The "real-world" study of 3,671 patients with major depression showed that no one drug or class of drugs works any better, the researchers said. The right combination must be found for each individual patient.

"The good news is that two-thirds of people can be relieved of their depression if they can hang in there for up to four treatment steps. That's pretty significant for a tough illness," said Dr A John Rush, a professor of psychiatry at the University of Texas Southwestern Medical Centre, who led the study.

But the bad news, he said, is that until patients are completely symptom-free, they risk relapse.

Writing in the American Journal of Psychiatry, Rush and colleagues said their findings come from the $35 million, six-year Sequenced Treatment Alternatives to Relieve Depression, or STAR*D, study paid for by the National Institute of Mental Health.

All the patients started on the same drug -- Forest Laboratories Inc.'s Celexa, a drug in a class called SSRIs, for selective serotonin reuptake inhibitors. This drug sent 37 percent of patients into remission.

Patients who did not get completely well were randomly assigned to take other SSRIs or Wyeth's Effexor, a drug in a similar class called selective serotonin/norepinephrine reuptake inhibitors or SNRIs.

COGNITIVE THERAPY

Patients could also add cognitive therapy -- a type of psychological counselling -- and this second step helped an additional 31 percent of patients.

Those who were still not cured went to one of four treatments: nortryptyline, an antidepressant in an older class called tricyclics; Remeron, also known as mirtazapine, which is a noradrenergic and specific serotonergic antidepressant; lithium; or T-3, a thyroid hormone.

This step helped 13 percent of patients. Then a fourth step involved yet another class of antidepressants called MAO inhibitors or a combination of the other drugs.

"It didn't turn out that there was a definitive large winner," Rush said in a telephone interview.

"It is almost as if there are different kinds of depression."

Rush said only a few patients opted for cognitive therapy but his group will publish findings on that later.

"Our treatments don't produce a 100 percent remission rate," he said. "That means we have got to find newer treatments or use the treatments that we have more effectively."

His team will look at genetic samples from the patients to see if there is a way of predicting who will respond to which treatment.