Wednesday, June 07, 2006

STAR*D is Bunk

There has been plenty of news over the past year about the long-awaited and much-anticipated STAR*D trial. It was designed to be a study to address real-world approaches to treating depression, in real-world settings.

The major articles are:

Trivedi MH et al. Evaluation of Outcomes with Citalopram for Depression Using Measurement-Based Care in STAR*D: Implications for Clinical Practice. American Journal of Psychiatry. 163:28-40 2006

Rush AJ et al. Bupropion-SR, Sertraline, or Venlafaxine-XR after Failure of SSRIs for Depression. New England Journal of Medicine. 354:1231-1242 2006


Trivedi MH et al. Medication Augmentation after the Failure of SSRIs for Depression. New England Journal of Medicine. 354:1243-1252 2006

Many reviews of the articles are already readily accessible, so I won't describe the articles here.

There are many very interesting pieces of data from these studies. One of the things that I found most interesting was the epidemiological survey of people who tend to get better with citalopram (young, educated, privately-insured women who make a decent amount of money).

Some of my concerns about the articles are:

  • The authors completely botch the use of the term remission. The STAR*D authors think that remission can be determined at a single point in time. They say that scoring below 7 on the HRSD-17 or below 5 on the QIDS reveals remission. This is not true. Remission can only be determined if a patient maintains that low score over time! For all we know, the "remitters" may have become depressed again the day the trial stopped. Remember, also, that remission is not cure. The definitions of "response" and "remission" were co-opted by psychiatry from oncology. If we co-opt the words, we must also maintain their actual meanings.


  • Speaking of the HSRD and the QIDS, take a gander at the "switch" paper, by Rush. The authors are running their trials from two completely different outcome measures! Look at page 1235: remission is based on a HRSD-17 score of less than 7. This is a completely different outcome measure from the primary outcome measure used in the first-step trial (the citalopram -only one): QIDS-C < 5. In a set of trials, why would the authors use different primary outcome measures?


  • It gets worse! In the "augmentation" trial by Trivedi, there are two different primary outcome measures! Come on, guys! Which is it? Look at pages 1244 and 1245. Page 1244 states that the primary outcome is based on the QIDS, whereas page 1245 states that the primary outcome is based on the HSRD. Did the authors base their results and conclusions on whichever scale gave them the best data?


  • Remission rates were somewhere between 28% and 33%. This isn't too great. I know many of you would say "at least we have that." I wouldn't be so sure. Take a look at the augmentation data. The augmenting drugs were bupropion and buspar. Buspar?? As an antidepressant? Buspar hasn't shown efficacy as an antidepressant. In this trial, it is effectively a placebo!. Here's the big point: we do have a placebo arm! Now, compared with placebo, the bupropion-augmentation remission rate was not significantly different: 29.7%-30.1%. Not much different from the citalopram remission rate, huh? Is this all placebo effect?



I should probably stop there. There are many very interesting things about these trials, most of which have not received attention so far (teaser...look at the suicide data). I will write about those later.

What do you think?

6 Comments:

Anonymous Roy said...

I'll reply to 2 points, both about remission.

It makes sense to apply a time component to "remission". Since one must meet criteria for two weeks to be "depressed", it makes sense that one should meet remission criteria for two weeks.

On the other hand, if we appy the oncology comparison, if one is cancer-free at Time1, what more do you need? The time component applies here because one may appear cancer-free, but time will show if this is, in fact, the case. I suppose, the same with depression. One could appear depression-free at Time1, but tincture of time will demonstrate if that remains true.

Thursday, June 08, 2006 8:58:00 AM  
Blogger bgp said...

It is amazing how little attention is paid to the idea of remission as a temporal phenomenon in psychiatry. I like your "two-week" proposal.

If one is cancer-free at Time1, but not at Time2, then we haven't treated the cancer, we just pushed it back (there may have been a few cancerous cells left following a treatment, too small for visualization on MRI or PET). It came back. By definition, the patient suffered a recurrence and was never in a state of remission (requiring probably several months of cancer-free survival, depending on the cancer. In rheumatology, for rheumatoid arthritis, remission involves a period of two months without symptoms). So, in response to "what more do you need?" there needs to be a persistence of freedom from cancer for the state of remission. Similarly for depression remission, there needs to be a persistence of freedom from depression.

"One could appear depression-free at Time1, but tincture of time will demonstrate if that remains true." Right. That's just what I'm getting at.

Thursday, June 08, 2006 6:34:00 PM  
Blogger The Neurocritic said...

You said:

"The authors are running their trials from two completely different outcome measures! Look at page 1235: remission is based on a HRSD-17 score of less than 7. This is a completely different outcome measure from the primary outcome measure used in the first-step trial (the citalopram -only one): QIDS-C < 5."

If you had designed the study, which of the two outcome measures would you have chosen? Inquiring non-clinician minds want to know!

Thanks.

Wednesday, June 14, 2006 3:51:00 PM  
Blogger bgp said...

So, the STAR*D QIDS and HRSD can be found on this Pitt site.

I haven't used either for research purposes. And I've only used the HAM-D in the clinical setting, not the HRSD. Looking over each, though, I like the QIDS because it allows for typical OR atypical depression (especially QIDS items 10 and 12).

I can't say that I'm opposed to either for research purposes, as long as the researchers are consistent with their outcome measures between trial arms. It's like wanting to evaluate student intellectual ability using the SAT and ACT. Both are fine (I know, there are psychometric problems with each), but you can't directly compare an ACT score with an SAT score. And if you're designing a trial to look at aptitude, you wouldn't design your arms with different outcome measures. You'd stick with either the SAT or the ACT in all arms.

Wednesday, June 14, 2006 5:40:00 PM  
Blogger The Neurocritic said...

Thank you for the link.

I can see it now. If I ever conduct a research study on depression using clinical outcome measures, the Acknowledgements would read, "The Neurocritic (anonymous blogger) thanks bpg (anonymous blogger) for helpful discussions."

Wednesday, June 14, 2006 8:08:00 PM  
Blogger The Neurocritic said...

Oops, I meant bgp...

Wednesday, June 14, 2006 8:12:00 PM  

Post a Comment

Subscribe to Post Comments [Atom]

Links to this post:

Create a Link

<< Home