The fear of a sudden collapse in white blood cells has shadowed clozapine for half a century, and it is the main reason the world’s most effective schizophrenia drug goes unused. Now one of the largest real-world investigations ever conducted has mapped exactly when that danger peaks and who is most exposed. In a nationwide cohort of 12,810 people starting clozapine in Taiwan, an international research team found that serious neutropenia forcing patients off the drug is rare, that the danger is heavily concentrated in the opening months of therapy, and that the risk curve bends decisively downward at roughly week 15. The study, published in The Lancet Regional Health – Western Pacific by Chi-Shin Wu of the National Health Research Institutes in Taiwan, Korinne Northwood and Dan Siskind of the University of Queensland, and colleagues, arrives at a moment when regulators in the United States and Europe are dismantling decades of restrictive blood-monitoring rules, and it supplies the kind of population-scale evidence they have been lacking.
Clozapine occupies a paradoxical place in psychiatry. For the roughly one in three patients with schizophrenia whose symptoms shrug off two or more standard antipsychotics, guidelines worldwide agree that clozapine should be the next step, and trial evidence shows it outperforms every alternative in reducing symptoms, preventing relapse, and lowering mortality. Its pharmacological profile is singular: minimal liability for the movement disorders that plague other antipsychotics, and possible protective effects against suicide and aggression. Yet clinicians persistently underuse it. The reason is haematologic toxicity. Neutrophils, the granulocyte front line of innate immunity, can plummet during clozapine treatment, a condition called neutropenia, and in its most severe form, agranulocytosis, the count can collapse to near zero, exposing patients to overwhelming infection. Cumulative estimates suggest neutropenia strikes roughly 3 to 4 percent of clozapine-treated patients and agranulocytosis about 0.9 percent. Those figures, small as they are, spawned mandatory blood-count surveillance programs on every continent, and those programs, while effective at preventing deaths, impose weekly clinic visits, laboratory waits, and bureaucratic checkpoints that delay treatment for months or push patients off the drug entirely.
The new study exploits an unusually complete data infrastructure. Taiwan’s National Health Insurance program, established in 1995, covers about 23 million people, roughly 99 percent of the population, and its claims database records every diagnosis and prescription dispensed through hospital and affiliated pharmacies. Because Taiwanese regulations mandate haematologic monitoring for all clozapine users, weekly blood counts for the first 18 weeks and at least every four weeks thereafter, the researchers could link prescription records with laboratory results across the entire country. Starting from more than 46 million claims filed by 1.6 million people with major psychiatric disorders between 2014 and 2023, they isolated 41,735 patients with at least one clozapine prescription, then defined new users as those beginning treatment after 2015 following a 365-day washout with no prior exposure. Requiring a baseline white blood cell or absolute neutrophil count plus at least one follow-up test produced the final analytic cohort of 12,810 people, who together contributed 50,062 person-years of observation. A parallel group of 6,153 re-initiation episodes from 4,850 patients who restarted clozapine after an interruption of at least 42 days allowed the team to examine whether restarting the drug resets the danger clock.
Methodologically, the study goes beyond simple event counting. To locate precisely when risk changes over the course of treatment, the researchers applied joinpoint regression, a technique that fits piecewise linear segments to weekly incidence rates on a log scale, using Poisson models with person-time offsets, and selects the number of trend inflections with the Bayesian Information Criterion to avoid overfitting. For risk factors, they used multivariable cause-specific Cox proportional hazards models, treating each neutropenia subtype as a competing event for the others and death as a competing risk, with concomitant medications modelled as time-varying exposures over rolling 28-day prescription windows. Because Taiwanese guidelines allow monitoring with either the absolute neutrophil count or the total white blood cell count, and because ANC measurements were not universally recorded, the team validated WBC thresholds as a proxy: against concurrent ANC measurements, WBC-based classification agreed 98.9 percent of the time (Cohen’s kappa 0.61), with 99.6 percent specificity, though only 54.1 percent sensitivity, a limitation the authors addressed by repeating every analysis in the 10,510-patient subgroup with ANC-defined outcomes, which produced consistent results.
