In 1934 a Hungarian psychiatrist injected camphor into a catatonic man to give him a seizure, because he believed that epilepsy and schizophrenia could not live in the same brain. The premise turned out to be wrong. The treatment it produced is still in use. And twenty years later a Swiss neurologist described the mirror image of Meduna's idea: patients with epilepsy who became psychotic just as their seizures stopped and their EEG became normal. This article follows that idea from the asylum to the modern epilepsy clinic, and asks what is left of it.
Meduna's wager
Ladislas von Meduna worked as a neuropathologist before he turned to patients. Looking at brains under the microscope, he was struck by a contrast: the glia proliferated in the brains of people with epilepsy and appeared sparse in those of people with schizophrenia. Clinical reports of the time added the impression that the two conditions rarely occurred together, and that the occasional patient with both seemed to improve in one when the other flared. From these observations he built a hypothesis of biological antagonism: if epilepsy and schizophrenia excluded each other, a seizure might push schizophrenia out1.
On 23 January 1934, at the Lipótmező asylum in Budapest, he gave an intramuscular injection of camphor in oil to a man who had been catatonic for four years. The seizure came after a long and frightening delay; after a series of them, the patient began to speak, eat and look after himself. Camphor was slow and unpredictable, so Meduna moved to pentylenetetrazol, sold as Cardiazol in Europe and Metrazol in the United States, which produced a seizure within seconds of an intravenous injection. He published his first series in 19351,2.
The patients' experience was brutal. Pentylenetetrazol produced intense dread in the seconds before consciousness was lost, and the convulsions caused fractures. In April 1938, in Rome, Ugo Cerletti and Lucio Bini induced the first seizure with electricity instead, after years of experiments in animals3. Electroconvulsive therapy was faster, more controllable and, once anaesthesia and muscle relaxants were added, far safer. It replaced chemical convulsion within a decade.
The premise was wrong
The antagonism hypothesis did not survive contact with data. In 1963 Eliot Slater and A. W. Beard described 69 patients at the Maudsley and National Hospitals who had developed a schizophrenia-like psychosis after years of epilepsy, most of them with temporal lobe seizures. The psychosis appeared, on average, about fourteen years after the epilepsy began, and the authors argued that the association was far too frequent to be chance4.
Modern epidemiology settled the question in the same direction. In a Danish population cohort of 2.27 million people, a history of epilepsy carried about two and a half times the risk of schizophrenia (relative risk 2.48) and nearly three times the risk of schizophrenia-like psychosis (2.93), and the association ran in both directions5. A 2014 meta-analysis put the pooled prevalence of psychosis in people with epilepsy at 5.6%, rising to 7% in temporal lobe epilepsy, with an odds ratio of about 7.8 compared with the general population6.
Epilepsy and psychosis are not enemies. They are neighbours that share a border, and the border runs through the temporal and limbic networks.
Meduna was therefore right for the wrong reason. Induced seizures did help some psychotic patients, particularly those with catatonia, but not because epilepsy and schizophrenia are opposites. The idea of antagonism might have disappeared entirely had it not been for a strange observation in the epilepsy clinic.
Landolt's paradox
Heinrich Landolt, working in Zurich in the early 1950s, recorded serial EEGs in patients with epilepsy who developed episodes of psychosis. In some of them he found something unexpected: during the psychotic episode the EEG became more normal, sometimes entirely normal, and the epileptiform activity returned when the psychosis cleared. He called it forcierte Normalisierung, forced normalization7. He noted that the phenomenon seemed to follow treatment with succinimides, the drugs that had just been introduced for absence seizures.
In 1965 Hubertus Tellenbach described the clinical side of the same coin without needing an EEG: some patients alternated between periods of seizures and periods of psychosis, as if one state replaced the other. He called it alternative psychosis8. Peter Wolf later preferred paradoxical normalization, a term that describes the finding without implying that anything is being forced9. The terms overlap but are not identical: forced normalization is defined by the EEG, alternative psychosis by the clinical course.
What forced normalization looks like now
Forced normalization is uncommon, and most of what we know comes from case reports and small series rather than cohorts10,11. The typical story is a patient with long-standing, frequent and often drug-resistant focal epilepsy, frequently of temporal or limbic origin, who becomes seizure-free or nearly so after a new antiseizure medication or a dose increase, and then, over days to weeks, develops a behavioural disturbance. Psychosis is the classic picture, but depression, mania, anxiety, dissociative symptoms and functional (non-epileptic) seizures have all been described11,12.
The drugs most often implicated are ethosuximide, the original culprit in Landolt's patients, vigabatrin, levetiracetam and topiramate; barbiturates and phenytoin have also been reported11,12. The same pattern has been described after epilepsy surgery that stops seizures. A recent change in treatment is so common in the published cases that it is part of the diagnostic criteria.
