The brain under treatment: what the imaging studies measured, and what they did not
Whether antipsychotics change the structure of the brain has been measured in five different ways, and the findings converge on the existence of an effect — not on which drugs, and with limitations the researchers themselves state, and which we pass on.
An animal experiment. Macaques given haloperidol or olanzapine for 17–27 months, at plasma levels comparable to patients', had 8–11% lower brain weight than the control group, across all major regions, with loss of both grey and white matter. Six animals per group; the authors present it as a possible confounder for human studies, not as proof.
A human cohort. 211 first-episode schizophrenia patients, 674 MRI scans over 7.2 years on average: greater cumulative exposure to antipsychotics was associated with smaller grey-matter volumes after statistical control for illness severity and substance use — a “subtle but measurable” effect, in the authors' words. And their own limitation: it was not randomised; those receiving more treatment already had smaller volumes at baseline, and severity may confound despite the adjustment.
Two meta-analyses. Across 30 longitudinal studies (1,046 patients, 780 controls) grey-matter loss correlated with cumulative exposure and not with duration or severity of illness; across 18 studies (1,155 patients) the same, with one difference: with first-generation drugs the loss increased with dose, with newer ones it did not — if anything, the reverse. Both state that correlation is not causation; the second discloses authors' fees from pharmaceutical companies.
And one randomised trial — a small one. It is the strongest in design, because it removes the problem of “who was more severely ill”, and the narrowest in scope: 72 people, a single drug, a single diagnosis. patients with psychotic depression who had already recovered on olanzapine and sertraline were randomised to continue olanzapine or to placebo for 36 weeks. Those who continued the drug showed cortical thinning that the others did not — equivalent, by the authors' estimate, to losing about 1.2% of the cortex in 36 weeks, when the normal annual change in adult life is 0.35%. The same paper records the other side: those who relapsed without the drug also showed thinning, and the authors' conclusion is that where psychosis is present, the risk of untreated illness outweighs it — while the balance shifts where alternatives exist or the drug is given off-label. They also note that the findings may not apply to other antipsychotics.
WHAT THIS FINDING IS, AND WHAT IT IS NOT
| IT IS | IT IS NOT |
|---|
| Converging evidence from animals, cohorts, meta-analyses and one randomised trial | Proof that the change is irreversible or clinically harmful |
| A reason for the dose to be the lowest effective one and the duration justified | A reason to stop — relapse without the drug is 64% versus 27% in the first year |
| Agreement on the existence of an effect — not on the drug: haloperidol in the comparisons, olanzapine in the only randomised trial | The same for every drug, dose and patient |
| An argument against off-label use — dementia, sleep, “agitation” | An argument against treatment in psychosis |
The practical significance is in the last row of the table. The same drugs measured here in psychosis are widely given to older people with dementia and to children off-label — see Part B and Part D. There the table reads the other way round: without the benefit that justifies the cost.
[64] Dorph-Petersen KA et al. Neuropsychopharmacology 2005;30(9):1649–1661 · [65] Ho BC et al. Arch Gen Psychiatry 2011;68(2):128–137 · [66] Fusar-Poli P et al. Neurosci Biobehav Rev 2013;37(8):1680–1691 · [67] Vita A et al. Biol Psychiatry 2015;78(6):403–412 · [68] Voineskos AN et al. JAMA Psychiatry 2020;77(7):674–683 · [69] Lieberman JA et al. Arch Gen Psychiatry 2005;62(4):361–370 · relapse: [33] Leucht S et al. Lancet 2012