Murray and colleagues' confident advice ^{Reference Murray, Quattrone, Natesan, van Os, Nordentoft and Howes1} to psychiatrists, encouraging them to leave fewer patients with schizophrenia on long-term medication, is based on one of several possible interpretations of a selected literature, and little clinical evidence. We consider that at times Murray and colleagues misrepresented the literature in their descriptions of what some papers report; sometimes these descriptions are misleading (e.g. their references to Mace et al (2015), Vita et al (2015) and Boonstra et al (2011)) or incorrect (Saha et al (2007)). If the evidence so strongly supports the authors' recommendations, why have they relied so heavily on single case reports and personal communications, and on qualifying words such as ‘doubts’, ‘possibilities’, ‘suggest’, ‘appear’, ‘raise the possibilities’, ‘several Japanese groups have suggested’? Why have they given prominence to a study (Harrow et al (2014)) that they admit has a ‘major confounder’, and another (Wunderink et al (2013)) that they describe as ‘a study less open to bias’, which others consider grossly flawed (reviewed by Catts & O'Toole ^{Reference Catts and O'Toole2} )? And why do they consistently refer to use of low or no doses, without ever describing the conditions that discriminate these indications?

Murray and colleagues assert that continuous antipsychotic medication loses its effectiveness over time, but do not present any clinical evidence for this, and that this putative treatment resistance is due to antipsychotic-induced dopamine receptor supersensitivity that has been found in animal studies. Indeed, the authors rely heavily on animal studies generally to make their case for a range of issues, without highlighting the fact that the relevant animal studies were all carried out on healthy animals. The authors seem overly confident that the results of these animal studies can be applied directly to the clinical situation, although no psychiatrist uses antipsychotic medication in healthy humans. The authors fail to see the complete disconnect between healthy animal research and clinical research on patients with schizophrenia. It seems to us that the reliance on animal studies by Murray and colleagues has resulted in their treatment recommendations being almost the opposite of others based on clinical literature. ^{Reference Catts and O'Toole2}

We suggest that Murray and colleagues are proposing a therapeutic dead end. With current practice, most patients stop their medication anyway (mainly owing to non-adherence to oral medication) – 60% of patients with first-episode psychosis do so within 60 days of hospital discharge: ^{Reference Tiihonen, Haukka, Taylor, Haddad, Patel and Korhonen3} so how will taking more patients off their antipsychotic medication improve the current overall recovery rates in schizophrenia of 13.5%, ^{Reference Jaaskelainen, Juola, Hirvonen, McGrath, Saha and Isohanni4} and the death rates that all agree are unacceptably high? Murray and colleagues' answer is more psychosocial intervention, but they present no evidence for the effectiveness of such intervention in unmedicated patients. The clinical evidence for antipsychotic medication reducing the mortality rate at all stages of the illness is of high quality and very consistent (summarised by Tiihonen ^{Reference Tiihonen5} ); the simple truth is that taking more patients off maintenance medication will result in more patients dying unnecessarily – the ultimate therapeutic dead end.