Krishnamoorthy and Trimble proposed criteria in 1999 that are still the standard reference13. Their primary criteria are:
- An established diagnosis of epilepsy, based on history, EEG and imaging.
- A behavioural disturbance of acute or subacute onset: psychosis, a significant change in mood, anxiety or another neuropsychiatric syndrome.
- Either (A) a reduction of more than 50% in the epileptiform activity on a waking EEG compared with a recording during a previous behavioural state, or (B) a report from the patient or family that seizures have stopped completely for at least a week, compared with the four weeks before.
Criteria 1, 2 and 3A are enough. If only the clinical criterion 3B is met, at least one supportive feature is needed: a change in antiseizure treatment within the previous month, the return of seizures as the behavioural disturbance resolves, or similar episodes in the past13,14.
Not every psychosis in epilepsy is forced normalization
The commonest error is not missing forced normalization but diagnosing it when the patient has something else. Psychosis in epilepsy is usually classified by its relation in time to the seizures, and the distinction changes treatment.
| Type | Relation to seizures | What it usually means |
|---|---|---|
| Postictal psychosis | Follows a cluster of seizures or a single severe one, typically after a lucid interval of hours to a few days, and resolves within days to weeks15 | The commonest acute psychosis in epilepsy. Treat the seizures more, not less; short-term antipsychotic or benzodiazepine. |
| Ictal psychosis | Occurs during seizure activity, often non-convulsive status epilepticus | An EEG emergency, not a psychiatric one. |
| Interictal psychosis | Chronic, between seizures, independent of their frequency; the picture Slater and Beard described4 | Long-term antipsychotic treatment alongside the epilepsy care. |
| Forced normalization / alternative psychosis | Appears as seizures stop or the EEG improves, often after a treatment change | Review the most recent change; seizures and psychosis may need to be balanced. |
| Drug-induced psychosis | Follows a new drug or dose without a clear change in seizure control | Overlaps with forced normalization; the same drugs are involved. |
The boundary between forced normalization and a direct psychiatric side effect of the drug is genuinely blurred. Levetiracetam and topiramate cause irritability, depression and occasionally psychosis in people whose seizures have not changed at all. When the two coincide, the published cases rarely allow the effects to be separated, and the patient is helped by the same decision either way.
Why it might happen
No mechanism has been established12. The proposals fall into three groups.
Networks. The temporal and limbic structures that generate focal seizures also take part in perception, emotion and the sense of what is real. A surface EEG samples the cortex and can be normal while deep structures remain abnormal. Wolf suggested that in forced normalization the epileptic process is not abolished but continues subcortically, spreading along different pathways and showing itself as behaviour instead of seizures9.
Neurotransmitters. The dopamine hypothesis is the most developed. Seizures and electroconvulsive therapy both change dopaminergic transmission, and it has been proposed that the suppression of seizures removes an influence that was keeping dopaminergic activity in check12. Changes in the balance between glutamate and GABA have also been invoked14. These are reasonable hypotheses with experimental support, not demonstrated mechanisms in patients.
Inhibition. One old idea is that a seizure is followed by a period of intense inhibition, and that frequent seizures keep a network in a chronically inhibited state. Removing the seizures removes that inhibition. The hypothesis has never been tested directly in humans, but it fits the second part of this story: the anticonvulsant effect of electroconvulsive therapy.
The paradox inside the treatment
Electroconvulsive therapy makes the antagonism idea look even stranger. A course of ECT raises the seizure threshold: later treatments need a larger stimulus to produce a seizure, and the seizures become shorter. Harold Sackeim argued that this anticonvulsant effect is part of how ECT works, and that the seizure matters for what the brain does to stop it16. ECT has been used, in small series, to treat refractory status epilepticus. The procedure that Meduna invented to cause seizures is, physiologically, an anticonvulsant.
What ECT does not do is confirm that seizures cure schizophrenia. Its main modern indications are severe depression and catatonia. In schizophrenia its best evidence is narrow: in a randomised trial of 39 patients whose illness had not responded to clozapine, half of those given ECT in addition to clozapine responded, compared with none of those who continued clozapine alone; 47% of non-responders improved when they crossed over to ECT17. That is an augmentation strategy for a selected group, under anaesthesia, with controlled seizures, and nothing like a spontaneous epileptic seizure.
What this means at the bedside
Practical points
- Psychosis is several times more common in people with epilepsy than in the general population. Ask about it, and ask the family.
- When a patient with epilepsy becomes psychotic, first place the psychosis in time: after a cluster (postictal), during seizures (ictal, needs an EEG), chronic (interictal) or as the seizures stop (forced normalization).
- Look at what changed in the last month: a new antiseizure medication, a dose increase, a withdrawal, surgery.
- An EEG during the episode is the only way to show normalization, and the only way to exclude non-convulsive status epilepticus.
