Table 3.
Multivariable Analyses for Outcome of Cytogenetically Normal Acute Myeloid Leukemia Patients
| Variable in Final Models | Hazard Ratio | 95% CI | P |
|---|---|---|---|
| All patients | |||
| Event-free survival* | |||
| CEBPA, mutated v wild type | 0.4 | 0.2 to 0.8 | .007 |
| WT1, mutated v wild type | 2.5 | 1.4 to 4.7 | .003 |
| FLT3-ITD, positive v negative | 2.1 | 1.3 to 3.4 | .002 |
| ERG expression, high v low | 1.6 | 1.0 to 2.6 | .04 |
| Age, each 10-year increase | 0.6 | 0.4 to 0.9 | .007 |
| Disease-free survival† | |||
| CEBPA, mutated v wild type | 0.4 | 0.2 to 0.8 | .014 |
| WT1, mutated v wild type | 3.3 | 1.7 to 6.6 | < .001 |
| FLT3-ITD, positive v negative | 1.9 | 1.1 to 3.4 | .006 |
| ERG expression, high v low | 2.7 | 1.2 to 6.1 | .04 |
| Overall survival‡ | |||
| CEBPA, mutated v wild type | 0.3 | 0.2 to 0.6 | < .001 |
| WT1, mutated v wild type | 3.4 | 2.3 to 7.0 | < .0001 |
| NPM1, mutated v wild type | 0.4 | 0.2 to 0.7 | .03 |
| FLT3-ITD, positive v negative | 2.1 | 1.2 to 3.7 | .004 |
| WBC, each 50-unit increase | 1.3 | 1.1 to 1.6 | .002 |
| Extramedullary involvement, no v yes | 2.0 | 1.2 to 3.2 | .01 |
| Molecular high-risk patients | |||
| Event-free survival§ | |||
| CEBPA, mutated v wild type | 0.3 | 0.2 to 0.6 | < .001 |
| WT1, mutated v wild type | 2.0 | 1.1 to 3.7 | .03 |
| Hemoglobin, each 1-unit increase | 0.9 | 0.9 to 1.0 | .04 |
| WBC, each 50-unit increase | 1.3 | 1.0 to 1.6 | .03 |
| Disease-free survival‖ | |||
| CEBPA, mutated v wild type | 0.4 | 0.2 to 0.7 | .004 |
| FLT3-ITD, positive v negative | 1.9 | 1.0 to 3.6 | .008 |
| Hemoglobin | 0.8 | 0.7 to 0.9 | .003 |
| Overall survival¶ | |||
| CEBPA, mutated v wild type | 0.4 | 0.2 to 0.8 | .009 |
| WT1, mutated v wild type | 2.8 | 1.5 to 5.3 | .002 |
| FLT3-ITD, positive v negative | 2.9 | 1.3 to 6.6 | .009 |
Hazard ratios greater than 1 indicate higher risk for an event for the first category listed for categorical variables. Hazard ratios less than 1 indicate lower risk for an event for the higher values of a continuous variable. Hazard ratios greater than 1 indicate higher risk for an event for the higher values of a continuous variable. Variables considered in the model were those significant at α = .20 from the univariable models. FLT3-ITD status and NPM1 mutation status were evaluated in all final models and included if determined to be confounded with the main analysis variable, CEBPA.
Abbreviation: FLT3-ITD, internal tandem duplication of the FLT3 gene.
Variables considered for model inclusion were CEBPA (mutated v wild type), FLT3-ITD (positive v negative), WT1 (mutated v wild type), ERG expression (high v low), BAALC expression (high v low), age, hemoglobin, platelets, WBC, and extramedullary involvement based on their significance from univariable analyses. On the basis of clinical importance, NPM1 (mutated v wild type) was tested in the final model and retained because of its confounding effect on CEBPA. Age did not meet the proportional hazards assumption and, therefore, was evaluated with an artificial time-dependent covariate in the model. The P corresponds to the Wald statistic of a 2-df test evaluating whether the coefficients for age and an artificial time-dependent covariate were equal to 0. The hazard ratio presented is for a 10-year increase in age, evaluated at 3 years.
