Phenylpropanolamine and the risk of hemorrhagic stroke. Kernan WN, Viscoli CM, Brass LM, Broderick JP, Brott T, Feldmann E, et al.
Overall Study Question
Over the past 20 years, more than 30 case reports have suggested a causal relationship between phenylpropanolamine (PPA), a common ingredient in over-the-counter appetite suppressants and cold remedies, and hemorrhagic stroke. To characterize this relationship, a multi-center, nested case-controlled study was conducted.
The three objectives of the study were to
1) estimate the association between hemorrhagic stroke and the use of appetite suppressants containing PPA and any first use of products containing PPA among women,
2) estimate the association between hemorrhagic stroke and the use of any product containing PPA among both men and women, and
3) to estimate the association between hemorrhagic stroke and type of PPA and type of PPA exposure among both men and women.
Are the Results of the Study Valid?
Were there clearly identified comparison groups that were similar with respect to important determinants of outcome, other than the one of interest?
No. Patients with symptomatic subarachnoid hemorrhage or intracranial hemorrhage were identified and matched with two controls for each subject according to age, sex, and race through random digit dialing.
Baseline demographics, however, were significantly different between subjects and controls. Compared to controls, study subjects were more likely to be black, have a lower level of education, smoke, have hypertension, have a family history of hemorrhagic stroke, report regular alcohol use, and report cocaine use the day of or before the stroke. Study patients were also more likely to have used agents containing caffeine or nicotine and less likely to have used nonsteroidal anti-inflammatory drugs in the period three days prior to the stroke. To compensate for this, the investigators adjusted for all measured and significant potential con-founders that, when added to the conditional logistic model, resulted in at least a ten percent variability in the matched odds ratio. The only variables that met these criteria were race, history of hypertension, and smoking status. The rest of the potential con-founders were deemed to be insignificant.
Were the exposures and outcomes measured in the same way in the groups being compared?
Yes. For each patient the focal time or index day was defined as the calendar day that marked the onset of symptoms. Patients were then categorized according to PPA use. The window of exposure was defined as 72 hours preceding the index day. Patients who had used PPA-containing products less than 24 hours prior to the index day, but not during the 14 days prior, were labeled “first users” and were included in a sub-analysis.
Yes. For each patient the focal time or index day was defined as the calendar day that marked the onset of symptoms. Patients were then categorized according to PPA use. The window of exposure was defined as 72 hours preceding the index day. Exposure to PPA was ascertained through direct patient interview with a structured questionnaire when possible, or by telephone interview if a meeting could not be arranged within 30 days of the index day. During the interview, patients were asked to produce all medications taken two weeks prior to the index day, or identify them from a book containing photographs of packages. The investigators claimed to limit the risk of recall bias by conducting a structured interview, blinding patients to the study hypothesis, and conducting phone interviews if meetings could not be arranged within 30 days. The ratio of patients interviewed in person to those who were interviewed by phone was not stated. Patients who had used PPA-containing products less than 24 hours prior to the index day, but not during the 14 days prior, were labeled “first users” and were included in a sub-analysis.
Was the follow-up sufficiently long and complete?
Yes. Since patients were identified retrospectively (i.e., after a stroke had occurred), long-term follow up was unnecessary. However, control subjects were identified within 30 days of the patient’s stroke to account for potential seasonal differences in PPA exposure.
Is the temporal relationship correct?
Is there a dose-response gradient?
Potentially. The median dose of PPA in current male and female users from the control group was 75 mg. In a supplementary analysis, the investigators calculated point estimates of the odds ratio for those patients using less than or greater than a 75 mg dose. The odds ratio for the higher dose approached significance (2.30, 95% CI 0.96-5.54, p=0.06), indicating that there may be a dose effect, however this cohort analysis did not have adequate power to detect such a difference.
What were the Results?
How strong is the association between exposure and outcome?
Twenty-seven of 702 (3.8%) hemorrhagic stroke patients and 33 (2.4%) of 1376 control subjects had used PPA less than three days before the index event. The investigators found that women having a hemorrhagic stroke were more likely to have used PPA-containing products in the prior three days than the control group, yielding an odds ratio (OR) of 1.98 (95% CI, 1.00-3.90, p=0.05). Furthermore, women having a hemorrhagic stroke were more likely to have used PPA-containing appetite suppressants than the control group (OR=16.58, 95% CI 1.51-182.21, p=0.02). This finding remained significant when both men and women were included in the analysis despite the fact that no males used PPA-containing appetite suppressants (OR=15.92; 95% CI, 1.38-184.13, p=0.03). No significant difference was found among males who used PPA products.
How precise is the estimate of risk?
The use of conditional logistic models for the statistical analysis of this study was appropriate, as this procedure permits the investigator to control for potential confounding differences between the two groups. The difference between the two groups at baseline, however, raises questions about the internal validity of this study. Although the authors claim to have controlled for any measured and significant confounder that varied the results by more than 10%, this study may have involved a sample size that was too small to accurately identify significant con-founders assuming that no variables were overlooked during the data collection. Only a prospective, randomized, clinical trial with comparable groups of subjects can confirm such an association. Such a trial is unlikely to be conducted because of ethical considerations.
Will the Results Help Me in Caring for My Patients?
Are the results applicable to my practice?
Yes. This study suggests that the use of PPA, especially as an appetite suppressant, may increase the risk of hemorrhagic stroke in female patients. It is unclear whether there is an increased risk among males who use PPA in cough and cold products or appetite suppressants as only six of 319 male patients and 13 of 626 male controls used PPA three days prior to the index event.
What is the magnitude of the risk?
The results of this study suggest that one of every 107,000 to 3,268,000 women will experience a hemorrhagic stroke, should they use appetite suppressants containing PPA. Although this risk may seem small, the U.S. Food and Drug Administration claim that in 1999, six billion doses of medicines containing PPA were sold in this country alone. This would suggest that a significant number of people could be at risk for this serious adverse event.
Should I attempt to stop the exposure?
As a result of this study, PPA-containing products will no longer be available for sale in Canada or the United States. Whether this rapid response is warranted is debatable. Ideally, a prospective, placebo-controlled trial with better matched subjects and controls would be able to confirm or disprove the results of this study. Regardless, the FDA has answered this question for us.
While this study demonstrated an apparent increased risk of hemorrhagic stroke among women who use PPA-containing products (primarily appetite suppressants) it was not powered to detect a difference among men. Despite attempts to control for baseline differences between groups in the statistical analysis, there were significant differences between the experimental and control groups. At this time it may be prudent to avoid PPA-containing products for both men and women but further investigation is warranted into risk factor identification and differentiating risk according to frequency and type of use (i.e., daily use of appetite suppressants versus short term use of cold remedies).
The information from this study would be more useful if the mechanism behind the increased risk of stroke was understood, and if differences in risk between men and women could be better elucidated. However, considering that the incidence of PPA use among stroke patients appears to be very low, it is unlikely that a study with adequate power would be feasible. Further in vitro study and perhaps animal model studies need to be done to identify the mechanism for this serious adverse event.