More than 70 years after the first use of chlorpromazine in psychosis, there remain widely different views on the prescribing of antipsychotic drugs in schizophrenia and schizoaffective disorder. While most practitioners recognise the need for long-term treatment, there is an incomplete consensus on the optimal duration of treatment, although most people with schizophrenia-like psychoses are prescribed antipsychotics for the whole of their lives, after diagnosis.

There are opposing views on if, or when, to stop, and on how to stop antipsychotic treatment. A large majority recognise the need for life-long treatment, while a smaller proportion advocate for at least attempting to withdraw treatment. When antipsychotics are stopped by patients themselves, relapse is by far the most likely outcome1 and almost everyone with a diagnosis of schizophrenia will experience a second relapse within 10 years of diagnosis2. Continuous delivery of an adequate antipsychotic dose reduces the risk of relapse almost to zero, even in people with several prior episodes3. These observations offer apparently incontestable support for the benefits of long-term or lifelong treatment with antipsychotics in schizophreniform conditions.

In this edition, Moncrieff and Horowitz examine in detail the issue of discontinuing antipsychotics. Beginning by casting doubt on the effectiveness of antipsychotics (a subject they, perhaps wisely, leave behind), they go on to question the idea that the removal of the antipsychotic’s therapeutic effect is the sole or main cause of relapse. They propose instead that relapse might be primed by the dopamine-blocking properties of the antipsychotic, such that relapse occurs because sudden cessation exposes a hypersensitive state, which in turn provokes the return of symptoms. Alongside this process, they suggest, runs a withdrawal syndrome that results from the precipitous loss of activity at other receptors. The consequences of dopamine hypersensitivity and the non-dopamine withdrawal symptoms are alleged to combine to produce something that is recognised and reported as relapse. Support for this view comes from the handful of cases of psychosis that occur in people stopping dopamine antagonists prescribed for self-limiting non-psychiatric conditions (although it is only a handful4) and from the observation that slow (as opposed to abrupt) discontinuation of antipsychotics reduces the risk of relapse. This second observation is open to interpretation, since although the rate of relapse is palpably changed by slow discontinuation, the extent of relapse (the proportion of people who ultimately relapse) is at least partly determined by the length of follow-up, with longer studies tending to show convergence in the extent of relapse5.

When discussing the mechanism of action, Moncrieff and Horowitz repeatedly drift into a discussion of the possibility that antipsychotics do not correct, even temporarily, any specific abnormality or pathology related to schizophrenia, but merely ‘suppress symptoms by altering normal brain functioning’. While this has some relevance to the question of whether or not to continue antipsychotics, it adds a layer of debate that is perhaps superfluous.

Moncrief and Horowitz go on to liken antipsychotics to lithium and its capacity to accelerate time to relapse and frequency of relapse when abruptly withdrawn6. This is an interesting suggestion: that antipsychotics, like lithium, are highly efficacious when taken, but if abruptly stopped, cause a worsening of the condition being treated. They submit that withdrawal symptoms provoked by abruptly stopping an antipsychotic bring forward an episode of the underlying condition, perhaps via intensified emotional reactivity, or insomnia. This phenomenon is apparently evident in studies comparing the rate of relapse following discontinuation of oral and long-acting forms of the same drug—relapse is certainly brought forward in those abruptly stopping oral treatment. However, an obvious alternative explanation is that relapse begins to occur when the concentration of the drug at the site of action falls below a certain threshold.

The benefits of discontinuing antipsychotic treatment include the lessening of the severity and frequency of adverse effects, particularly metabolic effects and weight gain, but also neurological effects such as akathisia and tardive dyskinesia. While these benefits are self-evident, it should be noted that antipsychotic use is associated with better mortality than non-use—people live substantially longer on antipsychotics7,8.

Buried deep in the highly technical and complex information that is carefully and objectively presented by Moncrieff and Horowitz is one indisputable fact—that some people with a diagnosis of schizophrenia can stop antipsychotic treatment for good and not relapse. We should not be surprised by this. In the pre-chlorpromazine era, many people with schizophrenia-like conditions were discharged from hospital, never to return9, and perhaps some of these people recovered. Schizophrenia may not be, for some people at least, a lifelong condition. We should therefore take more seriously the possibility that antipsychotics can be successfully stopped. In the absence of any means of identifying those who can stop and bearing in mind the probability that relapse worsens outcomes in the longer term (opinions differ), it could be argued that all people with a first episode of psychosis should be considered for careful, hyperbolic withdrawal of antipsychotic treatment. We should not forget, however, that antipsychotics are often stopped abruptly by patients themselves (or at least not taken consistently)10, and that this portends relapse in the large majority of people and a likely worsening of the condition itself. Stopping antipsychotics, especially abrupt stopping, is thus most likely to have a negative outcome for the patient. This observation is sufficient evidence to support the much wider use of long-acting injectable formulations of antipsychotics; formulations that can be withdrawn, if withdrawal is indeed to be considered, in a manner least likely to provoke relapse.