Patients with AML can be separated into 2 groups: with constitutive activation of PI3K (PI3K ⁺ group) and without PI3K activation (PI3K ⁻ group). (A) Blast cells from patients 1 and 2 were analyzed for PI3K/Akt activation at the indicated times, after starvation in serum- and cytokine-free medium, to determine constitutive and sustained activation of this signaling pathway. Proteins from 10⁶ cells were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and analyzed by Western blot using antibodies directed against p-Akt (Ser 473) and pFOXO3A (Thr 32). To ensure equal loading, blots were reprobed with anti-actin or anti-Akt antibodies. LY294002 was used at 25 μM for 30 minutes to inhibit PI3K/Akt. The functionality of the PI3K/Akt pathway was tested in patient 2, after 15-minute stimulation with FLT3 ligand at 50 ng/mL. Vertical lines were drawn to indicate cuts made to put together nonadjacent regions from the same gel. (B) Kaplan-Meier curves for overall survival according to the PI3K status. Classical log-rank test was used to compare survival distribution between subgroups. We tested the proportional hazard hypothesis using a test proposed by Grambsch et al.¹⁴  To investigate for a mixed effect (early and late effect) of PI3K/Akt activation, we tested the null hypothesis of no short-term and no long-term effect, using the model proposed by Broët et al.¹⁵  This latter test allows for testing 3 null hypotheses: (1) no long-term and short-term effect; (2) no short-term effect; and (3) no long-term effect (whatever a short-term effect is). Here, the long-term effect represents the effect on the proportion of patients alive after 30 months. (C) Cumulative incidence curves for relapse according to the PI3K status. Relapse probabilities were obtained from crude cumulative incidence estimates, treating death before relapse and relapse as competing risk events. Crude cumulative incidence curves for relapse were compared between subgroups using the Gray test.¹⁶