Defining risk is especially important given the potential morbidity associated with valvular heart disease. Definitive randomized, controlled studies cannot be performed. However, retrospective case–control and cohort studies, which are superior to uncontrolled surveys for the estimation of true risk, can be done. To estimate the risk for medication-associated valvular heart disease, we identified users of fenfluramine or dexfenfluramine who underwent echocardiography before beginning therapy with these medications and compared these echocardiograms to those obtained after therapy was discontinued.

Methods

Top Methods Results Discussion Author & Article Info References

Patient Selection

Patients who were prescribed fenfluramine or dexfenfluramine were identified through medication records at primary care practices affiliated with two academic medical centers in Massachusetts. In general, providers selected patients for therapy on the basis of regional guidelines [8].

Eligible patients used fenfluramine or dexfenfluramine for at least 14 days and underwent echocardiography at the study institutions after 1 January 1988 but before use of diet medication. We obtained consent from physicians and patients according to a protocol approved by the institutional review boards at each study institution.

Study Design

Clinical data were obtained through chart review and patient interview [9]. Patients were asked about changes in cardiopulmonary symptoms and exercise tolerance by using questions from a validated instrument [10].

A cardiopulmonary examination was performed by an examiner who was blinded to the results of echocardiography. For patients who refused to undergo examination, we abstracted data from the most recent physician examination performed after completion of diet medication therapy.

Echocardiography

Complete echocardiograms were obtained before and after therapy. All cardiac valves were interrogated with spectral, continuous-wave, and color Doppler echocardiography from multiple views in the standard manner [11, 12] by using mainly Hewlett-Packard Sonos series ultrasonographs (Andover, Massachusetts) with 2.5- to 2.7-MHz transducers and fundamental frequency imaging. The most recent premedication and initial postmedication echocardiograms were reread in a blinded manner by one of two trained echocardiographers.

Valvular Abnormalities

The following criteria were used to define valvular abnormalities.

Mitral or tricuspid valve: Leaflet thickening was assessed as normal, mild, moderate, or severe according to established criteria [13]. Regurgitation was graded by qualitative estimate of the composite color Doppler jet area obtained in at least two of four views and was confirmed by spectral Doppler recordings of high-velocity, holosystolic flow in the left atrium. Regurgitation was graded as trace (jet area < 2 cm²), mild (2 to 5 cm²), moderate (6 to 10 cm²), or severe (> 10 cm²) [11, 12, 14, 15].

Aortic valve: Leaflet thickening criteria were similar to those used for the mitral and tricuspid valves except that thickening could be focal or diffuse. Regurgitation was graded according to the composite jet area obtained in three views and the jet height in the outflow tract and was confirmed by spectral Doppler recordings of high-velocity, diastolic flow in the left ventricular outflow tract. Regurgitation was graded as trace (jet area 1 cm²), mild (2 to 5 cm²), moderate (6 to 10 cm²), or severe (> 10 cm²) [11, 12, 15, 16].

Pulmonic valve: Leaflet thickening criteria were similar to those for the aortic valve. Regurgitation was considered normal (jet area 2 cm²) or abnormal (>2 cm²).

Outcome Measures

The primary echocardiographic outcome was development of new valvular lesions meeting U.S. Food and Drug Administration (FDA) criteria (at least mild aortic regurgitation or moderate mitral regurgitation) [1] or progression of preexisting valvular disease by at least one degree of severity that also met FDA criteria. A change in severity by one degree was defined as progression from no disease to mild disease, mild disease to moderate disease, or moderate disease to severe disease.

Statistical Analysis

Descriptive statistics were used to characterize patients in the cohort. The percentage of weight lost was determined by using the patient's nadir body weight while taking medication and baseline body weight before taking medication. Paired analysis was used to determine the presence of the primary outcome of new or worsening valvular lesions. Each patient's baseline echocardiogram served as the control for any new changes. The 95% CIs were calculated according to the normal approximation of the binomial distribution.

To determine the generalizability of results obtained from this cohort, we compared the characteristics (obtained by chart review) of study patients with those of a random 8% sample of patients from the two largest group practices (approximately 80% of study patients) who were prescribed fenfluramine and dexfenfluramine but did not undergo echocardiography before medication use. We also compared enrolled patients with eligible patients who declined to participate. Comparisons were made by using the Fisher exact test for categorical variables and the Wilcoxon rank-sum test for continuous variables. In addition, we analyzed the combined secondary outcome of change in cardiopulmonary symptoms and exercise tolerance or the presence of cardiopulmonary examination findings.

Results

Top Methods Results Discussion Author & Article Info References

Patient Enrollment

Of 873 patients identified as fenfluramine or dexfenfluramine users, 76 had echocardiography before therapy was started. Physicians excluded 11 patients. Of the 65 remaining patients, 19 declined participation. Of the 46 patients included for analysis, 34 (74%) agreed to be interviewed. The investigators examined 34 (74%) patients; examination data were abstracted from the medical record for 10 of the remaining 12 patients.

Patient Characteristics and Generalizability

Baseline characteristics did not differ significantly between study patients and nonenrolled eligible patients (Table 1). Compared with a random sample of patients who used diet medications but did not have echocardiography before therapy for comparison, study patients were older (P < 0.001), were more likely to be hypertensive (P = 0.04), and were more likely to have coronary artery disease (P < 0.001).

