Lilly Publishes Study on How to Measure Placebo Effect In An Antidepressant Trial.
Wow. What a cool idea! Measure the placebo effct in a depression study!
I was looking for the HamD, a very commonly used depression scale, to see what the items are. To help me evaluate a study. I did a google scholar search, and came across a study that looked like it might have the actual items.
There are many of these scales. Typically, someone gives one of the provide responses to each of the questions: for the past two weeks, have you felt sad or blue? No, not at all, a little, somewhat, a lot.
You add the score for each question and that is your total depression score. If you score lower, you are less depressed. If you score higher, you are more depressed.
so, total score depends upon the various questions. What are the questions?
So, I pull up the study. PDFs-plus-internet is awesome for these types of things!
The study I look at is:The responsiveness of the Hamilton Depression Rating Scale. Journal of Psychiatric Research, 2000, v 34, pages 3-10. Authored by: Faries, Herrera, Rayamajhi, DeBrota, Demitrack, Potter.
The study notes the HamD scores across several antidepressant studies, comparing the drug to placebo. As is well-established, there will be a placebo effect: people taking the non-active pill get less depressed across a few weeks. As do the people taking the pill.
To declare that the pill has some effect, you simply look to see if the pill group has some degree of improvement beyond the placebo response.
The study does not have the various questions. So, My desired reason for looking at this study is not fulfilled. Bummer.
What it does have is an analysis of recognized subsets of the HamD questons, and how much each subset goes along with the treatment outcome or goes along with placebo effect.
So, here is an obvious depression-study, placebo-effect detail that has never occurred to me:
As well as getting more better, it is possible that the people getting less depressed due to pills versus placebo will have a different pattern of responses across the various questions that add up to the total depression score.
So, by analyzing questions within this scale, you might be able to figure out more specifically whether improvement is due to the med, or due to some other factors.
This study does just that. They report overall result for med- versus placebo, AND subscale results for three recognized sub-groupings of HamD items: "Bech," and "Maier Phillip." [they overlap each other but difer on a couple Qs.]
BOTH subscales were superior at distinguishing between the med and the placebo group than the overall HamD scale.
The authors conclude that, for med versus placebo studies, We depression researchers may be wiser to use the scales that seem to exclude some of the placebo response.
**My mind additionally throws in another idea: We depression researchers should maybe identify and include a few MORE items that are more closely responsive to placebo than med effect.**
This would help answer the question: to what degree is depression improvement due to placebo, and what degree due to the medication?
Add a few items to the HamD, and it goes from 17 to 20 or 21. Not much additional TIME. But a world of increased knowledge.
If it exists, it can be measured. If a placebo response exists, we can use our knowledge on deperssion symptomatology, and psychometrics, and develop a emasure of placebo response. To some degree, this study does just that. It is sub-optimal because the data were not specifically developed to do just that; it is more like: proof-of-concept.
to what extent has our observed improvement been seen with the questions that have been shown to be LESS responsive to placebo effects, in contrast to the items that have been shown to be MORE responsive to placebo effects?
Now, keep in mind: ALL questions measure depression. None alone is perfect. That is why you use several.
If it exists, it can be measured. But it will be measured imperfectly, since no measure is perfect.
It seems intuitive that some meds will have MORE of an antidepressant effect than others. If a med has more antidepressant effect compared to a med with less, you would see a greater response on that subset of questions. you could calculate the ratio between the "core," or less-placebo-influenced items, and the more-placebo-influenced items. A better antidepressant would be one with a greater ratio.
And yet another great idea occurs to me: you could look at this placebo-influece-subscale, and measure the degree that your specific study evoked a placebo, versus a med-based antidepressant effect, if you knew what a typical, desireable, med-based effet was: for example, if typical strong-antidepressant effect on a subscale is a change of 6 points in six weeks, and you get a good outcome OVERALL on the full HamD, and you do get a change of 6 points in your current study, then you know that your study had a normal amount of placebo influence. But if you get a good outcome, but a low response on the "core" subscale, then you know that, somehow, your study has a good amount of placebo effect built into it.
How? Could be any of many ways: very supportive, friendly, optimistic research assistants, clinicians wearing lab coats and ties versus wearing birkestocks and Hawai'ian shirts, Professional appeal of the med packaging, a prestigious institution, who knows.
With all of the drug studies we see, is this ever reported? discussed? No.
Why not? This has not occurred to anyone else? No one else read this 2000 study?
I know who IS aware of this study: Eli Lilly. Cuz 4 of the authors were Lilly employees.