Variables considered for model inclusion were CEBPA (mutated v wild type), FLT3-ITD (positive v negative), NPM1 (mutated v wild type), WT1 (mutated v wild type), ERG expression (low v high), hemoglobin, WBC, race (white v not white), and extramedullary involvement based on their significance from univariable analyses. NPM1 was retained in the final model along with a time-dependent covariate (because it did not meet the proportional hazards assumption), despite its borderline significance (P = .067), because of its confounding effect on CEBPA. FLT3-ITD did not meet the proportional hazards assumption. The P corresponds to the Wald statistic of a 2-df test test evaluating whether the coefficients for FLT3-ITD and an artificial time-dependent covariate were equal to 0. The hazard ratio presented is for FLT3-ITD–positive v –negative status, evaluated at 8 months from the date of complete remission. ERG did not meet the proportional hazards assumption. The P corresponds to the Wald statistic of a 2-df test evaluating whether the coefficients for ERG and an artificial time-dependent covariate were equal to 0. The hazard ratio provided was evaluated at 1.5 years after achieving complete remission.
Variables considered for model inclusion were CEBPA (mutated v wild type), FLT3-ITD (positive v negative), NPM1 (mutated v wild type), WT1 (mutated v wild type), ERG expression (high v low), BAALC expression (high v low), age, hemoglobin, platelets, WBC, percentage of blood blasts, and extramedullary involvement based on their significance from univariable analyses. FLT3-ITD did not meet the proportional hazards assumption and, therefore, was evaluated with an artificial time-dependent covariate in the model. The P corresponds to the Wald statistic of a 2-df test evaluating whether the coefficients for FLT3-ITD and an artificial time-dependent covariate were equal to 0. The hazard ratio presented is for FLT3-ITD–positive v –negative status, evaluated at 8 months on study. NPM1 did not meet the proportional hazards assumption and, therefore, was evaluated with an artificial time-dependent covariate in the model. The P corresponds to the Wald statistic of a 2-df test evaluating whether the coefficients for NPM1 and an artificial time-dependent covariate were equal to 0. The hazard ratio presented is for NPM1 mutated v wild type, evaluated at 1.5 years on study.
Variables considered for model inclusion were CEBPA (mutated v wild type), FLT3-ITD (positive v negative), WT1 (mutated v wild type), ERG expression (high v low), hemoglobin, and WBC based on their significance from univariable analyses. On the basis of clinical importance, NPM1 (mutated v wild type) and FLT3-ITD (positive v negative) were tested in the final model but were not retained because they were not confounded with CEBPA and were not significant in the final model.
Variables considered for model inclusion were CEBPA (mutated v wild type), FLT3-ITD (positive v negative), FLT3 TKD (positive v negative), WT1 (mutated v wild type), hemoglobin, WBC, and race (white v not white) based on their significance from univariable analyses. On the basis of clinical importance, NPM1 (mutated v wild type) was tested in the final model but was not retained because it was not confounded with CEBPA and was not significant in the final model. FLT3-ITD did not meet the proportional hazards assumption and, therefore, was evaluated with an artificial time-dependent covariate in the model. The P corresponds to the Wald statistic of a 2-df test evaluating whether the coefficients for FLT3-ITD and an artificial time-dependent covariate were equal to 0. The hazard ratio presented is for FLT3-ITD–positive v –negative status, evaluated at 8 months from the date of complete remission.
Variables considered for model inclusion were CEBPA (mutated v wild type), FLT3-ITD (positive v negative), WT1 (mutated v wild type), ERG expression (high vuu low), hemoglobin, platelets, and WBC based on their significance from univariable analyses. On the basis of clinical importance, NPM1 (mutated v wild type) was tested in the final model but was not retained because it was not confounded with CEBPA and was not significant in the final model. FLT3-ITD did not meet the proportional hazards assumption and, therefore, was evaluated with an artificial time-dependent covariate in the model. The P corresponds to the Wald statistic of a 2-df test evaluating whether the coefficients for FLT3-ITD and an artificial time-dependent covariate were equal to 0. The hazard ratio presented is for FLT3-ITD–positive v–negative status, evaluated at 9 months on study.