Parameter* Nonresponders to decitabine or treatment of choice Responders to decitabine or treatment of choice(CR+CRi+PR) P value
Median overall survival, months 4.3 17.4 <0.0001‡
Intravenous antibiotic use (1 per patient)
Number of patients 383 102  
Yes, n (%) 38 (9.9) 3 (2.9) 0.024§
Platelet transfusions (episodes per patient)
Number of patients 373 102  
Mean (SD) 3.03 (3.82) 1.19 (2.35) P1<0.0001; P2=0.796‖
Median (range) 1.8 (0–21) 0.3 (0–14)  
Number of transfusions/number of patient treatment cycles (proportion) 3231 / 1492 (2.17) 1025 / 1319 (0.78)  
Red blood cell transfusions (episodes per patient)
Number of patients 373 102  
Mean (SD) 2.93 (2.54) 1.42 (1.47) P1<0.0001 P2=0.872‖
Median (range) 2.3 (0–16) 1.1 (0–10.5)  
Number of transfusions/number of patient cycles (proportion) 3642 / 1492 (2.44) 1395 / 1319 (1.06)  
Dose modifications based on dose reductions and delays (1 per patient)
Number of patient cycles 1492 1319  
Number of patients with dose modification(s), n (%) 115 (7.7%) 63 (4.8%) 0.016¶
Table 2: Relationship between response to decitabine or treatment of choice and various potential indicators of treatment efficacy and safety. *All patients who had both response and parameter data available. † Either supportive care or cytarabine. ‡P values for overall survival of responders vs. non-responders. §2proportion Z-test to compare proportions between responders and nonresponders. ‖Based on zero-inflated Poisson regression model with responder (yes/no) as explanatory variable and log number of treatment cycles as offset. P1 is the P value based on the Poisson (non-“Certain Zero”) component and P2 is the P value based on the binomial component. ¶ Based on logistic regression with number of treatment cycles as the number of trials and responder (yes/no) as explanatory factor. CR, complete remission; CRi, CR with incomplete blood count recovery; NS, not significant; PR, partial remission; RBC, red blood cell; and SD, standard deviation.