- In suspected forced normalization the usual approach in published cases is to reduce or replace the most recently started drug, accepting some seizures if necessary, and to treat the psychosis with an antipsychotic. Neither choice has trial evidence; the decision is a balance agreed with the patient and family.
- Do not stop antiseizure treatment abruptly. Withdrawal seizures and status epilepticus are more dangerous than the psychosis.
- Seizure freedom is not automatically success. Ask how the patient is, not only how many seizures they had.
What is left of the antagonism
As a theory of two diseases, biological antagonism is dead. Epilepsy and psychosis co-occur far more often than chance, and the same patient can have both, at the same time or in turn. As a description of a small group of patients, something of it survives: in a minority of people with difficult focal epilepsy, the brain seems to move between two abnormal states, and suppressing one can uncover the other.
The question this leaves is a real research question, not a settled fact. Can the same change that silences epileptic activity in one network destabilise another network that holds a person's grip on reality? If that is true, it says something important about both conditions: that a seizure is not only a disorder but also, in some brains, a kind of regulation, and that controlling it has a cost that a seizure diary does not record.
Meduna's camphor was based on a mistake. The patients he treated, and the ones Landolt watched become ill as their EEGs became normal, still teach the same lesson: epilepsy is a disorder of networks that also carry thought and mood, and its treatment should be judged by the whole person.
References
- Fink M. Meduna and the origins of convulsive therapy. Am J Psychiatry. 1984;141(9):1034–41. PubMed 6147103
- Meduna L. Versuche über die biologische Beeinflussung des Ablaufes der Schizophrenie. I. Campher- und Cardiazolkrämpfe. Z Gesamte Neurol Psychiatr. 1935;152:235–62.
- Cerletti U. Old and new information about electroshock. Am J Psychiatry. 1950;107(2):87–94.
- Slater E, Beard AW. The schizophrenia-like psychoses of epilepsy. Br J Psychiatry. 1963;109:95–150.
- Qin P, Xu H, Laursen TM, Vestergaard M, Mortensen PB. Risk for schizophrenia and schizophrenia-like psychosis among patients with epilepsy: population based cohort study. BMJ. 2005;331(7507):23. doi:10.1136/bmj.38488.462037.8F
- Clancy MJ, Clarke MC, Connor DJ, Cannon M, Cotter DR. The prevalence of psychosis in epilepsy: a systematic review and meta-analysis. BMC Psychiatry. 2014;14:75. doi:10.1186/1471-244X-14-75
- Landolt H. Serial electroencephalographic investigations during psychotic episodes in epileptic patients and during schizophrenic attacks. In: Lorentz de Haas AM, ed. Lectures on Epilepsy. Amsterdam: Elsevier; 1958:91–133.
- Tellenbach H. Epilepsie als Anfallsleiden und als Psychose: über alternative Psychosen paranoider Prägung bei „forcierter Normalisierung“ (Landolt) des Elektroencephalogramms Epileptischer. Nervenarzt. 1965;36:190–202.
- Wolf P. Acute behavioral symptomatology at disappearance of epileptiform EEG abnormality: paradoxical or “forced” normalization. Adv Neurol. 1991;55:127–42.
- Calle-López Y, Ladino LD, Benjumea-Cuartas V, et al. Forced normalization: a systematic review. Epilepsia. 2019;60(8):1610–8.
- Krishnamoorthy ES, Trimble MR, Sander JWAS, Kanner AM. Forced normalization at the interface between epilepsy and psychiatry. Epilepsy Behav. 2002;3(4):303–8.
- Kawakami Y, Itoh Y. Forced normalization: antagonism between epilepsy and psychosis. Pediatr Neurol. 2017;70:16–9. doi:10.1016/j.pediatrneurol.2017.02.007
- Krishnamoorthy ES, Trimble MR. Forced normalization: clinical and therapeutic relevance. Epilepsia. 1999;40 Suppl 10:S57–64.
- Anzellotti F, Franciotti R, Zhuzhuni H, D'Amico A, Thomas A, Onofrj M. Nonepileptic seizures under levetiracetam therapy: a case report of forced normalization process. Neuropsychiatr Dis Treat. 2014;10:959–64. doi:10.2147/NDT.S60089
- Logsdail SJ, Toone BK. Post-ictal psychoses: a clinical and phenomenological description. Br J Psychiatry. 1988;152:246–52.
- Sackeim HA. The anticonvulsant hypothesis of the mechanisms of action of ECT: current status. J ECT. 1999;15(1):5–26.
- Petrides G, Malur C, Braga RJ, et al. Electroconvulsive therapy augmentation in clozapine-resistant schizophrenia: a prospective, randomized study. Am J Psychiatry. 2015;172(1):52–8. doi:10.1176/appi.ajp.2014.13060787