Questions

Clarifying antidepressant studies

From Gideon in Australia:

I was hoping you might have the time to answer a couple of questions.

Question:

I have been following your analysis of the STAR*D trial, and my key takeaway is that monoamine agonists appear to be effective in moderate to severe unipolar depression in the short term. However, I was somewhat confused by your comments on the Cipriani (2018) meta-analysis, where you clarify that monoamine agonists, with the exception of amitriptyline, are in fact not effective in depression. Is it the case that the Cipriani meta-analysis does not differentiate by severity of depression. If it did, then we would expect to see a more clinically significant effect in more severe cases?

Answer:

Yes, the overall meta-analyses of all antidepressant trials show a very small benefit, which is not clinically meaningful. This the case with the most recent by Cipriani and colleagues, although you have to look 200 pages into the online appendix to find the absolute effect size showing a small clinically meaningless benefit.

However, a prior meta-analysis (again deep in the text and ignored by the authors in their misleading claims in the media) showed that this overall small benefit hides a larger and more clinically meaningful benefit in severe depression.

This benefit is short-term though: acute, for 8 weeks, symptomatic; not long-term, not preventive, not showing disease modification.

When you look at the STAR*D trial, short-term benefit is replicated (overall about 3/4 of subjects), but long-term benefit is shown to be much less, about half as much as short-term response (overall about 1/3 of subjects).

These results are consistent with concluding that antidepressants may have some short term symptomatic benefits in severe cases, but they are not proven to have long-term preventive or disease-modifying effects.

Question:

Additionally, the NICE standard for a clinically meaningful effect—a 3-point reduction on the HRSD—seems rather minimal. If I were a patient, I would find such a change unimpressive, and it would be difficult to justify the costs, side effects, and therapeutic focus on medication. Even though the STAR*D trial demonstrated an additional 4-point improvement in moderate to severe cases, beyond the NICE threshold, I am curious to hear your perspective. In your experience, do patients perceive this level of improvement as meaningful in their day-to-day lives?

Answer:

You have to look at both absolute and relative benefit. In terms of overall absolute benefit, antidepressants produce a large improvement, with a Cohen’s d effect size of about one. That’s why clinicians think they are so great, and that’s why many patients think they are helped. In clinical practice, they seem to work great.

The problem is that in clinical practice, clinicians don’t see or know about confounding factors: other factors that can influence the outcome other than what they think is producing the outcome. The most important other factors are psychological effects, and the natural course of the illness. Randomization is the way that one can control for those other effects.

That’s why randomized clinical trials are needed: they can tell us what is true and not true in our clinical experience.

The randomized trials show that this benefit is not because of the drug, because it is almost the same with placebo, which reflects not just a psychological effect but also God’s effect, that is, Nature and the natural history of the illness. Depressive episodes are “episodes” because they always end, naturally, without any treatment, usually in a few months or so. In a 8 week trial of subjects previously depressed typically for 2 months or longer, God and Nature heal many. We doctors and drugs can’t take all the credit.

So the 3 point difference is really about showing that the drug can take the credit over Nature or doctors.

3 points is a small difference, nothing major: it’s like sleeping moderately better and eating somewhat more. That’s three points.

But remember that the absolute benefit is more like 15 points: So patients go from 25 to 10 with drug and from 25 to 13 with placebo. That’s the three point difference. They improve in both cases, though a bit more with drug. You’re right though, that’s a small benefit. 6 or 7 points would be more impressive but none of our drugs shown that benefit over placebo.

Esketamine - for all the hallucinogen (so-called psychedelic) lovers out there - only had a 4 point improvement over placebo in its large phase III trials. How transformative and revolutionary….

The standard of 3 points for FDA approval is a low bar, and yet so many drugs fail to meet even that standard.

Even a higher effect wouldn’t impress me though, because we’re still talking about a better tylenol for fever. We’re still saying that symptoms get better, not treating the cause of the symptoms, not getting long-term disease-modifying benefit, which very few psychiatric medications have been shown to have. Lithium is the exception, and clearly the most effective psychiatric medication.