limitations. Randomized, prospective treatment studies could not be conducted because of the abrupt withdrawal of dexfenfluramine and fenfluramine from the market. In addition, the dexfenfluramine group in our study was older, heavier, and more likely to have hypertension than other groups. Although there was potential for enrollment bias—with 331 patients not returning for follow-up—nonparticipants were evenly distributed among treatment groups. There were no statistical differences between participants and nonparticipants who demonstrated FDA criteria for AR and MR on the initial echocardiogram. Also, PASP was measurable only in 18.7% of participants in this study. This relatively low percentage is consistent with a recent report of 17.3% with measurable tricuspid regurgitation jets.¹⁰

In summary, there was no evidence of 1-year progression of MR or AR, with a statistically significant decrease in AR, and no changes in other valvular parameters, or in reports of cardiovascular events and physical findings, among treated and untreated groups.