Bleeding in essential thrombocythaemia: a retrospective analysis on 565 patients


Essential thrombocythaemia (ET) is a chronic myeloproliferative neoplasm characterized by persistent thrombocytosis, associated with an increased predisposition to vascular complications (Vardiman et al, 2002).

While older age (≥60 years) and previous thrombosis are the established risk factors for thrombosis (Barbui et al, 2004), there are few data available regarding haemorrhages, which occur in 5–30% of patients (Wolanskyj et al, 2005; Griesshammer, 2006; Vannucchi et al, 2007). Therapeutic algorithms based on expert recommendations consider extreme thrombocytosis (>1000 or >1500 × 109/l) and previous bleeding as risk factors for haemorrhage; however, it is unknown whether and which other baseline features might enhance the haemorrhagic risk.

We retrospectively analysed a cohort of 565 consecutive patients, diagnosed at the Institute of Haematology/Oncology ‘L. and A. Seràgnoli’, Bologna, between 1977 and 2010. ET was diagnosed according to the Polycythaemia Vera Study Group (Murphy et al, 1997) or the World Health Organization (WHO) criteria (Vardiman et al, 2002).

Main clinical/laboratory data were collected at presentation in all patients. Severe haemorrhage was defined as a life-threatening bleeding or an overt haemorrhage requiring one or more units of red cells/day (grade III-IV on the WHO scale). Response to cytoreductive treatment and qualitative evaluation of JAK2 V617F mutational status were assessed as previously described (Palandri et al, 2009).

The median follow-up of the patients is 7·8 years (range, 0·5–32·5). Median age was 64 years (range, 16–95); 61% were women. Mean platelet count was 843 × 109/l [median ± standard deviation (SD):763 ± 319 × 109/l]; 114 patients (20%) had a platelet count >1000 × 109/l. Mean leucocyte count was 9·5 × 109/l (8·9 ± 3·1 × 109/l); 143 patients (25%) had a leucocyte count >11 × 109/l. Mean haemoglobin concentration was 140 g/l (141 ± 16 g/l). JAK2 V617F mutation was determined in 360 patients, 30/62 (48%) patients with haemorrhages and 330/503 (66%) patients without haemorrhages, and was present in 55% and 65% of the patients, respectively (P = 0·31). Three patients were homozygous. The treatment strategy changed over the years. Generally, antiplatelet drugs were used in high-risk patients and/or with cardiovascular risk factors and/or symptoms due to microvessel disturbance. Cytotoxic therapy was reserved for high-risk patients and/or persistent microvessel disturbances on antiplatelet therapy. Overall, 484 patients (85·6%) received a cytoreductive treatment and 379 (78%) achieved at least a partial response. Five hundred and thirty-seven patients (95%) received an antiplatelet drug (acetylsalicylic acid, 100 mg daily, n = 395, 73·5%; ticlopididin, n = 78, 14·5%; indobuphene, n = 64, 12%). The time on antiplatelet agents varied from 0·5–380 months (median 65). There was no difference in terms of type and duration of antiplatelet/cytoreductive treatments between the two cohorts.

Over the 4433 person-years of follow-up time, 62 patients (11%) had one or more haemorrhages, for a total of 80 haemorrhagic events, resulting in a cumulative risk of haemorrhage of 14% at 10 years. The incidence rates of total and of severe bleeds were 13·9 and 5 events per 1000 person-years, respectively. Twenty-six major bleeds occurred in 24 patients (4%) at ET diagnosis (4) or during follow-up (22), after a median time from diagnosis of 29 months (range, 0–260). The sites of severe haemorrhages were: melena (nine), haematemesis (one), rectal (three), haematuria (three), metrorrhagia (three), articular (two), intracranial (three), retinal (one) and renal (one). At the time of bleeding, the median platelet count was 560 × 109/l (range, 122–1322) and median leucocyte count was 7·3 × 109/l (range, 4·8–12·7) In 20 cases (77%), patients were receiving antiplatelet/anticoagulant agents. In seven patients, a major thrombotic event was observed at the same time or after the haemorrhage. In one case, bleeding was the ultimate cause of death. The remaining 54 haemorrhages were minor. The haemorrhage-free survival was 89·2% at 20 years.

By multivariate (Cox) analysis, splenomegaly, platelet count >1000 × 109/l and leucocyte count >11 × 109/l retained prognostic significance (Table I). Taking these three parameters into account, it was possible to stratify patients in three subgroups characterised by an increasing risk for haemorrhage (Fig 1).

