{"id":119,"date":"2010-01-22T17:52:05","date_gmt":"2010-01-23T00:52:05","guid":{"rendered":"http:\/\/www.PsychologyLounge.com\/?p=119"},"modified":"2010-01-22T17:55:31","modified_gmt":"2010-01-23T00:55:31","slug":"new-study-shows-antidepressant-medication-fails-to-help-most-depressed-patients","status":"publish","type":"post","link":"https:\/\/www.PsychologyLounge.com\/new-study-shows-antidepressant-medication-fails-to-help-most-depressed-patients\/","title":{"rendered":"New Study Shows Antidepressant Medication Fails to Help Most Depressed Patients"},"content":{"rendered":"

A very interesting study recently published in the Journal of the American Medical Association <\/a>(JAMA) demonstrated very clearly that when it comes to antidepressant medication, the Emperor is wearing few if any clothes! The researchers did what is called a meta-study or meta-analysis. They searched the research literature for all studies that were placebo-controlled studies of antidepressants when used for depression. That means the studies had to include random assignment to either a medication group or a placebo (sugar pill) group. They eliminated some studies which use a placebo washout condition. (This means the studies first gave patients a placebo, and then eliminated all patients who had a 20% or greater improvement while taking placebo.) When they eliminated all studies that didn\u2019t meet their criteria, they were left with 6 studies of 738 people.<\/p>\n

Based on scores on the Hamilton Depression Rating Scale (HDRS), the researchers divided the patients into mild to moderately depressed, severely depressed, and very severely depressed. This is a 17 item scale that is filled out by a psychologist or psychiatrist, and measures various aspects of depression. It is used in most studies of depression. They then analyzed the response to antidepressant medication based on how severe the initial depression was.<\/p>\n

The two antidepressants studied were imipramine and paroxetine (Paxil). Imipramine is an older, tricyclic antidepressant, and Paxil is a more modern SSRI antidepressant.<\/p>\n

What did they find? They were looking at the size of the difference between the medication groups and the placebo groups. Rather than do the typical thing of just looking at statistical significance, which is simply a measure of whether the difference could be explained by chance, they looked at clinical significance. They used the definition used by NICE (National Institute of Clinical Excellence in England), which was an effect size of 0.50 or a difference of 3 points on the HDRS. This is defined as a medium effect size.<\/p>\n

What they found was very disheartening to those who use antidepressant medications in their practices. They divided the patients into three groups based on their initial HDRS scores: mild to moderate depression (HDRS 18 or less), severe depression (HDRS 19 to 22), and very severe depression (HDRS 23 or greater).<\/p>\n

For the mild to moderately depressed patients, the effect size was d=0.11, and for severely depressed patients the effect size was d = 0.17. Both of these effect sizes are below the standard description of a small effect which is 0.20. For the patients in the very severe group, the effect size was 0.47 which is just below the accepted value of 0.50 for a medium effect size.<\/p>\n

When they did further statistical analysis, they found that in order to meet the NICE criteria of effect size of a 3 points difference, patients had to have an initial HDRS score of 25 or above. \u00a0To meet the criteria of an effect size of .50, or medium effect size, they had to have a score of 25 or above, and to have a large effect size, 27 or above.<\/p>\n

What does this all mean for patient care? It means that for the vast majority of clinically depressed patients who fall below the very severely depressed range, antidepressant medications most likely won’t help. The sadder news is that even for the very severely depressed, medications have a very modest effect. Looking at the scoring of the HDRS, the normal, undepressed range is 0 to 7. The very severely depressed patients had scores of 25 or above, and a medium effect size was a drop in scores of 3 or more points compared to placebo patients. Looking at the one graph in the paper that show the actual drops in HDRS scores, the medication group had a mean drop of 12 points when their initial score was 25. That means they went from 25 to 13, which is still in the depressed range, although only mildly depressed. Patients who initially were at 38 dropped by roughly 20 points, ending at 18, which is still pretty depressed. And the placebo group had only slightly worse results.<\/p>\n

One interesting thing is how strong the placebo effects are in these studies. It seems that for depressions less serious than very severe, placebo pills work as well as antidepressant medication.\u00a0 Is this because antidepressants don’t work very well, or because placebos work too well? It’s hard to know. Maybe doctors should give their patients sugar pills, and call the new drug Eliftimood!<\/p>\n

So in summary, here are the main observations I make from this study.<\/p>\n