OBJECTIVE: Recent epidemiological studies identifying an association between some classes of gastric acid suppressants and Clostridium difficile colitis and community-acquired pneumonia prompted our analysis. Our objective was to retrospectively apply data mining algorithms (DMAs) to the Food and Drug Administration (FDA) drug safety database to see if they might have directed/redirected attention to the reported association of gastric acid suppressive drugs with C. difficile colitis and community-acquired pneumonia, prior to the published epidemiological findings that supported the association. DESIGN: Two statistical DMAs, proportional reporting ratios (PRRs) and multi-item gamma Poisson shrinker (MGPS), were applied to a spontaneous reporting system (SRS) database to identify signals of disproportionate reporting (SDRs). RESULTS: SDRs related to community-acquired pneumonia were observed for two proton pump inhibitors (lansoprazole and omeprazole), two H2 antagonists (famotidine and roxatidine), and one antacid (magnesium silicate hydroxide). For C. difficile colitis, an SDR was generated for one proton pump inhibitor (lansoprazole). CONCLUSIONS: Although our analysis suggests that there may be an association between the SDRs using SRS data and the epidemiological findings, these results may not have alerted public health professionals in advance of published studies to an association between proton pump inhibitors/gastric acid suppressants and C. difficile colitis or community-acquired pneumonia. However, the analysis reveals the potential utility of DMAs to direct attention to more subtle indirect drug adverse effects in SRS databases that as yet are often identified from epidemiological investigations.
OBJECTIVE: Recent epidemiological studies identifying an association between some classes of gastric acid suppressants and Clostridium difficilecolitis and community-acquired pneumonia prompted our analysis. Our objective was to retrospectively apply data mining algorithms (DMAs) to the Food and Drug Administration (FDA) drug safety database to see if they might have directed/redirected attention to the reported association of gastric acid suppressive drugs with C. difficilecolitis and community-acquired pneumonia, prior to the published epidemiological findings that supported the association. DESIGN: Two statistical DMAs, proportional reporting ratios (PRRs) and multi-item gamma Poisson shrinker (MGPS), were applied to a spontaneous reporting system (SRS) database to identify signals of disproportionate reporting (SDRs). RESULTS: SDRs related to community-acquired pneumonia were observed for two proton pump inhibitors (lansoprazole and omeprazole), two H2 antagonists (famotidine and roxatidine), and one antacid (magnesium silicate hydroxide). For C. difficilecolitis, an SDR was generated for one proton pump inhibitor (lansoprazole). CONCLUSIONS: Although our analysis suggests that there may be an association between the SDRs using SRS data and the epidemiological findings, these results may not have alerted public health professionals in advance of published studies to an association between proton pump inhibitors/gastric acid suppressants and C. difficilecolitis or community-acquired pneumonia. However, the analysis reveals the potential utility of DMAs to direct attention to more subtle indirect drug adverse effects in SRS databases that as yet are often identified from epidemiological investigations.