Patients and their doctors should weigh those potential risks and benefits when choosing between the drugs, the study concludes.
It found Avandia had the lowest treatment failure rate -- 15% -- compared with 21% for Glucophage and 34% for Micronase.
But Avandia was linked to greater weight gain, higher levels of LDL ("bad") cholesterol, and more swelling than the other two drugs.
However, a journal editorial suggests that metformin, which is sold generically and under several names, might be the "logical" starting point for diabetes drug treatment.
Researcher Steven Kahn, MB, ChB, of the University of Washington and the Veterans Affairs Puget Sound Health Care System, and colleagues did the study.
Participants were mainly white and obese. They were, on average, in their mid-to-late 50s.
The patients were assigned to take one of three diabetes drugs -- Avandia, Glucophage, or Micronase.
Glucophage is a brand-name version of metformin, which is also sold generically and under other brand names including Glumetza, Fortamet, and Riomet. Micronase is a brand-name version of glyburide, which is sold also generically and as Diabeta.
The patients were followed for an average of four years.
The chance of drug failure was 32% lower with Avandia than with Glucophage and 63% lower than with Micronase, based on the study's threshold for drug failure.
However, all the drugs had pros and cons.
Though Avandia was linked to greater weight gain, higher LDL levels, and more swelling, it also came with extra benefits.
Patients taking Avandia were less likely to have gastrointestinal problems than those taking metformin. They were less likely to have blood sugar that was too low than patients taking glyburide.
Avandia is the newest and most expensive of the three drugs.
"The potential risks and benefits, the profile of adverse events, and the costs of these three drugs should all be considered" in choosing diabetes drugs, conclude the researchers.
The study was funded by Avandia's maker, GlaxoSmithKline.
High Dropout Rate
Editorialist David Nathan, MD, writes that the findings' clinical significance may be "less impressive" than it seems at first glance.
Nathan works at the Diabetes Center and Department of Medicine at Boston's Massachusetts General Hospital and Harvard Medical School.
He points out that a "surprisingly high" proportion of patients quit the study (about 40%).
That "weakens the results," Nathan writes. He notes that patients left the study for various reasons, not just because of drug side effects.
Nathan concludes that metformin "remains the logical choice" when starting diabetes drug therapy, given Avandia's "modest" blood sugar benefit, risk of fluid retention and weight gain, and higher cost.
Nathan reports having received grants from Novo Nordisk, which makes diabetes care products, and drug company Sanofi-Aventis, the maker of Diabeta. He has also received GlaxoSmithKline funding for an educational program.