Bad enough that antidepressants fail to help an estimated one-third of people suffering from depression. Even worse is that it can take 6 to 8 weeks before that becomes clear: the patient dutifully swallows Zoloft after Zoloft or Paxil after Paxil, only to find after two months that she is no better off—at which point her doctor typically puts her on a different med, and the whole process of trial-and-error starts all over again. There’s got to be a better way—and now there may be.

Last September I wrote about a new use of EEGs—the decades-old technology that measures brain waves—in which psychiatrists compare the EEG of a patient to thousands of EEGs in a huge database that matches it to an effective treatment. (This is different from using EEGs to diagnose a mental illness, something that doesn’t seem to work, perhaps because there are many, many ways for a brain to have an underlying pattern of electrical activity that adds up to “depression” or “bipolar disorder” or other psychiatric disease.) CNS Response, the California company that runs the database, looks for matches between EEG and effective drug. In about 75% of cases, that produces surprising pairings—such as an anticonvulsive drug for a patient with depression—that the physician would never have thought of.

A new study being reported this afternoon at the annual meeting of the American Psychiatric Association finds another use for EEGs: predicting which patients will respond to the antidepressant they have just started. Rather than waiting for months, patients suffering from major depression—as nearly 15 million Americans do—take the drug for a week and then undergo an EEG (which is painless, noninvasive and relatively cheap, on the order of $150).

The study, led by Andrew Leuchter of UCLA and called BRITE (Biomarkers for Rapid Identification of Treatment Effectiveness), had 73 patients take the antidepressant escitalopram, which is sold as Lexapro and belongs to same category—selective serotonin reuptake inhibitors, or SSRIs—as Prozac and many others. That’s the quandary: all of the drugs supposedly work by targeting the brain’s serotonin system (which is actually a questionable claim, but that’s a story for another day), so which one will help a particular patient? Before starting the drug and again after taking it for 48 hours, for one week, and for two and seven weeks, the patients underwent EEGs. At the one-week visit, doctors assessed how well they were responding to the drug; the researchers also identified genetic markers that have been reported to predict how well patients will respond to SSRIs, and measured how much of the drug was in the patients’ blood, which is thought to be an indication of whether it is likely to work.

The bad news: the docs were terrible at predicting, based on how well the patients were doing after a week on Lexapro, whether the drug would alleviate their depression. The genetic markers fared no better. Neither did the blood levels.

But of the 38 patients who got a little better and the 28 who recovered completely by the end of the seven weeks, the EEG readings—measuring brain-wave changes after one week on the drug—did pretty well, predicting who would get a little better or even recover with 74% accuracy (compared to 51% accuracy for the docs’ evaluation).

“Early changes in frontal EEG signals carry important information about future clinical response,” Leuchter said in a statement, suggesting that EEGs have “the potential to help clinicians improve the care of patients suffering from depression.”

Caveats: the company that sells the EEG system, Aspect Medical Systems, funded the study, and Leuchter has been a paid consultant to Aspect, served on its board and received grant money from Aspect. (I have blogged before on how company-funded studies can be skewed.) And seven weeks is not exactly long-term. Still, anything that moves us beyond the current hit-and-miss approach to treating depression is to be welcomed.