Results and Discussion
We extracted 72,723 reports of suspected AEs related to antibiotic treatment among the 1,048,576 reports in the JADER database between April 2004 and January 2021. The number of reports of agranulocytosis was 3,221. As shown Table 1, 10 of 60 antibiotics were detected as signals; the RORs (95% CIs) for ampicillin/sulbactam, amikacin, cefmetazole, cefozopran, clindamycin, ciprofloxacin, imipenem/cilastatin, kanamycin, teicoplanin, and vancomycin were 2.65 (1.79–3.80), 2.49 (1.91–4.34), 4.48 (2.27–6.92), 2.77 (1.88–3.95), 1.64 (1.04–2.47), 2.01 (1.40–2.82), 2.78 (2.11–3.60), 6.05 (2.16–13.7), 2.05 (1.31–3.07), and 3.54 (2.73–4.54), respectively. Because neutrophils are produced in bone marrow, we speculated that antibiotics with high migration to bone marrow may have a high risk of inducing agranulocytosis before data mining. However, signals were not detected for antibiotics with high cerebrospinal fluid transferability, such as ceftriaxone [10] and cefotaxime [11], whereas teicoplanin [12], an antibiotic with poor cerebrospinal fluid transferability, was detected. Therefore, these findings suggested that the risk of agranulocytosis could not be clarified by analysis of migration to the bone marrow only. A retrospective study reported that the development of vancomycin-induced neutropenia was not associated with the dosage, trough concentration, or administration period [13]. Although other antibiotics also need to be investigated, antibiotic-induced agranulocytosis may not be a cumulative toxicity.
The incidence of neutropenia induced by vancomycin is approximately 2–18% [14, 15]. In contrast, Smith et al. reported that linezolid, an anti-methicillin-resistant Staphylococcus aureus (MRSA) agent, may be safe and effective for the treatment of febrile neutropenia [16]. In addition, carbapenems targeting Pseudomonas aeruginosa are empirically used for the treatment of febrile neutropenia [17], indicating that when using antibiotics with anti-Pseudomonas aeruginosa or anti-MRSA activity in the empirical treatment of high-risk patients, such as those with febrile neutropenia, it may be better to preferentially use antibiotics with the same spectrum for which no signal was detected in this study.
In the clinical setting, aminoglycosides are often administered withβ -lactam antibiotics to induce synergistic effects [18, 19]. Moreover, infective endocarditis caused by MRSA may be treated with vancomycin or teicoplanin in combination with aminoglycosides [20]. In the current study, because amikacin and kanamycin were detected as signals, we speculate that it may be better to use gentamicin as the aminoglycoside in combination therapy.
In recent reports of SRS data mining, WSP analysis has been performed to evaluate the time-to-onset data for target AEs [9]. For the time-to-onset analysis, we extracted 292 cases for the 10 antibiotics with signal detection (Table 2). The median durations (interquartile ranges) of agranulocytosis onset following treatment with amikacin, cefmetazole, ciprofloxacin, and kanamycin were within 1 week of the first treatment. The difference in the onset time may have been related to the variations in pharmacokinetics of each antibiotic and the onset mechanism of agranulocytosis; however, the details are unknown, and further research is needed. Although the 95% CIs of WSP β for eight antibiotics (excluding amikacin and ciprofloxacin) were over and excluded 1, indicating a wear out failure type, the profiles of amikacin and ciprofloxacin were consistent with the random failure type (Table 2). Accordingly, time-to-onset analysis using WSP is likely to be a useful tool for determination of the specific safety monitoring period for agranulocytosis.
Data mining using the JADER database has some limitations that should be considered because the JADER database has several biases. First, because the JADER database does not contain information for control patients, the intensity of RORs cannot be quantified and compared among antibiotics. Second, there are also some reporting biases, such as under-reporting and lack of data. Third, the number of reports was small and may have been underestimated because the PT codes 10018687 (granulocytopenia) and 10018681 (granulocyte count decreased) were not included in the current study. Fourth, because agranulocytosis may be induced by severe infections [21], the agranulocytosis reported by SRS may not be caused by administration of antibiotics.
In conclusion, this comprehensive analysis was the first report evaluating the incidences of antibiotic-induced agranulocytosis. Despite the various limitations of using the JADER database, we identified 10 antibiotics that may be associated with high risk of agranulocytosis, suggesting that absolute neutrophil counts in patients taking these drugs should be monitored carefully in the clinical setting. Finally, further clinical studies are needed to verify the mechanisms through which ampicillin/sulbactam, amikacin, cefmetazole, cefozopran, clindamycin, ciprofloxacin, imipenem/cilastatin, kanamycin, teicoplanin, and vancomycin may induce agranulocytosis.