Clozapine's neutropenia risk is front-loaded. Twenty years of Hong Kong data show when.

Clozapine is the only antipsychotic with established superiority in treatment-resistant schizophrenia, and it is chronically underused. The most commonly cited reason is the mandatory lifelong haematological monitoring imposed in many jurisdictions because of the risk of severe neutropenia [s1]. A 20-year population-based study published in The British Journal of Psychiatry on 28 August puts numbers on when that risk actually occurs — and the shape of the curve is the finding [s1].

The study

Researchers used Hong Kong's electronic health records to identify patients starting clozapine or a non-clozapine second-generation antipsychotic between 2003 and 2019, following them to 2023 [s1]. Outcomes were minor neutropenia, defined as an absolute neutrophil count of 1.0 to 1.5 × 10⁹/litre, and serious neutropenia, below 1.0 × 10⁹/litre; serious neutropenia leading to clozapine discontinuation within six weeks was classed as clozapine-associated [s1].

The cohort comprised 4,868 clozapine users and 38,277 users of other second-generation antipsychotics [s1].

What it found

Unadjusted incidences were higher on clozapine: minor neutropenia 7.05% versus 3.74%, serious neutropenia 2.53% versus 1.45% [s1]. Adjusted rates of serious neutropenia, however, were essentially the same in the two groups — 0.37 versus 0.36 per 100 person-years [s1].

The difference between those two comparisons is timing. In the first 18 weeks, clozapine carried a substantially higher risk of both minor neutropenia (incidence rate ratio 5.91; 95% CI, 3.96–8.70) and serious neutropenia (IRR 3.48; 95% CI, 1.70–6.84) [s1]. Bayesian change-point analysis showed the clozapine-associated risk declining sharply after 26 to 29 weeks and converging with non-clozapine second-generation antipsychotics somewhere between 82 and 135 weeks — roughly two years [s1].

Two subgroup observations complete the picture. No significant early excess risk was found in patients younger than 45, and males experienced a transient risk of minor neutropenia [s1].

The authors' recommendation follows directly: weekly monitoring for 18 weeks and monthly for up to two years, with potential for personalised schedules based on age and sex [s1].

Why the monitoring question is not settled anywhere

The variation between countries is larger than most prescribers realise. A 2022 international survey compared haematological monitoring guidelines across 102 countries — 35 in Europe, 24 in Asia, 20 in Africa, 11 in South America, seven in North America, and five in Oceania and Australia — and found they differ in monitoring frequency and in the thresholds that trigger discontinuation [s2].

Some of the gaps in that survey are striking. Only 5% of the included countries had explicit guidelines for clozapine rechallenge and 40% explicitly prohibited it [s2]. Just 7% had modified discontinuation thresholds for benign ethnic neutropenia [s2] — a distinction that determines whether a patient whose baseline neutrophil count is constitutionally low can start the drug at all. And none of the guidelines specified how long haematological monitoring should continue [s2].

That last omission is precisely the gap the Hong Kong analysis addresses. Monitoring protocols have generally been open-ended not because evidence supports indefinite testing but because no one had characterised when the risk ends.

The 2022 survey also found that the most stringent guidelines were in Europe and the least stringent in Africa and South America, with a positive association between a country's stringency score and its healthcare expenditure per capita (r = 0.43, P < 0.001) [s2]. Stringency, in other words, tracks what a health system can afford to test rather than what the drug does.

What it does not establish

This is a retrospective cohort in one health system, and the definition of "clozapine-associated" serious neutropenia — discontinuation within six weeks of the event [s1] — is a proxy that inherits whatever the local discontinuation practice was. A jurisdiction with a lower discontinuation threshold would classify differently.

More importantly, an observational study of people who stayed on clozapine under weekly monitoring cannot tell you what would happen if the monitoring stopped. The declining risk curve is measured inside a system that was catching and acting on falling counts. It is evidence about when events occur, not a demonstration that later monitoring is dispensable.

What it does support is the narrower and still consequential claim that the current global default — indefinite, unspecified-duration testing [s2] — is not derived from the timing of the risk it exists to catch.

Sources

• [s1] "Incidence of clozapine-associated neutropenia in patients with schizophrenia: 20-year population-based study in Hong Kong," The British Journal of Psychiatry, 28 August 2026. https://doi.org/10.1192/bjp.2026.10726
• [s2] "Clozapine haematological monitoring for neutropenia: a global perspective," Epidemiology and Psychiatric Sciences, 25 November 2022. https://doi.org/10.1017/S204579602200066X