While our study certainly has some limitations, the letter from Mr. Kalman and his colleagues is unfortunately characterized by incorrect statements and invalid analyses of the data. They initially refer to the age distribution of "reports" related to ephedra. However, as we clearly stated, our analysis was limited to adverse reactions, which are defined as "adverse events occurring with normal, prescribed, labeled, or recommended use" (1). Fewer than 25% of these occur in persons younger than 20 years of age (2). Although the TESS annual report lists only nine herbs, it includes totals for all single-herb and multiherb products. Ephedra-containing products accounted for 64% of all herb-related adverse reactions in the 2001 report, as reported in our paper.

Mr. Kalman and colleagues then made the mistake of determining mortality rates to be the number of herb-related deaths divided by the number of contacts to poison control centers. A mortality rate is determined by the number of deaths divided by the total exposed population in 1 year, and the exposed population can be estimated only from sales data or population-based surveys. The correct statement extrapolated from the TESS data would be that ephedra accounted for six deaths and less than 5% of herbal product sales in 2001, whereas all other herbs accounted for one death (from kava) and more than 95% of herbal product sales. Because spontaneous reporting has been estimated to detect less than 1% of all reactions to herbal products (3), these deaths associated with ephedra may be merely the "tip of the iceberg." Mr. Kalman and colleagues also referred to adverse reaction frequency as the proportion of adverse reactions among reports to poison control centers; however, this statistic has no relevance for comparing the safety of herbal products.

While not mentioned in Mr. Kalman and colleagues' letter, an important limitation of our study was the difficulty in estimating ephedra's percentage of herbal product unit sales. For this reason, we included a sensitivity analysis in our study that varied this percentage from 0.82% to 13.5%. Our estimation that ephedra was responsible for 0.82% of the herbal product market in 2001 was based on unit sales of products that were identified as "herbal formulas or singles" by a natural products information company (SPINS, Inc., San Francisco, California). After the article was published, the company realized that information about some products, which contain ephedra and other herbs but which the company classifies as "vitamins/supplements" or "grocery" items rather than "herbal products," had not been provided to us. When information from these additional categories was included, ephedra's percentage of herbal sales was 4.3% rather than 0.82% (the figure we originally published). Of note, the 4.3% figure is well within the range used in our original sensitivity analysis. When this higher estimate was used, the relative risk for an adverse reaction from ephedra compared with all other herbs decreased from 220 (95% CI, 200 to 240) to 40 (CI, 37 to 44). The natural products company has contacted us and verified that all of their ephedra and other herb sales data are captured within the three categories that were used to obtain the lower estimate.

We stand by our original conclusions: The risk for adverse reactions with ephedra is dramatically elevated compared with other herbs. Our results are consistent with a recently published systematic review, which found that ephedra is associated with an elevated risk for nausea, vomiting, psychiatric symptoms, autonomic hyperactivity, and palpitations (4).