Introduction: Bleeding is a main cause of morbidity and mortality in adult immune thrombocytopenia (ITP). Treatment is indicated in case of bleeding or of low platelet count. However, the threshold of platelet count associated with bleeding and therefore to initiation of treatment is debated. Most of guidelines recommend the threshold of <30 x 109/L. However, some authors recommend other thresholds, as low as 10 x 109/L. Moreover, the risk factors for bleeding are not well known in ITP. The aim of this study is to assess the risk of bleeding by platelet count and to identify these risk factors.

Methods: We studied all the patients included between June 2013 and December 2016 in the CARMEN (Cytopénies Auto-immunes : Registre Midi-PyréneEN) registry. This multicenter registry is aimed at the prospective follow-up of all incident ITP adults in the French Midi-Pyrénées region (South-West of France, 3 million inhabitants). Each investigator follows every patient newly diagnosed for ITP in routine visit or hospital stay and detailed information on patients' characteristics and management of ITP are recorded prospectively. We also included all adult patients newly diagnosed for ITP at the French National Referral Center for autoimmune cytopenias (Créteil) from November 2015 to December 2016 where data are prospectively recorded in the same database. ITP was defined by platelet count <100 x 109/L and exclusion of other causes of thrombocytopenia. We described the frequency of any bleeding, of mucosal bleeding and of severe bleeding (defined by macroscopic hematuria, gastrointestinal or central nervous system bleedings) at ITP onset by platelet count and by subgroups of interest: age groups (according to quartiles: ≤45, 45-64, 65-79 and ≥80 years), sex, arterial hypertension, history or current symptoms of gastroduodenal ulcer, Charlson's comorbidity score, exposure to non-steroidal anti-inflammatory drugs (NSAIDs), antiplatelet drugs, anticoagulant drugs and serotonin reuptake inhibitors (SRIs). Lastly, we assessed in the whole cohort the factors associated with any bleeding, mucosal bleeding and severe bleeding using multivariate logistic regression analysis.

Results: 302 newly diagnosed ITP adults were prospectively included in the study. Median age was 66 years, 49.7% were females, 46.8% had a Charlson's score ≥1, 3.7% had a history or current symptoms of gastroduodenal ulcer, 16.0% had secondary ITP, 19.5% were exposed to antiplatelet drugs, 7.6% to anticoagulant drugs, 7.6% to NSAIDs and 6.6 to SRIs at the time of diagnosis. Median platelet count was 18 x109/L. At diagnosis, 57.9% experienced any bleeding symptom, 30.1% mucosal bleeding, and 6.6% severe bleeding (including 4 central nervous system bleedings).

The rates of bleeding, mucosal bleeding and severe bleeding by platelet count is shown in Figure 1. The rate of any bleeding was >50% if platelet count was <20 x109/L. The rate of mucosal bleeding was >40% if platelet count was <10 x109/L. In contrast, a similar risk of severe bleeding was observed with all platelet count categories. Only slight differences were observed in other subgroups, except an increased frequency of any bleeding in case of exposure to NSAIDs, as well as an increased rate of severe bleeding in elderly and patients exposed to anticoagulant drugs and SRIs.

In multivariate analysis, the factors associated with any bleeding at ITP diagnosis were a low platelet count (<10 x109/L vs. >20 x109/L: OR, 46.3, 95% CI: 19.4-110.6; between 10 and 19 x109/L vs. >20 x109/L: OR, 5.1, 95% CI: 2.3-11.2), female sex (OR for males, 0.4, 95%CI: 0.2-0.8) and exposure to NSAIDs (OR, 4.4, 95 %CI: 1.1-18.7). Only a low platelet count was associated with mucosal bleeding (<10 x109/L vs. >20 x109/L: OR, 6.2, 95% CI: 3.4-11.6; between 10 and 19 x109/L vs. >20 x109/L: OR, 2.6, 95% CI: 1.1-6.1). Lastly, only exposure to anticoagulant drugs was associated with severe bleeding (OR, 4.5, 95% CI: 1.4-14.4). Analyses restricted to primary ITPs led to similar results.

Conclusion: Platelet counts <20 x109/L and <10 x109/L were thresholds with major increased risk for both any and mucosal bleedings. Platelet count, female sex and exposure to NSAIDs should be considered to assess the risk of any bleeding. In contrast, severe bleeding seemed not linked to platelet count but to individual factors, particularly exposure to anticoagulant.


No relevant conflicts of interest to declare.

Author notes


Asterisk with author names denotes non-ASH members.