Development of Valvular Heart Disease

At baseline, eight patients (17.4%) had valvular abnormalities that met FDA criteria; seven (15.2%) had such abnormalities at follow-up. Additional patients had valvular changes that did not meet FDA criteria. Two patients with previous abnormalities that met FDA criteria regressed (Table 2). Two patients with normal mitral and aortic valves developed both mild aortic and mitral thickening at follow-up, two patients developed new mild aortic thickening, and two developed new mild mitral thickening.

Of 34 patients for whom data were available, 14 (41%) reported new or worsening cardiopulmonary symptoms or exercise intolerance after using diet medications. Of 44 patients for whom physical examination data were available, 24 (54%) had a murmur and 27 (61%) had a murmur or some abnormal physical finding that may reflect valvular abnormality (such as an abnormal heart sound or pedal edema).

Of the 46 study patients, 2 (4.3% [95% CI, 0.6% to 14.8%]) developed the primary outcome of new or worsening valvular disease. Both patients were female and had used fenfluramine-phentermine. Neither patient used other diet medications or serotonergic agents. The first patient, who was 52 years of age, had a bicuspid aortic valve and mild aortic regurgitation on a blindly read echocardiogram obtained 1.9 years before medication use. Her follow-up echocardiogram (obtained after 62 days of therapy) showed progression of aortic regurgitation to moderate severity and new mild mitral regurgitation. Her pulmonary artery pressure was elevated (39 mm Hg); no adequate baseline estimate was available. Clinical readings of baseline and follow-up echocardiograms differed from the blinded readings in that aortic regurgitation was read clinically as mild to moderate without interval change. The clinical and blinded readings were otherwise consistent. This patient took fenfluramine-phentermine for 307 days and reduced her initial body mass index (30.5 kg/m²) by 12%. With the exception of a heart murmur that was present before medication use, her interview and physical examination were unremarkable and did not suggest progression of valvular disease.

The second patient, who was 66 years of age, had hypothyroidism. An echocardiogram obtained 8.1 years before medication use was normal. A follow-up echocardiogram (obtained after 402 days of therapy) revealed new moderate aortic regurgitation and new mild mitral valve thickening. Her pulmonary artery pressure was elevated at 38 mm Hg on follow-up (no adequate baseline estimate was available). She used fenfluramine-phentermine for 250 days and reduced her initial body mass index (38.9 kg/m²) by 16%. She declined the medical interview and examination; review of her medical record revealed a normal cardiopulmonary examination and no reports of new cardiopulmonary symptoms after medication use.

When the 8 patients with FDA-defined valvulopathy before medication use were excluded, the risk for new valvular abnormalities was 2.6% (CI, 0.1% to 13.8%). Among all patients who took fenfluramine-phentermine, the risk was 8.3% (CI, 1.0% to 27.0%); after exclusion of those with existing valvular disease, the risk was 4.0% (CI, 0.2% to 24.9%).

Discussion

Top Methods Results Discussion Author & Article Info References

New or worsening valvular heart disease developed in 4.3% of a select cohort of patients who took fenfluramine or dexfenfluramine and had undergone echocardiography before beginning therapy. When patients with existing valvulopathy were excluded, the risk was 2.6%. Neither patient with the primary outcome had new or worsening cardiopulmonary symptoms or examination findings. Several patients showed progression and regression of valvular insufficiency and thickening that did not meet FDA criteria; the clinical relevance of these findings are uncertain.

A temporal relation between medication use and the development or progression of valvular disease was seen in two patients. However, one of these patients had aortic valve disease at baseline, which may have been responsible for disease progression. In addition, the discrepancies between the blinded and clinical echocardiographic interpretations suggest that these changes were subtle. The only patient who developed a clearly new valvular abnormality had a baseline echocardiogram that was obtained more than 8 years before medication use. Thus, the temporal relation in this patient is less convincing. These outcomes may have been coincidental and not causally related to use of diet medication.

The risk for new or progressive valvular heart disease that we report is much lower than that implied by previous prevalence studies. Several reasons may explain this discrepancy. We performed an incidence study examining the development of new or progressive disease after medication use. Consequently, our study is not comparable to previous prevalence studies, which may have detected pre-existing but unrecognized valvular disease. Because some of these previous studies are unpublished, it is unclear whether important methodologic differences contributed to higher disease prevalences. Our grouping of fenfluramine and dexfenfluramine users may have diluted the risk associated with fenfluramine-phentermine use. Referral bias may account for the low prevalence we report in that patients at higher risk for valvular disease may have been preferentially enrolled in other studies. The referral bias in our study design, which required a previously obtained echocardiogram for inclusion, would tend to result in a higher incidence of valvular disease. Finally, our study patients may have used medications for a shorter time than did patients in other studies, and there may be a dose-response effect.

Interpretation of these results is tempered by the study limitations. This was not a randomized, controlled trial, and the issue of confounding and selection bias must be considered. Enrolled patients were similar to nonenrolled eligible patients. However, study patients were older and had more comorbid illnesses than other patients taking diet medication; the risk for new or progressive valvular abnormalities is probably similar to or higher than that among diet medication users in general. Nonetheless, the small number of patients and the wide CIs raise the possibility that the actual risk for medication-associated valvulopathy was underestimated or overestimated. The small number of patients also limits our ability to draw conclusions about the sensitivity of screening by history and physical examination alone, although neither patient with the primary outcome showed any symptoms or examination findings that would have alerted their physicians. Likewise, we are wary of drawing conclusions from post hoc subgroup analyses.

The relatively low risk for medication-associated valvulopathy demonstrated in our study strongly supports the need for larger, carefully controlled studies to examine the true risks of short- and long-term fenfluramine and dexfenfluramine use and to study the cost-effectiveness of routine echocardiography former users of these medications.