Combining Paxil and Pravachol Can Raise Blood Sugar

The side effect -- which the authors said could push glucose-intolerant people into a diabetic state or make existing cases of diabetes harder to manage -- was unexpected, and its discovery illustrates the power of electronic health records to help bring to light previously unknown problems with medical treatments.

The authors estimate that as many as 715,000 people may be taking the two drugs, paroxetine and pravastatin, marketed as Paxil and Pravachol. But perhaps more notable are the methods used.

Researchers at Stanford University identified the drug combination as a possible concern by combing a massive Food and Drug Administration database of side effects for signs of diabetes-related drug interactions. They quickly and cheaply tested their theory using hundreds of thousands of electronic health records for patients at Stanford University Hospital, Vanderbilt University Medical Center, and the hospitals of Boston-based Partners HealthCare.

The new study offers a peak into the "totally unprecedented'' research potential of such large, sophisticated patient databases, said Dr. David Blumenthal, former national coordinator for health information technology .

The electronic records make it possible to study a whole population of patients in real time, "in the wild, so to speak, as we are cared for, because that's the most realistic study sample,'' said Dr. Isaac Kohane, an author of the study and co-director of the Harvard Medical School Center for Biomedical Informatics. "You can come in with a question and literally, in weeks rather than years, answer important epidemiological questions.''

Nicholas Tatonetti, a doctoral student at Stanford, developed algorithms to search the FDA database, looking not for specific reports of an increase in blood sugar but for the associated symptoms, including fever and fatigue. He came up with a list of possible problematic combinations. When his adviser, Dr. Russ B. Altman, looked at the list, pravastatin and paroxetine jumped out at him. He knew they were each used by millions of US patients.

But the FDA database is so big and so messy that it often produces false results, Altman said. If validating their concerns required scouring paper records -- like those still used by many doctors -- or enrolling patients in a costly clinical trial, his lab never would have pursued it, he said.

Instead, the researchers searched the Stanford hospital electronic records for a specific kind of patient, one who had started one of the drugs and had a glucose test, then started the second drug and had another glucose test. They found nine non-diabetic patients who fit the bill.

They reached out to colleagues at the other two institutions who found another 127 non-diabetic patients, the bulk of whom came from Partners hospitals. Overall, the patients taking both drugs experienced an average increase in glucose of 19 milligrams per deciliter after starting on the second drug, and the increase was much greater among 239 diabetic patients taking both medications.

"Given the power of computing, you can do this study in minutes if you've got a big structure like this,'' Blumenthal said. "You could do 100 studies in an hour instead of doing one study in a year…. And, by the way, you can do this not just with drug events. You can do it with the relationship between treatments and symptoms. You can track the spread of an influenza event.''

For a complete list of Drugs That Can Affect Blood Glucose Levels (pdf available) just follow this link.