Table I.   Cox proportional hazards regression model analysis of risk factors associated with total haemorrhagic events. By univariate (log-rank) analysis, baseline splenomegaly (P < 0·0001), positive history of haemorrhage (P = 0·005), platelet count higher than the median value (P = 0·003) and >1000 × 109/l (P < 0·0001) and leucocyte count >11 × 109/l (P < 0·0001) significantly correlated with total haemorrhagic events.
FactorTotal haemorrhagic events
  1. HR, hazard ratio; 95% CI, 95% confidence interval.

  2. *Age ≥60 years and/or previous thrombosis.

Sex (male versus female)0·70·4–1·40·33
High risk for thrombosis*1·60·9–2·80·12
Previous haemorrhage1·80·7–4·70·25
Median haemoglobin, g/l0·70·3–1·40·32
Median haematocrit, %1·10·5–2·20·88
Platelet count >1000 × 109/l2·31·3–3·70·003
Median platelet count, ×109/l1·80·6–5·50·28
Leucocyte count >11 × 109/l1·91·1–3·20·015
Median leucocyte count, ×109/l0·90·5–1·90·89
Antiplatelet therapy1·10·3–2·30·13
Figure 1.

 Cumulative risk of haemorrhage. Patients were stratified according to the presence of the three risk factors for haemorrhage (splenomegaly, leucocyte count >11 × 109/l, platelet count >1000 × 109/l). Line A: high risk (66 patients with two or three risk factors; Line B: intermediate risk (160 patients with one risk factor); Line C: low risk (339 patients, no risk factors). Total haemorrhagic events were considered. The cumulative risk for haemorrhage at 10 years was 8·8% in the low-risk, 16·4% in the intermediate risk (P = 0·09) and 45·8% in the high risk group (P < 0·001).

Taken together, these three features are typical of a subgroup of ‘PVSG-diagnosed ET’ now recognized as ‘pre-fibrotic myelofibrosis’ (PMF-0) (Barbui et al, 2010). The possibility that some of the reported cases could be PMF-0 cannot be completely excluded; however, the three features were concomitantly present in a very small minority of the cases (six patients, 1%). The correlation between spleen enlargement and haemorrhage might relate to secondary portal hypertension, as six of the 12 patients who presented with gastrointestinal bleedings also had splenomegaly; moreover, splenomegaly was the only feature that retained prognostic relevance when risk factors for severe haemorrhages were separately investigated (Hazard Ratio 4·65, 95% confidence interval 1·8–4·6, P = 0·001). Extreme thrombocytosis is already recognized to be responsible for haemorrhagic complications. This paradox is attributed to an acquired von Willebrand syndrome and to the abnormalities typical of neoplastic ET platelets (Michiels et al, 2006). Leucocytosis has been found to induce platelet aggregation, to trigger the expression of tissue factors and to activate and damage endothelial cells (Falanga et al, 2005). It is possible that the endothelial dysfunction could also be a contributory cause of bleeding. Analogously, JAK2 V617F mutation was first proposed as a risk factor for thrombosis, and, more recently, the homozygous state of this mutation was also found to correlate with a higher haemorrhagic risk (Vannucchi et al, 2007). In the present study, the presence of the JAK2 mutation was not associated with an increased rate of bleeding; however, only very few of our patients were homozygous.

Although it is possible that antiplatelet drugs might have increased the haemorrhagic risk, the influence of these agents on bleeding was not evaluable in our study, because only a small fraction of patients (n = 25, 4·5%) had never received an antiplatelet agent. Moreover, the type and duration of antiplatelet drug administration were also very variable during the observation period. With the limitations due to the retrospective nature of the analysis, this study showed that patients with marked thrombocytosis, leucocytosis and spleen enlargement are at higher risk for bleeding. In these patients, the use of antiplatelet drugs should be carefully considered, balancing the haemorrhagic and thrombotic risks, which may be increased by the same features.


FP, NV, NP: Concept and design, analysis and interpretation of the data. SP: Collection of the data. FP, NV, NP, LC, DS: Critical writing and revising the intellectual content. EO: Molecular biology analysis. NV, MB: Final approval of the version to be published.

Conflicts of interest

The authors declare no competing financial interests.


The study was supported by Fondazione del Monte di Bologna e Ravenna, by European LeukemiaNet funds and by BolognAIL grants.