Relating Subjective and Objective Pharmacovigilance Association Measures - Clustering Challenges in Biological Network - page 283

tuberculosis nos

lupus-like syndrome

systemic lupus erythematosus

disseminated tuberculosis
pulmonary tuberculosis

medication error

0.01 0.10 1.00 10.00

Reporting Ratio

Fig. 15.1. Pharmacovigilance map for the drug infliximab, obtained by plotting U ab vs. R ab for
each adverse event appearing in one or more records that list infliximab. The vertical line corresponds
to R ab =1and separates those adverse events showing a positive association with infliximab (points
to the right of the line) from those showing a negative association (points to the left of the line). The
dashed horizontal line corresponds to the Bonferroni significance limit discussed in the text.

the logarithm:

U ab =log 10 u ab = − log 10 ρ ab . (15.5)

Fig. 15.1 shows a plot of U ab vs. R ab for the drug infliximab and all adverse
events that appear in AERS records listing this drug, plotted with a logarithmic
scale on the horizontal axis. Each point represents a specific adverse event, and
those lying in the upper right portion of the plot (large values of both U ab and R ab )
exhibit a strong positive association with the drug (i.e., these adverse events appear
significantly more often than expected by random chance alone). Conversely,
points with R ab < 1 correspond to adverse events that appear less often than
expected by random chance alone, so points lying in the upper left portion of the
plot exhibit strong negative associations with the drug infliximab.
Heeley et al. [9] display their results somewhat similarly, as a plot of the χ 2
statistic for each drug/adverse event combination against its associated P ab value.
An important extension included here is a simple correction for multiple com-
parison, based on the Bonferroni approximation [21]. That is, the uncorrected