Predictors of Infection and Bleeding in Patients with Chemotherapy Induced Thrombocytopenia. Session Type: Poster Session 215-II

Nancy M. Heddle, Richard J. Cook, Chris Sigouin, Kathryn E. Webert, Paolo Rebulla

Author address: 

Department of Medicine, McMaster University, Hamilton, ON, Canada; Statistics and Actuarial Science, University of Waterloo, Waterloo, ON, Canada; Centro Trasfusionale e di Immunologia dei Trapianti, IRCCS, Ospedale maggiore, Milan, Italy


Background: Large amounts of data are frequently collected while conducting a randomized controlled trial (RCT); however, typically these data are not optimally explored to develop and investigate hypotheses regarding the pathophysiology of disease. Data from a previously conducted platelet transfusion trial involving patients with acute leukemia, were analysed to explore factors related to bleeding and infection. Specific goals were to identify risk factors for bleeding and infection; to explore relationships between different grades of bleeding (WHO classification); to determine if patients with cutaneous bleeding were more susceptible to infection; and to determine the effect of hemoglobin levels on the onset of bleeding. Methods: Multivariate Cox and Anderson-Gill regression models were fit using backwards elimination to identify significant independently predictive factors for the events of interest. Potential risk factors for bleeding and infection included in the models were: administration of antifungal medication; documented evidence of infection; body temperature; antibiotic use; antifibrinolytic use; antiviral medication; chemotherapy; hemoglobin; platelet and red cells transfusions; steroid use; previous evidence of Grade 1 bleeding, and platelet count. Results: The risk factors for bleeding varied depending on the WHO bleeding classifications considered. The Table below summarizes the significant risk factors identified. WHO Grade 1 or 2 bleeding was a predictor of WHO bleeding Grades 3 or 4 (RR 3.05; p 0.02); Grade 1 bleeding was predictive of WHO bleeding Grades 2, 3 and 4 (RR 2.55; p 0.02); Grade 2 bleeding did not predict bleeding of Grades 3 and 4 (RR 2.8; p 0.31). There was no evidence that cutaneous symptoms of bleeding (Grades 1 and 2) increased the risk of infection (RR 1.48; p 0.25). A 10 gm/L increase in hemoglobin was associated with a lower risk of Grades 2, 3 and 4 bleeding (RR 0.78; p = 0.048). Patients receiving prophylactic platelet transfusions at a trigger of 20 x 109/L had a higher incidence of infection than those transfused at a trigger of 10 x 109/L (RR 1.86; p = 0.03). Conclusion: The data suggests that: higher hemoglobin values may protect against bleeding in thrombocytopenic patients; cutaneous bleeding does not predispose patients treated for acute leukemia to infection; Grade 1 bleeding may be an important clinical observation to identify patients that have an increased risk of more serious types of bleeds; and, patients transfused prophylactic platelets at a higher trigger may have a higher rate of infection than patients transfused at a lower trigger value. Information from this modeling analyses can be used for hypothesis generation and to support previous hypothesis that have been proposed. Table 1: Summary of Risk Factors for WHO Bleeding and Infection Significant Risk Factors Relative Rate 95% CI P Value EVENT: Bleeding (Grade 2,3 & 4) Body Temperature 1.87 1.40 to 2.49

abstract No: 


Full conference title: 

American Society of Hematology 45th Annual Meeting
    • ASH 45th (2003)