Prediction of thrombosis risk in patients with paroxysmal nocturnal hemoglobinuria

Paroxysmal nocturnal hemoglobinuria (PNH) is an acquired hemolytic disease with thrombosis as a major complication. The mechanism of thrombosis and related risk factors in PNH patients are still not well characterized. We retrospectively enrolled 99 patients with newly diagnosed PNH at our institute from 2011 to 2016. According to binary logistic regression model analysis, we first identified four baseline clinical risk factors which may be associated with incidence of thrombosis in the PNH cohort, including PNH clone sizes (fluorescent aerolysin of neutrophil) <= 80 (OR 1.056, 95%CI 1.016-1.097, P = 0.005), hemoglobin <= 75 g/L (OR 4.202, 95%CI 0.984-17.954, P = 0.053), platelet > 100 x 10(9)/L (OR 6.547, 95%CI 1.490-28.767, P = 0.013) and rs495828 = G (OR 5.243, 95%CI 1.314-20.916, P = 0.019). These independent risk factors were combined together to develop a risk model to evaluate thrombosis risk (AUC = 0.756, 95%CI 0.607-0.905, P < 0.001). Our risk model revealed a higher cumulative incidence of thrombosis and an earlier thrombosis events in PNH patients with high risk (risk score >= 23) compared with those with low risk (risk score < 23, P < 0.001 and P = 0.043, respectively). Although with some limitations, we set up a prediction model for thrombosis risk in patients with PNH for the first time, but it needed to be verified in a prospective study with larger patients and longer follow-up time in the future.