Antipsychotic non-responders have a different frontal metabolite profile, by a small margin

Design

Individual participant data mega-analysis of magnetic resonance spectroscopy studies with linear mixed models, plus conventional meta-analysis

Population

1,189 participants from 18 studies: 476 antipsychotic non-responders, 427 responders, 286 healthy controls

Primary outcome

Group differences in glutamate, glutamate plus glutamine, choline, myo-inositol, N-acetylaspartate, GABA and glutathione across four brain regions

Effect

Non-responders vs responders: medial frontal glutamate Glass delta 0.21 (p=0.02), glutamate plus glutamine 0.29 (p=0.002), choline 0.22 (p=0.03), myo-inositol 0.35 (p=0.001); 0.41 prospectively in first-episode psychosis and 0.64 for myo-inositol in treatment resistance

Predicting who will not respond to antipsychotics would change when clozapine is offered, which is currently far later than it should be. This mega-analysis pooled individual participant data from 18 magnetic resonance spectroscopy studies - 1,189 participants: 476 non-responders, 427 responders and 286 healthy controls - and added five more studies to the conventional meta-analysis.

Compared with responders, non-responders had raised medial frontal glutamate (Glass delta 0.21, p=0.02), glutamate plus glutamine (0.29, p=0.002), choline (0.22, p=0.03) and myo-inositol (0.35, p=0.001), with similar elevations against healthy controls. Two subgroup findings strengthen the case: raised glutamate plus glutamine was seen prospectively in first-episode psychosis (0.41, p=0.002), so it is not simply a consequence of chronic illness or of failed treatment, and myo-inositol elevations were largest in patients meeting treatment-resistance criteria (0.64, p=0.001).

The effect sizes are the limitation. A Glass delta of 0.2 to 0.35 means the distributions overlap heavily - most non-responders have values within the responder range - so nothing here separates two individuals in a clinic. Spectroscopy is also not routinely available, and measurement varies between scanners and sequences.

So this is a mechanistic result with a clinical direction rather than a test. Glutamatergic and inflammatory or glial markers, rather than dopamine, distinguish the patients who do not respond, which supports developing treatments aimed elsewhere than the D2 receptor. The clinical instruction remains unchanged and is worth repeating: recognise non-response early and move to clozapine, because no biomarker is going to do that for you.

• Define non-response prospectively: two adequate trials at adequate dose and duration, then clozapine
• Do not delay clozapine waiting for a biomarker - none of these separates individuals
• Myo-inositol and choline are glial markers, pointing away from a purely dopaminergic model
• Spectroscopy findings vary by scanner and sequence; this is not a portable measurement
• The prospective first-episode signal argues the difference precedes treatment failure rather than following it

The statistics, in plain English

Glass delta is a standardised difference using the control group's standard deviation; 0.21 to 0.35 is small by convention, meaning the two groups' distributions overlap by roughly 85-90%. Statistical significance in 1,189 people tells you the average difference is real, not that it is useful for classifying anyone. Several metabolites in four brain regions were tested, so some findings would be expected to reach p<0.05 by chance; the consistency of the glutamatergic signal across the prospective subgroup is what makes it more than that.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.