When Platelet Counts Soar Past a Million, Hematologists Split on How to Act

For most people, a routine blood test is a matter of curiosity. For patients with essential thrombocythemia, a chronic blood cancer in which the bone marrow produces too many platelets, the numbers on a complete blood count can dictate the course of a lifetime of monitoring and therapy. When platelet counts climb into the extreme range, exceeding one million per microliter of blood, clinicians face a set of questions that guidelines have never fully resolved. A new survey of hematologists, published as a brief report in the Annals of Hematology, offers one of the clearest snapshots yet of how specialists actually behave at the bedside when platelets reach a million and beyond, and it adds an unusual twist by comparing their answers with recommendations generated by artificial intelligence programs.

The study, conducted by Özge Sönmez and Ahmet Emre Eşkazan of the Division of Hematology at Cerrahpaşa Faculty of Medicine, Istanbul University-Cerrahpaşa, used a web-based questionnaire to probe how hematologists manage extreme thrombocytosis, the technical term for platelet counts that far exceed the normal range of roughly 150,000 to 450,000 per microliter. Fifty-six hematologists participated in the survey, which was supported by the Turkish Society of Hematology and approved by the local clinical studies ethics committee at Istanbul University-Cerrahpaşa. The authors set out to map real-world clinical behavior in a domain where evidence is thin, expert opinion varies widely, and the consequences of over-treatment or under-treatment can be serious for patients with essential thrombocythemia, one of the classical Philadelphia-negative myeloproliferative neoplasms.

The central finding concerns the platelet threshold at which physicians choose to intensify therapy. Cytoreductive treatment, which uses medications such as hydroxyurea to suppress platelet production in the marrow, was favored by 41.1 percent of respondents once platelet counts reached or exceeded 1000 × 10⁹ per liter, the celebrated million-platelet mark. But when the threshold was raised to 1500 × 10⁹ per liter, the proportion of hematologists opting for cytoreduction jumped to 83.9 percent. In other words, the million mark alone is not a universal trigger for action; it is a point of debate. Only when counts approach one and a half million does a strong majority of specialists converge on drug-based platelet lowering, a pattern that mirrors the cautious, risk-adapted philosophy embedded in contemporary myeloproliferative neoplasm guidelines.

Why does the million mark inspire such hesitation? The answer lies in the biology of extreme thrombocytosis. Very high platelet counts do not simply increase the risk of blood clots; paradoxically, they also raise the risk of bleeding. The mechanism is acquired von Willebrand disease, a condition in which the sheer excess of circulating platelets adsorbs and proteolytically cleaves von Willebrand factor, the adhesive protein that platelets and clotting depend on. As platelets multiply, they effectively devour their own hemostatic infrastructure, and the largest von Willebrand multimers disappear first. A patient with a platelet count of a million or more may therefore present with gastrointestinal bleeding or bruising rather than thrombosis, and giving aspirin to such a patient in the belief that high platelets mean high clot risk can precipitate dangerous hemorrhage.

This biological nuance explains the second pillar of the survey: laboratory assessment. A striking 78.6 percent of hematologists reported testing for von Willebrand factor ristocetin cofactor activity, abbreviated vWF:RCo, at the time of diagnosis of extreme thrombocytosis. This assay measures the functional capacity of von Willebrand factor to support platelet adhesion under shear stress, and it is the workhorse test for detecting the acquired form of the deficiency. The near-universal uptake of this test among survey respondents suggests that the field has internalized the lesson that platelet number alone is an unreliable guide to bleeding risk, and that documenting acquired von Willebrand disease is now considered a standard part of the initial evaluation when counts soar into the extreme range.

Aspirin decisions revealed an equally disciplined logic. In the presence of cardiovascular risk factors and in the absence of acquired von Willebrand disease, 91.1 percent of hematologists favored aspirin therapy for patients with platelet counts of at least 1000 × 10⁹ per liter, and 78.6 percent favored it at counts of at least 1500 × 10⁹ per liter. The slight decline at the higher threshold is telling: the very patients with the most extreme counts are also the most likely to harbor acquired von Willebrand disease, so the antiplatelet drug that protects one patient’s arteries becomes a hemorrhagic hazard for another. The survey thus captures, in numerical form, the delicate balancing act that defines extreme thrombocytosis management, where the same count can argue for and against the same drug depending on a single laboratory result.

