A higher STAT3β/α mRNA ratio correlates with clinical prognosis and survival in AML patients. (A) STAT3β/α mRNA ratio (normalized to β-ACTIN) in AML patients (n = 94) and HSPCs from healthy controls (n = 8), and STAT3β/α mRNA ratio in AML patients with a favorable (n = 23), intermediate (n = 22), or adverse (n = 49) prognosis. Data were compared using the Student t test and Kruskal-Wallis test. (B) Kaplan-Meier plot showing the survival of patients with a STAT3β/α mRNA ratio in the highest, medium, or lowest tertile (n = 48). (C) Estimated survival functions resulting from multivariable Cox regression. In addition to the logSTAT3β/α mRNA ratio in its continuous form, the model includes and thus adjusts for patient age, blast count, and cytogenetic risk category at diagnosis. For illustrative purposes, curves were plotted for the maximal, median, and minimal observed STAT3β/α mRNA ratios. The model P value refers to the hypothesis that all variables in the model collectively predict survival; the logSTAT3β/α mRNA P value tests whether the ratio predicts survival independent of other variables. Of note, the estimated survival changes continuously with the STAT3β/α mRNA ratio. Therefore, the survival can be plotted for any STAT3β/α mRNA ratio value. To visualize the magnitude of survival difference between the maximally observed (0.034) and the minimally observed (0.002), we plotted survival function for these values and for the median value as reference. Modeling the effect of STAT3β/α mRNA ratio in its continuous form allowed avoidance of arbitrary groups. (D) Change of estimated HR (thick line) with 95% confidence interval (CI; dashed lines) associated with different STAT3β/α mRNA ratio levels. The median STAT3β/α mRNA ratio was chosen as reference, not affecting the statistical analysis. The curve was plotted for a range of STAT3β/α mRNA ratios spanning from the minimum to maximum observed ratio. Estimated HRs are given for the upper and lower ends of the curve. *P < .05, ***P < .001.