Appetite-suppressant drugs and the risk of primary pulmonary hypertension

Summary

Appetite suppressing agents including phentermine, fenfluramine and dexfenfluramine are increasingly being used for the treatment of obesity. Phentermine is chemically related to the amphetamines and shares their potential for drug dependence and exacerbation of cardiovascular problems. Fenfluramine and dexfenfluramine are not amphetamine-like, but have been associated with rare cases of unexplained ("primary") pulmonary hypertension, some fatal. This case-control study was designed to look at risk factors for the development of primary pulmonary hypertension, in particular in relation to the use of anorexic drugs.

Methods

• Cases: Cases were prospectively and actively recruited from hospitals in four countries (France, Belgium, the United Kingdom and the Netherlands) and were considered if the diagnosis of pulmonary hypertension was first made by right heart catheterization between September 1, 1992 and September 30, 1994. Patients with secondary pulmonary hypertension were excluded. In Belgium, an attempt was made to find all new cases of primary pulmonary hypertension, allowing calculation of incidence.
   
• Controls: Four controls were selected for each case from the medical practice of the case patient's general practitioner and matched for sex, age and number of visits per year (greater than or less than two per year). When the case patient's GP could not serve as a source of controls, a physician practicing in the same region was used.
   
• Exposures: Detailed, face to face interviews with cases and controls were conducted by non medically trained interviewers who were unaware of the study's hypothesis. Exposure to antihypertensive drugs, oral contraceptives, thyroid extract and anorexic drugs was determined. The anorexic drugs examined were fenfluramine, dexfenfluramine and several amphetamine-like agents (diethylpropion, clobenzorex, fenproporex, mazindol and phenmetrazine). The index date for determining exposure was the first onset of symptoms (this same date was then used for each case patient's corresponding controls); exposures were only considered if they occurred before the index date and were classified as recent (within 12 months prior to the index date) or "past use" (greater than 12 months prior to the index date).

• Case patients: A total of 95 patients were included in this study, 80 with definite and 15 with probable primary pulmonary hypertension. The female:male ratio was 2.3:1, the mean age 44.7 years. 91% presented with dyspnea, severe in 2/3. Time from symptom onset to diagnosis was greater than one year in 2/3 of patients.

Two patients reported a family history of primary pulmonary hypertension, three had HIV (without AIDS), seven reported a history of cirrhosis, confirmed in four. None of these were found in the controls.
 

• Appetite-suppressant use: Among the 95 case patients, 31.6% reported definite use of appetite suppressants before the index date, vs. 7.3% among the 355 controls. This yielded an adjusted odds ratio of 6.3 (adjusted for multiple risk factors including use of cocaine, smoking and high body-mass index).

For recent use (within 1 year prior to the index date) the OR was 10.1, for past use only 2.4.

Use of anorexic drugs for less than 3 months was associated with an OR of only 1.8 (not significant). Use of appetite suppressant drugs for greater than 3 months (the principal outcome of this study) was associated with an adjusted odds ratio of 23.1.

When case patients with a family history, a history of cirrhosis, HIV or IV drug use were excluded, the adjusted OR increased slightly, to 8.6.

Only 2 case patients (2.1%) and 3 controls (0.8%) reported using amphetamine-like agents without fenfluramine or dexfenfluramine.
 

• Incidence estimate: From the Belgian data, where all incident cases of newly diagnosed primary pulmonary hypertension were assumed to have been captured, the incidence of the disease was estimated at 1.7 per million inhabitants per year.

In their discussion, the authors review a number of potential sources of bias and confounding.

• Patients with pulmonary hypertension might be more likely to remember exposure than controls (recall bias). However, the use of these agents among controls was close to what sales figures would have predicted.
   
• Patients using anorexic agents might be more likely to have their disease discovered earlier, due to closer surveillance. This is unlikely, since the time from symptom onset to diagnosis was similar in the two groups.
   
