Williams and colleagues ^{Reference Williams, Penington, Gupta, Rickett, Agorinya and Tsiachrista1} provide the most granular national evidence to date that what early intervention in psychosis (EIP) services do matters – not just whether they exist. Their study links audit and routine National Health Service (NHS) data across nearly 15 000 individuals to show that specific care components of EIP are associated with measurable differences in coercive outcomes. Among the components examined, the findings on clozapine are arguably the most consequential. Eligible patients who were not offered clozapine had a 51% higher hazard of psychiatric intensive care (PICU) admission, and a 139% higher hazard of physical restraint. Those who received it were indistinguishable from ineligible patients. The therapeutic ceiling, in other words, is not a property of the drug, and not where we have set it; it reflects the point at which current services stop short.

Although Williams and colleagues did not set out to examine haematological monitoring directly, we introduce it here as wider context and interpretation rather than as an outcome they assessed. It is tempting to read this as another argument for streamlining haematological monitoring. We have argued elsewhere that the 1990 protocol – weekly full blood counts for 18 weeks, then fortnightly, then monthly indefinitely – is disproportionate to actual long-term agranulocytosis risk, which becomes negligible after roughly 2 years of exposure. ^{Reference Fernandez-Egea and McCutcheon2} Reform of these requirements is overdue, and a recent global Delphi consensus has set out how neutrophil thresholds and monitoring frequency might safely be reduced. ^{Reference Siskind, Northwood, Pillinger, Chan, Correll and Cotes3} But framing clozapine under-prescription as a monitoring problem alone risks misdiagnosing the disease.

Only 4.2% of the Williams cohort received clozapine and 1.9% refused it, against a treatment-resistance rate estimated at roughly 20–25% in first-episode cohorts, ^{Reference Siskind, Orr, Sinha, Yu, Brijball and Warren4} though this figure varies with the definition used, duration of follow-up and diagnostic composition of the population, given that up to 18% of people presenting to EIP services have a primary affective psychosis. ^{Reference Catalán, Swidzinski, McCutcheon, Aymerich, Pedruzo and Knox5} The gap reflects a constellation of barriers that go beyond blood tests. Some are conceptual: within EIP, where the dominant narrative is one of recovery and trajectory change, we suggest that the prospect of initiating clozapine may be experienced as a ‘defeat’ – an acknowledgement that the early intervention paradigm has failed for this particular patient – rather than as the timely use of the most effective treatment available for a clearly identifiable subgroup. Treatment resistance thereby becomes a phenomenon to be deferred rather than recognised. Other barriers are infrastructural: clozapine initiation requires the kind of physical health monitoring, side-effect anticipation and coordinated medical input that mental health services have historically outsourced rather than embedded. Where physical health is not already integrated into routine mental health care, the additional clinical scaffolding clozapine demands looks exceptional, when it is in fact ordinary medicine.

That delay carries measurable cost. In our own cohort of 254 clozapine-treated patients, earlier initiation was associated with better outcomes across positive symptoms, negative symptoms, functioning and well-being. ^{Reference Muñoz-Manchado, Perez-Revuelta, Banerjee, Galindo-Guarin, Bernardo and González-Saiz6} Yoshimura and colleagues ^{Reference Yoshimura, Yada, So, Takaki and Yamada7} similarly identified a critical window of approximately 3 years before clozapine-resistant illness develops – a window that maps almost exactly onto the duration of EIP care. The Williams findings now anchor this in hard service-level outcomes: PICU admission, restraint and police involvement, falling disproportionately on Black, mixed-ethnicity and younger patients. The inequity is therefore not incidental; it is a predictable consequence of under-treatment concentrated in already disadvantaged groups.

The forthcoming reduction in monitoring requirements is therefore an opportunity to redesign rather than merely to deregulate. ^{Reference Fernandez-Egea8} We have shown that community-based clozapine initiation supported by wearable physiological monitoring, remote consultations and structured staff check-ins is feasible and acceptable, improving patient engagement and enabling earlier detection of adverse effects; its effect on restrictive practice and on inequities in coercive care remains to be tested. ^{Reference Krishnadas, Saunders, Donnelly, Okubadejo, Ayadurai and Meiser-Stedman9} The same logic extends to the broader challenge Williams and colleagues identify, that of intercepting deterioration before it reaches the PICU. Released capacity should be reinvested upstream: in reframing clozapine as a continuation of evidence-based EIP care rather than its endpoint; in operationalised treatment-resistance screening within EIP teams; and in integrating physical health monitoring so that the demands of clozapine initiation feel like routine care. Technology-enabled community support extends this logic further, shifting the system’s point of intervention before crisis becomes inevitable. The choice facing UK services is not whether to prescribe clozapine more often, but whether to build services in which timely prescribing is the path of least resistance. Williams and colleagues have shown the cost of the current arrangement.