Schizophrenia follows a relapsing–remitting pattern in a significant proportion of people. ^{Reference Jørgensen, Bøg and Kabra1} Periods of relapse of psychotic illness are strongly linked to severe negative impacts on day-to-day functioning, social relationships and general health. ^{Reference Tessier, Roger, Gregoire, Desnavailles and Misdrahi2} The mechanisms underlying relapse of psychotic illness remain unclear, but established risk factors include periods of external stress, sleep disturbance, drug abuse and discontinuation of maintenance antipsychotic treatment. ^{Reference Rivelli, Fitzpatrick and Nelson3} There is considerable concern that the more frequent, severe and longer-lasting the periods of psychotic relapse, the less likely that the person will return to their baseline functioning. ^{Reference Emsley, Chiliza, Asmal and Harvey4} However, despite considerable evidence that maintenance antipsychotic treatment reduces relapses, a significant proportion of people with a psychotic illness do not adhere to treatment with antipsychotic medication. The article by Moncrieff et al reports no significant differences in social functioning, symptom scores or quality of life outcomes after 24 months in patients with schizophrenia who had a relapse of their illness versus those who did not. However, the authors did find a detrimental impact of relapse on participants’ employment status. Their data comprise secondary analysis of a randomised controlled trial assessing the impact of antipsychotic medication dose reduction versus maintenance treatment on social functioning. ^{Reference Moncrieff5} This original study found that antipsychotic medication dose reduction was associated with an increased number of relapses of greater severity and several other adverse outcomes, but findings did not support the a priori hypothesis of significant improvement in social functioning or neurocognition. All the participants who relapsed in this primary antipsychotic reduction study were included in the secondary analysis of the publication. Here, we place these findings in a wider context, examining issues with antipsychotic discontinuation: the impact of relapse on brain physiology, the generalisability of the participant sample recruited into the primary study and, finally, the potential impact on clinical practice.

There is no doubt that the course of schizophrenia is heterogenous, with approximately 15–20% of patients non-responsive to treatment from onset, ^{Reference Siskind, Orr, Sinha, Yu, Brijball and Warren6,Reference Lally, Ajnakina, Di Forti, Trotta, Demjaha and Kolliakou7} 20% experiencing only one episode within 5 years ^{Reference Siskind, Orr, Sinha, Yu, Brijball and Warren6,Reference Lally, Ajnakina, Di Forti, Trotta, Demjaha and Kolliakou7} and with multiple individual courses in between. Thus, relapse remains an important reality for many people with schizophrenia. One recent meta-analysis of antipsychotics for relapse prevention reported relapse rates of 61% on placebo, compared with 24% on antipsychotics, within 9–12 months, ^{Reference Ceraso, Lin, Schneider-Thoma, Siafis, Heres and Kissling8} and rehospitalisation rates were 18 and 7%, respectively. At present, because it is not possible to predict those people who will suffer a relapse on discontinuing antipsychotic medication, some will be receiving treatment without clear benefit. This calls for improved methods of identifying patients likely to be at higher risk of relapse and those with worse outcomes. Perhaps most concerning for patients and clinicians is that periods of relapse may not only have direct impacts such as hospitalisation and/or loss of a work or relationships, but also a more general destabilising effect, making subsequent antipsychotic treatment less effective. ^{Reference Takeuchi, Siu, Remington, Fervaha, Zipursky and Foussias9} The most prominent theories regarding the mechanism underlying relapse in psychosis are largely variants of ‘receptor supersensitivity’, whereby continuous dopamine blockade by antipsychotic medication results in both increased numbers and sensitivity of existing dopamine receptors. Therefore, abrupt withdrawal of antipsychotic medication can lead to a rebound psychosis. It seems sensible that, if one is seeking to minimise risk of relapse, antipsychotics should be tapered gradually over a period of months or even years, using progressively smaller dose reductions.

The original study challenges the idea that relapse leads to long-term deterioration in schizophrenia or recurrent psychosis, suggesting that social and clinical functioning can return to pre-relapse levels within 24 months. Obviously, the generalisability of the results is predicated on the representativeness of the sample of patients participating in the study. There are several pertinent observations in this regard: first, the study involved older participants, predominantly white and with longer-term conditions, and thus may not be generalisable to first-episode psychosis. One possibility is that, while relapse may not significantly impact long-term outcomes in patients with chronic conditions, prevention of relapse early in the course of illness may be critical. Second, the study in secondary care services may not apply to people with a longer duration of stability preceding dose reduction often managed in primary care. Third, the baseline antipsychotic dose was relatively low, suggestive of a more stable cohort with a higher degree of symptom control. Fourth, the inclusion criteria selected participants that were considered by a clinician as not posing a serious risk of harm to self or others if their antipsychotic medication was reduced, and excluded those with insufficient spoken English. These criteria potentially favour inclusion of those participants having positive relationships with treating teams. The language criterion further excludes communities that are less well integrated into healthcare services. Indeed, variability in participation rates between the different participating NHS trusts was sufficient to impact the planned analyses, suggesting differences in the profile of people recruited into the study because it is unlikely this was due to differing illness profiles between sites. Finally, the use of broad metrics of social functioning and employment status may not be sufficiently subtle to reflect the functional impacts of relapse, while the limited social opportunities during the pandemic period may have reduced the number of options for all participants and thus the potential for differences between the more and less able groups.

There is a clear need for clinicians to tailor longer-term antipsychotic treatment to the individual – balancing patient choice and likelihood of adverse effects against the functional and neurological risks related to relapse. The extant data reflect risks of antipsychotic withdrawal at a group level but offer little guidance for an individual with schizophrenia. Thus, decisions about antipsychotic reduction are more a balance between risk mitigation and management during medication withdrawal rather than choice. This discussion speaks to the need for a better understanding of the prediction of outcomes following medication withdrawal, including with respect to self-harm, violence and social and work relationships; and is linked to assessment of the psychological and neurobiological change associated with periods of relapse, to develop a more personalised approach to this complex issue.