On Wednesday, the US Food and Drug Administration approved the first consumer DNA test that promises to tell buyers which drugs might work best for them according to their genetic profile. The manufacturer, 23andMe, won FDA approval by being very careful not to overpromise what its test can do— so careful that the approval highlights the limited usefulness of the test and how much we still don’t know about this field of medicine.

The test is related to a segment of medicine called pharmacogenetics, which is the study of how genes influence how we react to drugs. Variations in our genes can affect how quickly we break down a drug, how much we absorb, or how efficiently it gets to where it needs to be in the body.

The 23andMe test will tell consumers about 33 genetic variants that are associated with how well common drugs — blood thinners like Plavix and Coumadin, for example — work, but it’s explicit that consumers should not change their medications based on these results. The test is only supposed to provide some potentially useful information that people can discuss with their doctor. To win approval, the company had to run a demographically representative study proving that 97 percent of users understand that the test is not a medical recommendation, says 23andMe chief legal and regulatory officer Kathy Hibbs. It’s not yet clear when the test will be available for purchase or how much it will cost.

Say you take the 23andMe test and bring the results to your doctor. The doctor still isn’t supposed to suggest changing medication until they have you genetically tested again by an independent lab. “It seems to me that if a patient has an interest in their pharmacogenetic profile that could impact medication decisions, they’re probably better off just asking the physician about what testing can be done, since confirmatory testing is expected anyway even if you got 23andMe,” says Boadie Dunlop, an Emory University psychiatrist who researchers biological markers that predict responses to drug treatments. (Dunlop is also a consultant for the personalized medicine company Assurex Health.)

At the moment, “there’s very little data supporting medication choice based on pharmacogenetics,” adds Dunlop. He’s also an author of a paper that was published in The Pharmacogenetics Journal in May that found that different pharmacogenetic tests can make different treatment recommendations.

The team analyzed four tests (CNSDose, Genecept, GeneSight, and Neuropharmagen) and found that they could reliably detect whether someone has a certain genetic variant, says Dunlop. But they don’t necessarily test every gene that could be important. The tests they researched also didn’t agree on what medical recommendations to make when they detected that a patient had a particular gene variant. “There’s a lot of supposition, and the companies have their own algorithms that are proprietary, and that’s where you get the most variation in recommendations,” Dunlop adds. In fact, a day after the 23andMe approval, the FDA warned against other non-FDA-approved tests that make such suggestions.

Even though 23andMe has been more successful than other companies in gaining FDA approval, that doesn’t mean its test is useful or necessary. Eric Topol, a geneticist at the Scripps Research Institute, points out that the genome variants they are analyzing are very limited. It’s simply too early. A review published by the American Psychiatric Association task force concludes that while initial data on the association of genetics and drugs has been promising, there’s not enough evidence to justify the widespread use of pharmacogenetic tests. When it comes to non-psychiatric drugs like Plavix and Coumadin, various studies suggest that there is promising evidence for treatment strategies linked to genetics, but also many unanswered questions.

Ultimately, we’re all interested in making medication more efficient. But even though products promising a future filled with personalized medicine are starting to hit drugstore shelves, the research still suggests we’re not at a point where our genes can help determine our dosages — yet.