Data analysis
We selected the reporting odds ratio (ROR) for the detection of signals associated with antibiotic-induced agranulocytosis. As shown Fig. 1, the ROR and 95% confidence interval (95% CI) were calculated from a two-by-two contingency table. Signal detection was defined as follows: ROR greater than 1, lower limit of the corresponding 95% CI greater than 1, and case number greater than 2 [8]. Conversely, the condition of ROR less than 1 was considered no exposure-event association. The adjusted odds ratio was not calculated by multivariate analysis because of the small number of reports of agranulocytosis. Statical analysis was conducted with SPSS Statistics version 27 (IBM Japan, Tokyo, Japan).
The median of period until agranulocytosis onset from the time of first administration for each patient was evaluated using the Weibull shape parameter (WSP), which does not require a reference population [9]. The elapsed time from the first administration was calculated by subtracting the first administration date from the agranulocytosis expression date. Fifty cases were excluded because the date of first administration or onset of agranulocytosis was not available. The combination of WSP β and 95% CI could be interpreted as follows: when β was equal to 1 (random failure) and 95% CI of βincluded the 1, the hazard was estimated to be constant over time; ifβ was greater than 1 (wear out failure) and the 95% CI (lower) of β exceeded 1, the hazard was considered to increase over time; if β was less than 1 (early failure) and the 95% CI (upper) ofβ was less than 1, the hazard was considered to increase at an early stage and then decrease. Time-to-onset analyses were performed using JMP version 15 software (SAS Institute, Cary, NC, USA).