Methods
Data sources
Obtain ASCII data packages from the US Food and Drug Administration
(FDA) Adverse Event Reporting System (FAERS) for case reports from the
first quarter of 2004 to the fourth quarter of 2022, which are divided
into seven tables, including patient demographic information (DEMO),
drug information (DRUG), reported indications for drug use (INDI),
treatment start and end dates (HER), and adverse events (REAC), Adverse
event outcome (OUTC) and reporting source (RPSR). Import data cleaning
and signal mining into SAS9.4 software. Using the common names of
acyclovir, valacyclovir, ganciclovir and famciclovir to retrieve the
obtained data, and obtain the ADR report records of the first suspected
drugs.
Disproportionality
analysis
Exclude duplicate reports, information uncertainty reports, and other
reports containing abnormal information from the standardized data, and
filter out data with a number of reports ≥ 3. According to the four grid
table (Table 1), calculate the corresponding ROR, PRR, IC, and the
corresponding 95% confidence interval (CI) lower limit, and count the
number of signals to exclude reports that do not meet the threshold
requirements. The larger the lower 95% CI limit of the ROR value, the
stronger the signal, indicating a stronger connection between the target
drug and the target ADR. In this study, in order to overcome the high
false positive rate caused by estimation errors in traditional frequency
methods to some extent, we chose signals that meet the standards of the
three algorithms mentioned above for research. The criteria for
determining the signal strength of the BCPNN method are: ① no signal
(-): IC ≤ 0; ② Signal weak (+): 0 < IC ≤ 1.5; ③ In the signal (++): 1.5
< IC ≤ 3; ④ Signal strength (+++): IC>3.