• If the disease had actually begun earlier than assumed, some of the exposures might not have antedated the disease onset and thus would not have been true exposures. The authors recalculated their results moving the index date back twelve months, which actually increased the odds ratio slightly.
   
• Finally, adjusting for obesity, a possible confounder, did not alter the results.

Editorial

An editorial (complete text from NEJM) by Drs. JoAnn Manson and Gerald Faich argues that the risk of primary pulmonary hypertension must be balanced against the health risks of obesity. They go on to estimate the risk-benefit ratio in the following fashion:

• From the Nurses' Health Study (Dr. Manson was lead author), the risk of death among those with body-mass indexes between 29 and 32 was 60 to 70 percent higher than among those with BMI's between 25 and 27. This yielded 1260 excess lives lost per million women per year due to obesity.
   
• Intentional weight loss lowers obesity-related risk factors and lowers all cause mortality (they cite a 1995 article by Williamson et al., from the American Journal of Epidemiology 141:1128-41).
   
• A 12-month trial of dexfenfluramine yielded specific values for weight loss achieved among the participants.
   
• Applying these values for weight loss achieved to the excess mortality observed in the Nurses' Health Study yielded 280 deaths prevented per million obese persons treated with dexfenfluramine per year.
   
• Assuming a background risk of 1.2 cases of primary pulmonary hypertension and an increase in risk with dexfenfluramine treatment of 23, there would be about 28 cases per million of PPH caused by use of this agent; assuming a mortality rate of 50% for the disease, there would be 14 deaths caused by dexfenfluramine per million patients treated. This yields a benefit:risk ratio of 280 lives saved for 14 deaths caused, or 20:1.

Comment

The editorial has generated controversy, reported in the media, because its authors have had financial ties to the pharmaceutical industry producing these drugs. This issue aside, I believe that the analysis presented in the editorial, which yielded a benefit:risk ratio of 20:1 is problematic, for a number of reasons.

• The number of deaths that can be prevented by weight loss was obtained by looking at the excess mortality due to obesity found in the Nurses' Study and assuming that these excess deaths could be prevented by weight loss. This is not a reasonable assumption. To take an example from another domain, ventricular ectopy post-myocardial infarction is associated with a higher death rate in the year post-MI. Suppressing VPC's however, with certain medications turned out to increase mortality in the CAST study (presumably through pro-arrhythmia). By analogy, what counts is not the excess mortality due to obesity but the decrease in mortality that can actually be obtained through weight loss.

One of the very few studies that prospectively looked at the change in mortality associated with voluntary weight loss (the above-cited AJE article) did, in fact, find that weight loss was associated with a 20% decrease in all-cause mortality, but only in subjects with obesity-related co-existing conditions. When all-comers were looked at, those with and without co-existing medical problems, the mortality was decreased by about 800 per million person-years (by my calculations from the article), less than the 1260 assumed in the editorial.
 

• Even assuming a significant decrease in mortality from intentionally lost weight, as shown in the study from the AJE, the method of weight loss in that study was not through derivates of fenfluramine but, presumably, in large part from diet and exercise. Diet and exercise provide additional health benefits which may not be matched, "pound for pound" by weight loss using anorexic agents. We cannot simply extrapolate the benefits of weight loss through lifestyle modification to weight loss using drugs.

• Finally, the excess mortality of 14 per million person years quoted here could increase if these agents are used for longer periods of time. Currently, dexfenfluramine is only FDA approved for treatment up to one year.

Randomized trials looking at these drugs vs. placebo, and at different drugs taken for different periods of time are needed soon, since the public demand for these medications is likely to increase rapidly. In the absence of randomized trials, close surveillance of patients receiving these agents for prolonged periods of time will be essential. Perhaps a national registry should be considered.

9/1/96

Reader comments

The Lancet has an editorial that relates to this study.

October 3, 1996

 The New England Journal of Medicine has published an editorial piece about the conflict of interest issue related to the editorial about this study. The authors of that editorial respond in a letter to the editor.

Journal Club on the Web main screen

Submit a comment about this article