Clozapine is the only antipsychotic shown to reduce overall mortality and suicide in schizophrenia — and yet it carries the heaviest monitoring burden in psychiatry, built almost entirely around a single rare complication. This episode asks a more useful question: when people treated with clozapine die, what do they actually die of, and how much of it could we have prevented? The discussion follows a 17-year retrospective cohort study of every clozapine-treated patient who died under Cambridgeshire and Peterborough NHS Foundation Trust between 2009 and 2025 — 87 deaths across a catchment of roughly one million people and a treated population of around 410 at any one time. Each death was reviewed individually and assigned to one of three predefined clusters: suicide, expected deaths from chronic or progressive medical illness, and non-intentional unexpected deaths. The distribution is the first surprise. Suicide accounted for 11 deaths (12.6%), an annual rate of about 0.15% — at the low end of what is expected in treatment-resistant schizophrenia, consistent with clozapine's well-established protective effect. Expected deaths made up 29 cases (33.3%). The largest group, 47 deaths (54.0%), was non-intentional and unexpected: sudden cardiac or respiratory compromise, gastrointestinal obstruction, infection, metabolic disturbance, and a substantial number — 12 cases — where no definitive cause could be established even after post-mortem and detailed clinical review. By cause, malignancy led (21 of 87, 24.1%), predominantly solid tumours of the lung, breast, pancreas and oesophagus, several of them tobacco-related; only one death was attributable to a haematological malignancy. Cardiovascular disease followed (14.9%), then respiratory and infective causes including pneumonia (10.3%). Median duration of clozapine treatment at death was 14 years. The episode then turns to what distinguished those who died unexpectedly. Compared with a clozapine-treated comparison cohort alive in 2019, they were significantly older at the reference point (55.0 vs 48.6 years) and markedly more likely to smoke (75.0% vs 34.6%). Prescribed clozapine dose and high-dose exposure did not differ significantly — a finding that pushes the explanation away from pharmacology alone and towards accumulated physical and behavioural risk. Temporal patterns add a further layer. Mortality peaked in 2021 with 20 deaths, well after the first pandemic wave, and only four deaths across the whole cohort had COVID-19 documented as a direct or contributing cause. The excess is therefore hard to attribute to the virus itself, and more plausibly reflects disrupted physical health monitoring, reduced access to services and delayed help-seeking. Uptake of the annual primary care physical health check moved in parallel — between 16.7% and 29.0% before the pandemic, more variable and generally higher afterwards. The conversation closes on the clinical implication, which is deliberately awkward for current practice. If more than half of deaths are unexpected and driven by smoking, cardiometabolic disease, infection and gastrointestinal complications, then the mandatory haematological monitoring may be doing much of its work indirectly — as scaffolding for regular clinical contact, physical health review and early detection — rather than through the blood counts themselves. Any move to relax monitoring for stable patients, an idea gaining international traction, therefore needs to be paired with a deliberate redesign of clozapine services that preserves that contact, protects cardiometabolic and infectious disease surveillance, and takes smoking cessation seriously as a mortality intervention. A frank discussion of where the real risks lie in clozapine treatment — and why the monitoring we have may be right for the wrong reasons.
