Thursday, September 18, 2008

Psychotherapy Research Is Lame Too


In my previous post, I criticised the manner in which psychotherapy research is conducted. In this post, I discuss four (2 psychotherapy & 2 medication) studies. The emphasis will be on the populations used and how the external validity of the results (i.e., they arn't generalizable) are greatly compromised. The actual results of these studies won't be discussed; however, a reference for each study is included. A specific type of psychotherapy known as behavioral activation therapy (BAT or BA) has this acute major depression treatment study (1), and this two-year relapse prevention study (2). Aripiprazole has this 26 week bipolar I maintenance study (3), and this 74 week extension study (4).

The first BA study initially included 388 subjects who completed a comprehensive intake assessment. Based on the exclusion criteria presented in this post (5), 250 (64%) subjects were eligible for randomization; however, 9 declined participation. In all, 241 (62%) were included in this study. The majority were excluded because of "subthreshold" or "low severity" depression. What he have left are 241 subjects with pure MDD with moderate to severe symptoms. Those subjects were randomized to one of four arms: cognitive therapy (45), behavioral activation (43), paroxetine (100), or placebo (53). At the end of 16 weeks, 172 (71%) subjects completed this study. By treatment arm, 39 (86%) of the CT group completed the study, followed by 36 (83%) for BA, 56 (56%) for paroxetine, and 41 (77%) for placebo. One important caveat, the placebo arm was dissolved after week 8. This means for the remaining 8 weeks, there were only three active arms. If you subtract out the placebo arm, then only 188 (78%) patients with MDD received active treatment, and 131 (54%) of them completed the study. Approximately 1 out of 2 subjects lasted 16-weeks.

In the first aripiprazole study, 633 recently manic subjects were recruited. After exclusion criteria was enforced, 567 (89%) made it into the initial open-label 6-18 weeks stabilization phase. The study does not state why 11% were excluded. It's also important to note that 333 of the 567 subjects were from an aripiprazole acute mania study (i.e., 58% of subjects included were already shown to be responsive to aripiprazole). At the end of the stabilization phase, 361 (63%) subjects had discontinued (primarily for side effects, 22%), while 206 (36%) remained in the study. Out of those 206 subjects, 161 participated in the 26-week maintenance phase. In simpler terms, only 28% of the original 567 advanced to the actual area of investigation. 83 subjects were randomized to placebo, and the other 78 subjects were randomized to aripiprazole. At the end of 26 weeks, 28 subjects completed the placebo arm (34%), and 39 subjects completed the aripiprazole arm (50%). That's a total of 67 subjects, meaning, 88% of the subjects initially recruited dropped out of the study. Only 41% of the subjects who advanced to the 26-week phase remained in the study.

In the second BA study, those subjects who responded to acute treatment were eligible for this continuation study. 106 (61%) subjects out of 172 that finished the previous study were included. The subjects who originally received CT or BA did not receive continued treatment during the two-year follow-up period. These subjects who received paroxetine were re-randomized to either continued medication or switched to placebo. After the first year of follow-up, the paroxetine group was tapered off treatment and the placebo group was dropped from the study. At the end of year one, 55 (51%) subjects had either dropped out or relapsed (9 relapses each occurred in the three treatment arms, 12 in placebo). At the end of year two, only 46 (43%) subjects completed the study. Since the placebo arms were dropped at the half way point in each study (thus limiting comparisons with active treatments), only 167 (69%) subjects actually received active treatment, and only 27% of those subjects completed the final study.

In the second aripiprazole study, 66 of the 67 subjects who completed the 24-week study advanced to the 74-week maintenance phase. 27 subjects were from the original placebo arm and 39 subjects were from the original aripiprazole arm. At the end of the study, 22 (81%) of the placebo subjects discontinued while 32 (82%) of those treated with aripiprazole dropped out of the study. A grand total of 12 subjects out of the initial 567 completed the entire study. That's a paltry 2%. If you want to be more liberal and count only those subjects who initially entered the 26-week study (161), then a ginormous 7% of those subjects completed the study. That's sad when you remember that 333 subjects were already shown to have had a response to aripiprazole, and the remainder were stabilized on aripiprazole.

What I am trying to illustrate is how unimpressive these numbers actually are. Large percentages of subjects are lost even before the actual studies begin. Secondly, since these studied populations are not representative of actual clinical populations, the positive results in these studies are pretty meaningless. Prospective studies that follow subjects for one to two years are quite rare. However, when done, it's striking how very few subjects actually complete these studies. So BA was shown to be comparable to paroxetine after 16-weeks. Since the placebo arm was dropped after week 8, we have no meaningful comparison group. Although aripiprazole was shown to be a maintenance treatment, the numbers were so small in the end that the results become moot. Sadly, as far as these studies go, this is as good at it gets.

No comments: