Most of the case reports suggested by Dr. Salem and colleagues were already included in our review (n = 10) (1). Six of the suggested reports were excluded because of use of cytotoxic drugs, pancytopenia, or a neutrophil count nadir of 0.5 x 10⁹ cells/L or greater. Only 2 of the listed case reports (on carbutamide and nifedipine) were not identified because they were not listed as case reports in the MEDLINE publication type field. As causality is "possible" in these reports, the inclusion of these cases during the next database update will not cause any change in our Table 2. We believe that the first report of drug-induced agranulocytosis for a drug is most likely to be published in English-language journals with high international impact, and thus, those have been included in our review.

In contrast to Dr. Ibáñez and colleagues, we believe that the Council for International Organizations and Medical Sciences time relationship criteria are important to use (2). First, if the time window is too narrow, a delay between onset of agranulocytosis and diagnosis might obscure the causative drug. Second, some drugs associated with agranulocytosis have long half-lives and thus may be present in the body even more than 1 month after drug withdrawal. With respect to acetaminophen, an increased risk for granulocytosis has been reported (3), and this side effect is also listed in the German summary of product characteristics (4). We believe that treatment with granulocyte colony-stimulating factor or granulocyte macrophage colony-stimulating factor is also advantageous for primary symptomatic patients: If we analyzed the duration of neutropenia in these patients, it was borderline significantly different between treated and untreated patients if their neutrophil count nadir was less than 0.1 x 10⁹ cells/L (P = 0.063) but not if it was 0.1 x 10⁹ cells/L or greater (P = 0.77). If all primary symptomatic and asymptomatic patients were analyzed together, treated patients had a lower duration of neutropenia if the neutrophil count nadir was less than 0.1 x 10⁹ cells/L (P = 0.019) but not if it was 0.1 x 10⁹ cells/L or greater (P = 0.199).

Like Dr. Mossad, we also identified several case reports of ganciclovir-induced bone marrow damage, including agranulocytosis. However, "nonchemotherapy drug–induced agranulocytosis" generally refers to drugs causing agranulocytosis by noncytotoxic, idiosyncratic mechanisms. This categorization was difficult for some drugs. In the case of ganciclovir, the mechanism of action (inhibition of viral and human DNA polymerase), the high frequency of neutropenia listed in the summary of product characteristics (>10%), and the bone marrow suppression after overdosing also mentioned in the summary of product characteristics, suggest a nonidiosyncratic mechanism of neutropenia and agranulocytosis. Of note, other drugs, such as azathioprine, methotrexate, or colchicine, were also not included because of their known direct cytotoxic properties.