Multivariable analyses for achievement of complete remission (CR) and overall survival (OS)
| End point/variables in final models . | OR . | HR . | 95% CI . | P . |
|---|---|---|---|---|
| CR* | ||||
| BAALC expression, high vs low | 0.27 | 0.08-0.93 | .04 | |
| ERG expression, high vs low | 0.20 | 0.06-0.65 | .007 | |
| Age, 10-year increase | 0.47 | 0.25-0.91 | .02 | |
| OS† | ||||
| BAALC expression, high vs low | 1.84 | 1.04-3.28 | .04 | |
| FLT3-ITD, positive vs negative | 2.29 | 1.30-4.05 | .004 | |
| NPM1, mutated vs wild-type | 0.47 | 0.24-0.90 | .02 | |
| CEBPA, mutated vs wild-type | 0.42 | 0.20-0.89 | .02 | |
| log2(WBC)‡ continuous, 2-fold increase | 2.14 | 1.29-3.53 | .01 | |
| End point/variables in final models . | OR . | HR . | 95% CI . | P . |
|---|---|---|---|---|
| CR* | ||||
| BAALC expression, high vs low | 0.27 | 0.08-0.93 | .04 | |
| ERG expression, high vs low | 0.20 | 0.06-0.65 | .007 | |
| Age, 10-year increase | 0.47 | 0.25-0.91 | .02 | |
| OS† | ||||
| BAALC expression, high vs low | 1.84 | 1.04-3.28 | .04 | |
| FLT3-ITD, positive vs negative | 2.29 | 1.30-4.05 | .004 | |
| NPM1, mutated vs wild-type | 0.47 | 0.24-0.90 | .02 | |
| CEBPA, mutated vs wild-type | 0.42 | 0.20-0.89 | .02 | |
| log2(WBC)‡ continuous, 2-fold increase | 2.14 | 1.29-3.53 | .01 | |
Odds ratios less than 1 correspond to lower odds of achieving CR for higher values of continuous variables and the first category listed for dichotomous variables. Hazard ratios greater than (less than) 1 correspond to an increased (decreased) risk of death for higher values of continuous variables and the first category listed for dichotomous variables.
OR indicates odds ratio for achieving a complete remission; HR, hazard ratio of death for overall survival; CI, confidence interval; CR, complete remission; and OS, overall survival.
Variables, other than BAALC, considered for model inclusion were age, sex, race, hemoglobin level, platelet count, log2(WBC), percentages of blood and BM blasts, type of induction treatment (presence or absence of PSC-833), FLT3-ITD, NPM1 mutation status, CEBPA mutation status, MLL-PTD, FLT3-TKD, and ERG expression. FLT3-ITD, CEBPA mutational status, ERG expression, log2(WBC), age, and sex had a P < .20 in univariable analyses and were fit with BAALC expression in a full model. The final model is shown.
Variables considered for model inclusion were the same as those considered for the CR model with the exception of type of induction therapy. FLT3-ITD, NPM1 mutation status, ERG expression, hemoglobin level, platelet count, log2(WBC), percentage of blood blasts, and race had a P < .20 in univariable analyses and were fit with BAALC expression in a full model. CEBPA mutation status was added as a significant variable with BAALC, FLT3-ITD, NPM1, and log2(WBC) already in the model. The final model is shown.
log2(WBC) did not meet the proportional hazards assumption. The P value corresponds to the Wald statistic of a 2 degree of freedom test evaluating whether the coefficients for log2(WBC) and an artificial time-dependent covariate were equal to zero. The risk of an event is larger for those with higher WBC soon after going on study, but this risk decreases over time. The hazard ratio presented is for a 2-fold increase in WBC, evaluated at 3 months.