Design: Individual participant data mega-analysis of 18 studies plus meta-analysis of 23 studies
Population: 1,189 participants (476 non-responders, 427 responders, 286 controls) in the mega-analysis; 1,844 in the meta-analysis
Primary outcome: Group differences in brain neurometabolites by antipsychotic treatment response
Effect: Non-responders: medial frontal glutamate Glass Δ=0.21 (P=.02), Glx Δ=0.29 (P=.002), choline Δ=0.22 (P=.03), myo-inositol Δ=0.35 (P=.001)

This mega-analysis pooled individual participant data from 18 magnetic resonance spectroscopy studies — 1,189 people, of whom 476 were antipsychotic non-responders, 427 responders and 286 healthy controls — with a further meta-analysis of 23 studies and 1,844 participants. The question was whether brain neurometabolite concentrations differ between those who respond to antipsychotics and those who do not.

Non-responders had raised medial frontal glutamate (Glass Δ = 0.21, P=.02), glutamate plus glutamine (Δ = 0.29, P=.002), choline (Δ = 0.22, P=.03) and myo-inositol (Δ = 0.35, P=.001) compared with responders, with similar elevations against controls. The glutamate plus glutamine difference also appeared prospectively in first-episode psychosis (Δ = 0.41, P=.002), and myo-inositol elevation was largest in those meeting treatment-resistance criteria (Δ = 0.64, P=.001).

This is not a test to order. Effect sizes of 0.2 to 0.6 describe group differences with heavy overlap, and no clinical spectroscopy protocol exists to act on them. Its value is as a target signal: it supports continued work on glutamate-acting and inflammatory-pathway treatments in a population where the dopamine hypothesis has never explained non-response.

Do not order spectroscopy on the strength of this — these are group differences, not a diagnostic threshold
The prospective first-episode finding is the strongest part; it was not purely retrospective
Myo-inositol is an inflammation-associated marker, which is where the treatment implication sits
Treatment resistance still needs defining clinically — adequate dose, adequate duration, confirmed adherence
Worth knowing when a patient asks whether their non-response has a biological basis

Why it matters

It gives a biological handle on antipsychotic non-response that is not dopaminergic.

The statistics, in plain English

Glass's delta of 0.21 to 0.64 means the two groups' distributions overlap substantially — most non-responders have values found in many responders. A P value of .02 across 1,189 people says the average difference is unlikely to be chance; it says nothing about whether you could classify an individual. The largest effect, myo-inositol at 0.64 in treatment resistance, is still well short of what a clinical marker needs.