Data Mining Approach Shows Promise in Detecting Unexpected Drug Interactions

Up to 1 million patients in the United States may be taking 2 medications that can lead to unexpected increases in blood glucose levels when used simultaneously. Data mining techniques have revealed that the combination of the antidepressant paroxetine and the cholesterol-lowering medication pravastatin may cause this adverse effect (Tatonetti NP et al. Clin Pharmacol Ther. doi: 10.1038/clpt.2011.83 [published online ahead of print May 25, 2011]).

“If a physician has a patient on these 2 medications and their diabetes becomes harder to control, the physician may want to consider changing the medications,” said principal investigator Russ Altman, MD, PhD, professor of bioengineering, genetics, and medicine at Stanford University.

To uncover the paroxetine-pravastatin association with high glucose effects, the investigators mined the US Food and Drug Administration's (FDA’s) Adverse Event Reporting System (AERS) (http://tinyurl.com/mbjsdt) for drug interactions with diabetes-related associations. After finding an intriguing indication that pravastatin and paroxetine taken together may influence glucose homeostasis, they retrospectively evaluated changes in blood glucose in 104 patients with diabetes and 135 without diabetes who had received both of these drugs. To accomplish this, they analyzed data in electronic medical record (EMR) systems from 3 institutions and assessed mean random blood glucose levels before and after treatment with the drugs. Administration of both pravastatin and paroxetine together was associated with an average increase in blood glucose levels of 19 mg/dL (1.0 mmol/L) overall and 48 mg/dL (2.7 mmol/L) in individuals with diabetes. Neither drug administered singly was associated with such changes in glucose levels, and no increase was seen with other combinations of selective serotonin reuptake inhibitors and statins.

“FDA AERS sometimes gets criticized as filled with biases and not useful. It can certainly be improved, but it has some useful information, for sure,” said Altman. “Also, EMRs were really critical in our study because they allowed us to validate our FDA-derived predictions at minimal cost.” Lead author Nicholas Tatonetti added that the expanding presence of EMR systems represents a new and growing opportunity to study drug effects in real time. “This is especially important in the case of drug-drug interactions where the effect may not appear until a very large cohort of patients has been exposed,” he said.

The researchers also looked at the impact of the 2 medications in laboratory mice that were fed a high-fat, high-calorie diet and became insulin resistant. Neither medication alone increased fasting glucose levels in these prediabetic mice, but when they were given paroxetine and pravastatin together for 3 weeks, their glucose levels increased dramatically.

“Random data mining of spontaneous reports often leads to blind alleys, but this creative study used the approach to generate a wholly unexpected hypothesis that the authors were then able to confirm in observational studies of patients' glucose levels in multiple sites, as well as in an animal experiment,” said Jerry Avorn, MD, professor of medicine at Harvard Medical School and chief of the division of pharmacoepidemiology and pharmacoeconomics at Brigham and Women's Hospital in Boston, who was not involved with the work. “This is a very interesting application of informatics employed on a large scale to identify new drug-drug interactions, which are generally not well spotted in clinical trials,” he added.

The approach has the potential to help expose numerous other unexpected adverse effects from different drug combinations. “Perhaps best of all, it doesn't require a super computer to do this type of research. Any modern laptop can run these algorithms in minutes,” said Tatonetti.