A CP was associated with older age, higher CD34 + cell count, higher prevalence of BM fibrosis grade 3, lower JAK2 mutant burden, TN status, and U2AF1 mutations. Patients with ≥2 cytopenias were more likely to have karyotype abnormalities and mutations in CBL and U2AF1.

After a median follow-up of 94 (95% CI 79–115) months, 118 (55%) deaths were reported, with a median OS of 65 (95% CI 54–87) months. The OS of patients with CP (median 54 months, 95% CI 44–72) was significantly shorter compared to the proliferative group (median 96 months, 95% CI 64–139) (HR 1.7, 95% CI 1.2–2.4, P = 0.0026) (Fig. 1E). Patients harboring ≥ 2 cytopenias had an inferior OS (median 43 months, 95% CI 19–55) compared to patients with one sole cytopenia (median 64 months, 95% CI 45–76) (HR 1.9, 95% CI 1.1–3.2, P = 0.0146) (Supplemental Fig. 2A). Remarkably, a severe CP was associated with significantly inferior OS compared to patients with not-severe cytopenias (HR 2.9, 95% CI 1.7–4.8, P <0.0001), with median of 28 (95% CI 19–47) and 72 (95% CI 52–91) months, respectively (Supplemental Fig. 2B). Upon multivariate Cox proportional hazards analysis, severe thrombocytopenia, severe anemia, PB blast count ≥ 2%, HMR category and ≥2 HMR mutated genes independently predicted for inferior OS (Supplemental Table 2); severe thrombocytopenia showed the highest HR (5.8, 95% CI 2.5–13.7).

At last follow-up, a total of 28 (14%) patients transformed to acute leukemia. After competing risk analysis, the CuI of LT was not statistically different among cytopenic and proliferative patients, with 5-year rates of 15% (95% CI 8–23) and 12% (95% CI 6–20), respectively (Fig. 1F). The number and severity of cytopenias did not impact the CuI of LT (Supplemental Fig. 2C, D), although there was a trend for patients with severe compared to not-severe cytopenias (5-year CuI of LT 23%, 95% CI 10–38 and 10%, 95% CI 4–20, respectively; Grey test P = 0.0719).

In summary, the current study provides a comprehensive analysis of the CP in a large cohort of WHO-defined pre- and overt PMF. We showed that cytopenic features, that are more common in overt than pre-PMF, are associated with distinct high-risk clinical and molecular features predominantly in pre-PMF. Of note, U2AF1 mutations emerged as a distinct abnormality of CP in both PMF subtypes, suggesting that they might contribute to ineffective hematopoiesis and reinforcing their adverse prognostic role [13, 14]. A CP was associated with inferior OS in both PMF subtypes, and with a higher risk of LT in pre-PMF. While in pre-PMF the adverse prognostic impact of a CP was independent of the number and severity of cytopenias, in overt PMF the impact on OS seemed to be affected mainly by the CP severity, with severe thrombocytopenia having the greatest impact. Finally, we highlighted that a CP is an important risk factor for fibrotic progression in patients with pre-PMF, particularly for those presenting with anemia and thrombocytopenia. Overall, our results further expand the characterization of the cytopenic features in PMF with novel insights as regards the distinction between pre- and overt PMF. Despite the limitations associated with its arbitrary definition, identification of the CP is straightforward, does not require invasive or advanced technologies and, above all, can be performed longitudinally.

Cytopenia represents a significant challenge in the contemporary management of PMF. Currently, there are few agents aimed at treating cytopenic PMF, including immunomodulatory drugs, hypomethylating agents, and JAK inhibitors such as momelotinib and pacritinib, and development of new agents specifically tailored to this patient population remains an unmet need. The association with U2AF1 mutations may prompt the study of splicing modulators [14].