When the first psychotic episode is really an inflamed brain: screening 143 patients for autoimmune encephalitis
- Among 143 consecutively admitted first-episode psychosis patients screened at a tertiary psychiatric hospital, 5 (3.5 percent) were confirmed to have autoimmune encephalitis on Graus criteria.
- Detected antibodies were heterogeneous (anti-AMPH, anti-AMPA2, anti-GABAb, anti-Yo, plus one cerebrospinal-fluid antineuronal case), arguing against any single test as a sufficient gate.
- One of the five patients showed no clinical red flags at all, and in another the antibody appeared only in cerebrospinal fluid and not in serum, so clinical screening alone would have missed both.
- Of the five, one died in the acute phase and three recovered well enough to return to work, underscoring that the cost of a missed diagnosis is measured in lives and function.
Every clinician who works with acute psychosis carries a quiet anxiety: somewhere in the stream of first-episode admissions sits a patient whose hallucinations and disorganisation are not a primary psychiatric illness at all, but an autoimmune attack on the brain that antipsychotics will not touch and immunotherapy might reverse. The team led by Vera Fominykh at a Moscow tertiary psychiatric hospital decided to stop guessing and measure. Over the study window they screened 784 consecutive referrals, applied inclusion and exclusion criteria, and prospectively followed 143 first-episode psychosis (FEP) patients through acute-phase serum and cerebrospinal-fluid testing and a 9 to 12 month follow-up.
The headline number is 3.5 percent confirmed autoimmune encephalitis (AE) by Graus criteria – five patients. That figure sits comfortably inside the range other FEP cohorts have reported, which matters: it tells us this Russian sample behaves like its international peers and that the finding is not an artefact of an unusual catchment. What makes the paper clinically useful is not the prevalence but the texture of how those five were caught.
The antibody profile was deliberately scattered: anti-AMPH, anti-AMPA2, anti-GABAb, anti-Yo, and a case positive only on cerebrospinal-fluid antineuronal testing. There was no single dominant target, which quietly undermines the hope that one cheap serum panel could serve as a triage gate. The authors built their workflow as two layers – a clinical red flags assessment first, then a deeper laboratory and imaging set (cerebrospinal-fluid antibodies, brain MRI, oncological screening) for anyone flagged or laboratory-positive.
The two instructive failures sit precisely at the seams of that workflow. One patient had no red flags whatsoever, and would have been waved through any purely clinical screen. Another carried the antibody only in cerebrospinal fluid, invisible to serum. Together they make the paper's central argument concrete: red flags plus serum caught four of five; the fifth required lumbar puncture. A protocol that stopped at clinical judgement, or at a blood draw, would have missed roughly a fifth of the true cases in this cohort.
That argument has weight because of the stakes the follow-up revealed. One of the five died in the acute stage; three returned to work after immunosuppressive treatment. This is the rare psychiatric finding where a laboratory step changes not just a diagnostic label but a trajectory – from a lifelong antipsychotic regimen that does nothing to a time-sensitive immunotherapy that restores function.
Why a 3.5 percent finding should reshape an admission protocol
A prevalence this low is exactly the kind of number that gets argued away at the bedside – too rare to justify routine lumbar puncture, the reasoning goes. This study reframes that calculus. The relevant question is not how often AE appears but how catastrophic and how reversible it is when missed, and on both counts the answer here is stark.
The limits of the red-flag heuristic
Psychiatry leans heavily on clinical pattern recognition, and red-flag panels encode that instinct into a checklist. Fominykh and colleagues show the checklist is necessary but not sufficient: it is the entry point to investigation, not a substitute for fluid analysis in the subset it selects, and it will not select everyone who needs testing.
In a cohort of 143 first-episode psychosis patients, clinical red flags plus serum testing caught four of five autoimmune encephalitis cases; the fifth was visible only in cerebrospinal fluid.
This was a single-centre cohort at one East-European tertiary hospital, so the 3.5 percent figure may not transfer to community or general-hospital settings with different referral patterns. The confirmed-case count was small (five patients), which limits any inference about which red flags matter most. The non-interventional design describes diagnostic yield rather than testing whether the proposed algorithm improves outcomes against an alternative.