The headline numbers are reassuring. Across follow-up, the cohort recorded 524 isolated minor neutropenia events, defined as ANC 1000–1500 per microlitre or WBC 2500–3000 per microlitre, for an incidence of 10.5 per 1000 person-years. Serious neutropenia, ANC below 1000 or WBC below 2500, occurred far less often: 84 events led to clozapine cessation within six weeks, an incidence of just 1.7 per 1000 person-years, while 144 serious events did not prompt discontinuation, typically because an alternative cause such as acute infection was identified, at 2.9 per 1000 person-years. Re-initiators showed nearly identical rates. In other words, for a typical patient starting clozapine today, the annualized probability of a blood-count crisis severe enough to end treatment sits well below one percent, and most of whatever risk exists is front-loaded into the first months.
The temporal mapping is the study’s most consequential contribution. Joinpoint analysis placed the single inflection point for isolated minor neutropenia at week 14 and for serious neutropenia leading to cessation at week 15 among new users, with a sharp early bend at week 6 for serious events without cessation. Re-initiators showed the same early pattern, with inflections at weeks 11 and 19. After those bends, incidence fell substantially and stayed low, echoing registry and pharmacovigilance data from the United States, the United Kingdom, Europe, Australia, and New Zealand, where most serious events cluster within the first 18 weeks and risk becomes very low after about two years. The Taiwanese team is careful to note that weekly early testing probably inflates early detection of transient abnormalities, so the curve reflects both biology and surveillance intensity, but the alignment with Western datasets collected under different monitoring rules argues the pattern is not an artifact of the Taiwanese schedule.
Risk stratification proved even starker than the time course. Baseline haematologic status was the strongest predictor in the entire analysis: compared with patients starting above 8000 white cells per microlitre, those starting below 4000 faced a 13.29-fold higher hazard of minor neutropenia, an 11.42-fold higher hazard of serious neutropenia causing cessation, and a 9.90-fold higher hazard of serious neutropenia without cessation, with incidence in the lowest band reaching 61.7 minor events per 1000 person-years. Age acted as a second powerful gradient: patients aged 55 and older had 3.11 times the risk of minor events and 6.41 times the risk of treatment-ending serious events compared with those under 35, a pattern the authors attribute plausibly to declining haematopoietic reserve, comorbidity, and cumulative medication exposure. Lower income was linked to minor neutropenia only, possibly reflecting poorer baseline health and nutrition.
The concomitant medication findings carry direct prescribing implications. Valproic acid, taken by about 39 percent of the cohort, was associated with roughly double the hazard of minor neutropenia and an 81 percent higher hazard of serious cessation events, consistent with prior case-control evidence of additive myelosuppression. Carbamazepine, though prescribed to only a small fraction, showed the strongest serious-event associations, tripling the hazard of cessation-level neutropenia, a warning that matters because the drug’s own known capacity to cause agranulocytosis compounds clozapine’s. Benzodiazepines and other antipsychotics were linked to minor events only, while diuretics and corticosteroids, likely markers of acute illness and heightened testing rather than direct bone-marrow toxins, were associated with selected serious outcomes. The authors are explicit that these are risk markers shaped by confounding, not proven causal effects, but they give clinicians concrete signals for which co-prescriptions warrant vigilance.
Equally important is what the study says about restarting the drug. Re-initiation after an interruption of six weeks or longer did not materially raise neutropenia incidence, and 99.3 percent of re-initiation episodes occurred in patients with no documented neutropenia history. Patients who had previously experienced neutropenia did show numerically higher recurrence rates, particularly of serious events without cessation, but only 46 such episodes existed in the dataset, too few for precision. The authors therefore stop short of endorsing rechallenge in prior-neutropenia patients, calling instead for pooled analyses, though the population-level signal offers reassurance that treatment gaps do not inherently reactivate the early-period hazard.