When asked why they initiate cytoreductive treatment at all, respondents cited three goals with roughly equal frequency: prevention of thrombosis, prevention of bleeding, and control of symptoms such as headaches, erythromelalgia, and visual disturbances. This even distribution is itself informative. It shows that hematologists do not view cytoreduction solely as an anti-thrombotic strategy. Lowering the platelet count can restore von Willebrand factor levels and relieve bleeding, and it can also ameliorate the microvascular symptoms that plague many patients with essential thrombocythemia regardless of their thrombotic risk score. The therapeutic rationale, in practice, is tripartite, and the survey suggests that experienced clinicians weigh all three aims rather than fixating on any single one.

The most novel element of the report is its comparison with artificial intelligence. The authors posed the same clinical scenarios to AI programs and recorded the recommendations the models produced. The AI-generated advice supported von Willebrand testing, endorsed cytoreductive treatment particularly at platelet counts of 1500 × 10⁹ per liter or above, and generally recommended aspirin only for patients with counts of at least 1000 × 10⁹ per liter who did not have acquired von Willebrand disease. Across several clinical scenarios, the differences between the human specialists and the machine recommendations were few. The authors are careful to frame this comparison as exploratory, and they stop short of claiming that AI can substitute for clinical judgment, but the convergence is noteworthy at a time when large language models are being proposed as clinical decision aids.

The significance of this convergence deserves careful interpretation. On one hand, it suggests that the consensus positions in hematology, such as testing for acquired von Willebrand disease and reserving aggressive cytoreduction for the most extreme counts, are encoded deeply enough in the published literature that AI systems trained on that literature reproduce them faithfully. On the other hand, it raises the question of whether AI is genuinely reasoning or simply echoing the dominant patterns of its training data, including their blind spots. Where expert opinion is divided, as it clearly is at the million-platelet threshold where only about four in ten specialists opt for cytoreduction, an AI that confidently recommends a single course of action could flatten legitimate clinical nuance into false certainty. The survey’s exploratory design is a prudent acknowledgment of that risk.

For patients with essential thrombocythemia, the practical message of the study is reassuring in one respect and cautionary in another. Reassuring, because the management of extreme thrombocytosis appears to follow a coherent, biologically informed playbook: check von Willebrand function, tailor aspirin to its result, and reserve cytoreductive drugs chiefly for the most extreme counts or for patients with thrombosis, bleeding, or disabling symptoms. Cautionary, because the survey also exposes genuine variability at the margins, with more than half of specialists declining to treat at exactly one million platelets while others intervene. As Sönmez and Eşkazan conclude, extreme thrombocytosis itself remains an important driver of treatment decisions, particularly above 1500 × 10⁹ per liter, and the near-agreement between experienced hematologists and AI recommendations offers a provocative glimpse of a future in which algorithms may help standardize care, provided they are validated rigorously and deployed with the same humility that the survey’s authors bring to their own exploratory analysis.

Subject of Research: Clinical management of extreme thrombocytosis in essential thrombocythemia, comparing hematologist survey responses with artificial intelligence recommendations

Article Title: When platelets reach a million: insights from hematologists

Article References: Sönmez, Ö., & Eşkazan, A. E. (2026). When platelets reach a million: insights from hematologists. Annals of Hematology. https://doi.org/10.1007/s00277-026-07183-x

Image Credits: AI Generated

DOI: 10.1007/s00277-026-07183-x

Keywords: essential thrombocythemia, extreme thrombocytosis, platelets, cytoreductive therapy, acquired von Willebrand disease, aspirin, von Willebrand factor, myeloproliferative neoplasms, hematology, artificial intelligence, clinical survey, thrombosis