The randomized trials were negative, but a meta-analysis finds a signal
STUDY: Bak M et al, Journal of Clinical Psychopharmacology 2026
STUDY TYPE: Systematic review and meta-analysis
FUNDING: Independent
Background
Olanzapine orally disintegrating tablet (ODT) is valued in emergency settings for its faster onset of action, peaking about an hour earlier than standard tablets (3.5 vs 4.4 hours).
When ODT was first released as Zyprexa Zydis, we saw reports of weight loss after switching to it, but the trials were not randomized. Later, four randomized trials found no difference, and the idea was put to rest.
Now, a meta-analysis detects a signal, so how do we make sense of these conflicting results?
The Study
- Ten studies (five randomized trials, five cohort studies) comparing orally disintegrating olanzapine with standard tablets.
- Follow-up ranged from 4 weeks to a year.
Results
ODT caused less weight gain in antipsychotic-naïve (2 lbs / 0.9 kg difference, p = 0.008), but not in patients already on antipsychotics (they had a 1.3 lb / 0.61 kg difference which was not statistically significant).
The positive results were not just due to the uncontrolled trials. They held up when the randomized trials were analyzed separately, although only two randomized trials enrolled antipsychotic-naive patients.
The analyses tells us about the difficulties of interpreting data. Here are two opposing ways to interpret them.
Argument for a Real Difference: Assay Sensitivity
The best way to show a difference is to test the treatment in patients who are likely to respond (assay sensitivity).
Antipsychotic-naive patients are more sensitive to weight gain, so it makes sense that trials focused on this population would detect a difference.
Pooling data may also explain the positive signal. It is possible that small trials will miss a real signal that only becomes detectable when they are pooled together. Randomized trials are 4-times as likely to produce false-negatives than false positives [1], although in reality we’re less likely to see these negative trials because of publication bias.
[1] A trial’s nominal false-positive rate is α = 5% and its nominal false-negative rate is β = 10–20%, so a “negative” trial is roughly 4-fold more likely to be wrong than a “positive” one is by design.
Argument for No Difference: Publication Bias
A skeptic’s view is that this is due to the tendency to publish positive results over negative ones.
The easier the trial is to do, the more this principle applies. Imagine thousands of clinicians who switch patients to Zyprexa Zydis. Many check weight on them, and the lucky few who see a difference on them will publish. If they suggest to the patients that weight loss might follow, a placebo-driven shift in healthy behaviors will further the bias.
That may also explain the randomized trials. The only one that was positive on its own was easy to do: A secondary analysis of a randomized trial. Most researchers scan their results for novel signals after they’ve assessed the primary outcome. But a secondary analysis of null findings is not likely to reach publication. In this case, the secondary analysis was small (38 subjects), barely positive (p value 0.04) [2], and happened to involve antipsychotic-naive patients, tilting the balance for that group.
[2] Small trials with marginal p values are a danger zone. In anesthesia journals, over 50% of them had false-positive results.
A Biological Rationale?
When statistics are uncertain, a clear biological rationale may tip the balance. In statistics, this is the pretest probability: How likely was this idea to begin with?
The theory is that olanzapine ODT mostly skips first-pass metabolism in the liver, producing fewer metabolites. However, this doesn’t hold up, as there is no evidence that its metabolites cause weight gain. The most-studied metabolite, N-desmethyl-olanzapine, protects against weight gain and diabetes in rodents.
Practice Implications
- As a hopeful psychiatrist, I want this signal to be true.
- But with no sound biological rationale and too many sound statistical reasons to reject it, I’ll view it as a hypothesis in need of further testing.
—Chris Aiken, MD
Director, Psych Partners
Editor in Chief, Carlat Psychiatry Report
Special thanks to the primary author, Maarten Bak, MD, PhD, for corresponding on these dilemmas.
References
Four negative randomized trials:
- Kusumi I, Honda M, Uemura K, et al. Effect of olanzapine orally disintegrating tablet versus oral standard tablet on body weight in patients with schizophrenia: a randomized open-label trial. Prog Neuropsychopharmacol Biol Psychiatry. 2012;36(2):313-317.
- Bobo WV, Epstein RA Jr, Shelton RC. Effects of orally disintegrating vs regular olanzapine tablets on body weight, eating behavior, glycemic and lipid indices, and gastrointestinal hormones: a randomized, open comparison in outpatients with bipolar depression. Ann Clin Psychiatry. 2011;23(3):193-201.
- Karagianis J, Grossman L, Landry J, et al. A randomized controlled trial of the effect of sublingual orally disintegrating olanzapine versus oral olanzapine on body mass index: the PLATYPUS Study. Schizophr Res. 2009;113(1):41-48.
- Bitter I, Treuer T, Dilbaz N, et al. Patients’ preference for olanzapine orodispersible tablet compared with conventional oral tablet in a multinational, randomized, crossover study. World J Biol Psychiatry. 2010;11(7):894-903. doi:10.3109/15622975.2010.505663
Secondary analysis of a randomized trial (marginally positive):
- Arranz B, San L, Dueñas RM, et al. Lower weight gain with the orally disintegrating olanzapine than with standard tablets in first-episode never treated psychotic patients. Hum Psychopharmacol. 2007;22(1):11-15.
One comment
Christopher Fichtner, MD
August 15, 2026 at 9:24 pm
When the first studies were published suggesting that the ODT might be less likely to cause weight gain, I changed my practice to use only the ODT almost all the time, with rare exceptions. My reason for this is the lack of any conceivable down side in doing so, as well as the importance of the potential benefit if the phenomenon is real. Negative results in controlled studies don’t rule out the possibility that select individuals within those studies may have benefited significantly from the ODT with regard to the variable (weight) of interest, while other individuals in those studies did not; that is the nature of all of our aggregate sample trials analyzed statistically. That a meta-analysis finds a signal is helpful to know. I find no compelling clinical reason to ever select the standard oral tablet over the ODT, except perhaps in the rare instance where weight gain is a goal. As I have practiced consistently using the ODT most of the time, and in some cases switched patients coming to me on the standard oral tablet over to the ODT, my clinical observations seem to confirm that this is a good way to go, though it does not reliably protect everyone from weight gain.