The findings land amid a genuine regulatory shift. In 2025 the US Food and Drug Administration eliminated the Clozapine Risk Evaluation and Mitigation Strategy, the REMS program that had funneled all prescribing and dispensing through a centralized authorization system, and the European Medicines Agency has formally revised its monitoring framework, retaining weekly counts for the first 18 weeks and monthly counts to one year but allowing intervals to stretch to every 12 weeks after one year and annually after two years in patients without a neutropenia history. This study supplies independent, population-based confirmation from an East Asian health system that such de-escalation aligns with the underlying epidemiology, and it argues the logic further: monitoring intensity could be personalized not only by time since initiation but by age and baseline white cell count, with older patients and those starting below 4000 cells per microlitre meriting the closest surveillance while lower-risk patients could safely face a lighter long-term burden. The team frames its results as descriptive evidence supporting, not itself prescribing, future trials of risk-stratified schedules, and acknowledges residual confounding by unmeasured factors such as acute infection and the constraints of prescription-based exposure data. But the core message stands with unusual clarity for a drug-safety question: clozapine’s feared toxicity is uncommon, front-loaded, and predictable from a single baseline blood test and a patient’s age, and that predictability is precisely what a smarter monitoring system should be built to exploit.
Subject of Research: Temporal patterns and risk factors of clozapine-associated neutropenia in a nationwide population-based cohort of new and re-initiating clozapine users in Taiwan
Subject of Research: Medicine
Article Title: Temporal patterns and risk factors of clozapine-associated neutropenia: a nationwide population-based cohort study
Article References: Wu, C.-S., Northwood, K., Wu, M.-S., Wang, S.-H., Huang, W.-L., Liu, C.-C., & Siskind, D. (2026). Temporal patterns and risk factors of clozapine-associated neutropenia: a nationwide population-based cohort study. The Lancet Regional Health – Western Pacific, 74, Article 101970. https://doi.org/10.1016/j.lanwpc.2026.101970
Image Credits: AI Generated
DOI: 10.1016/j.lanwpc.2026.101970
Keywords: clozapine, neutropenia, agranulocytosis, treatment-resistant schizophrenia, white blood cell count, haematologic monitoring, joinpoint regression, drug safety, Taiwan National Health Insurance, antipsychotic treatment, risk stratification, pharmacovigilance
Cite Scienmag News
APA MLA Chicago
Ophelia Keating. (August 30, 2026). Nationwide study maps when clozapine triggers neutropenia and who’s at risk. Scienmag. https://scienmag.com/nationwide-study-maps-when-clozapine-triggers-neutropenia-and-whos-at-risk/
Ophelia Keating. “Nationwide study maps when clozapine triggers neutropenia and who’s at risk.” Scienmag, 30 August 2026, https://scienmag.com/nationwide-study-maps-when-clozapine-triggers-neutropenia-and-whos-at-risk/. Accessed 30 August 2026.
Ophelia Keating. “Nationwide study maps when clozapine triggers neutropenia and who’s at risk.” Scienmag. August 30, 2026. https://scienmag.com/nationwide-study-maps-when-clozapine-triggers-neutropenia-and-whos-at-risk/
Copy citation Download RIS
Tags: blood monitoring guidelines for clozapineClozapine neutropenia riskClozapine neutropenia risk assessmentduration of neutropenia risk after clozapine initiationearly detection of neutropenia in antipsychotic therapyinternational regulatory changes in blood monitoring for clozapinenationwide cohort analysis of clozapine-induced neutropenianationwide cohort study on clozapine safetypopulation-based study on clozapine safetypopulation-scale evidence on clozapine safetyreal-world data on clozapine adverse effectsreal-world evidence on clozapine side effectsregulatory changes in blood monitoring for clozapinerisk factors for clozapine-induced neutropeniarisk factors for neutropenia in schizophrenia treatmentrisk management intiming of neutropenia onset in clozapine therapytiming of neutropenia onset in schizophrenia treatment
