Spectrum IX by Ellsworth Kelly (2014), one of the artist’s last works
Seven medications beat placebo — but the tradeoffs matter as much as the efficacy
STUDY: Yalin N et al, European Neuropsychopharmacology 2026;108:112818, PMID: 41855620
STUDY TYPE: Network meta-analysis with updated systematic review
FUNDING: National Institute of Mental Health
Background
There are many ways to look at evidence. This paper compares meds for bipolar depression through the network meta-analysis lens, which means it lines them up as if they are all in a the same controlled trial, with the placebo groups lumped into one giant placebo arm.
There are a lot of assumptions in that view, so this isn’t the final word, but it adds to our perspective
The Study
The analysis included 101 randomized controlled trials of pharmacological agents for acute bipolar depression. Primary outcomes were reduction in depressive symptoms and risk of manic switch.
Seven medications beat placebo with moderate certainty, supported by large randomized trials. They all had similar benefits, with response rates ranging from 42-57% and remission rates from 28-44%. Where they diverged was in their risks, listed here:
- Olanzapine-fluoxetine and olanzapine: Most weight gain (significant in 1/3 of patients), and QTc prolongation
- Quetiapine: High sedation, modest weight gain
- Lurasidone: High akathisia (3-fold), modest weight gain
- Cariprazine: High akathisia (4-fold), modest weight gain
- Lumateperone: High sedation
- Lamotrigine: Well tolerated (but severe rash risk)
Lamotrigine is slow to build but holds steady over time. Its efficacy is probably bigger than meta-analyses suggest. Most of its trials only lasted five weeks, and at that point it barely works. Its bigger effects are at 6-8 weeks, and for long-term prevention.
Lamotrigine is also the best tolerated of the bunch. Most of the others have big tolerability problems. That means we’ll often need to draw from the low-certainty pile in practice. Let’s take a closer look at those.
CHANGE IN DEPRESSION: The 7 meds at the top are supported by large trials, while the 9 below rest on small trials. The lines represent the confidence intervals, so the true efficacy likely falls somewhere within those lines. K is the number of trials, N the total number of subjects.
Beyond Large Trials
Nine treatments were rated with “low quality evidence,” meaning they separated from placebo but only in small randomized controlled trials:
- Esketamine
- Pramipexole
- Vaproate alone or with glutamatergics (cytidine or memantine)
- Antidepressants: Venlafaxine and fluoxetine
- Supplements: Omega-3 and coenzyme Q10
Among those, venlafaxine had the highest risk of causing mania. Fluoxetine looks promising, but considering its close cousins sertraline, paroxetine, and citalopram failed in large trials of bipolar depression, it is likely a fluke. As a class, the antidepressants did no better than placebo in this analysis, and in other studies they worsen long-term mood outcomes for some, causing mixed states and rapid cycling.
Esketamine is likely effective, but is a short term treatment and this is a long-term illness.
Valproate is an interesting option. With FDA-approval in mania, we often forget it may also treat depression (and anxiety). Its side effects (sedation, weight gain, hair loss) will be a deal breaker for some, and it is risky in pregnancy. Still, it has a role, particularly in bipolar I and in patients with anxiety or alcohol use disorder.
That leaves two treatments that I view more favorably: Pramipexole and coenzyme Q10. Both have a sound theoretical basis — D3 agonism for pramipexole and mitochondrial health for coenzyme Q10. They are well tolerated and have a meaningful benefits. Even if their efficacy was only at the bottom end of the confidence intervals in this graph, they would still rank higher than many of the options backed by large trials.
But small trials often produce inflated effect sizes, so the confidence problem here is real. I have used pramipexole in over a thousand bipolar patients, and coenzyme Q10 in over a hundred. My guess is that pramipexole’s true effect is large, while coenzyme Q10’s is small.
Studies outside of bipolar back that up. Pramipexole had a large effect size in a large, 12-month trial of treatment-resistant depression. Coenzyme Q10, however, has only a small benefit in non-bipolar depression. Both are well tolerated and not known to cause mania, although pramipexole has a 1-3% chance of inducing hedonic dysregulation (ie, compulsive gambling) and a low risk of hallucinations (especially beyond 2-3 mg).
What Didn’t Work
Just as important is knowing what doesn’t work. Not just so you can avoid it, but so you can know what to deprescribe.
Those are listed below, and I’ll call out three where this analysis may have missed the mark:
- Lithium works. Its efficacy is hit-or-miss in acute depression, but is among the most effective for long-term prevention. It is especially effective in classic bipolar, where there are pure manias or hypomanias followed by pure depressions, and full recovery between the episodes (whether bipolar I or II).
- Modafinil may treat some symptoms. Yes, it failed in a large trial, but it likely treats symptoms that are important to patients but don’t move the needle on a rating scale (eg, fatigue, executive functioning).
- High-dose thyroid may work. It has a positive controlled trial in treatment-resistant bipolar depression, and another in rapid cycling.
Missing from the analysis are celecoxib, which worked in a small trial of treatment-resistant bipolar depression, and three non-pharmacologic therapies with good evidence:
- TMS
- ECT
- Lightbox
- Psychotherapy
Bipolar I vs II
The study did not separate bipolar I from II, but if they had we’d likely find greater efficacy in bipolar II with lumateperone, quetiapine, and pramipexole.
Practice Implications
- Meta-analyses only look at a small slice of the data, and network meta-analyses add assumptions that bias the results. I’ve tried to fill in those gaps here, but uncertainty is high.
- Lithium may not shine in acute trials, but rises to the top for long-term prevention.
- Lamotrigine is the best tolerated option. It is slow to build, but holds steady over time. Its efficacy is probably larger than this study conveys, as most of its trials only lasted 5 weeks and showed no benefit (it takes 4 weeks just to titrate the dose).
- Among the antipsychotics, lumateperone is the best tolerated, followed by lurasidone.
- After that, consider pramipexole and coenzyme Q10.
—Chris Aiken, MD
Director, Psych Partners
Editor in Chief, Carlat Psychiatry Report
What’s Your Take? Share in Comments
- Which of these have you had success with in bipolar depression? Which are hard to tolerate?
