Effectivity of a modified Sanz risk model for early death prediction in patients with newly diagnosed acute promyelocytic leukemia.

Early death is the main obstacle for the cure of patients with acute promyelocytic leukemia (APL). We have analyzed risk factors of early death from 526 consecutive newly diagnosed APL patients between 2004 and 2016. The overall incidence of early death was 7.2% (38/526). The peak hazard of early death occurred in the first 0-3 days. Multivariate logistic analysis demonstrated white blood cell (WBC) counts [odds ratio (OR) = 1.039; 95% confidence interval (CI): 1.024-1.055; P < 0.001], age (OR = 1.061; 95% CI: 1.025-1.099; P = 0.001) and platelet counts (OR = 0.971; 95% CI: 0.944-0.999; P = 0.038) were independent risk factors for early death. Furthermore, receiver-operator characteristic (ROC) curve analyses revealed a simple WBC/platelet ratio was significantly more accurate in predicting early death [areas under the ROC curve (AUC) = 0.842, 95% CI: 0.807-0.872) than WBC counts (AUC = 0.793; 95% CI: 0.756-0.827) or Sanz score (AUC = 0.746; 95% CI: 0.706-0.783). We stratified APL patients into four risk subgroups: low risk (WBC ≤ 10 × 109/L, platelet >40 × 109/L), intermediate risk (WBC/platelet <0.2 and age ≤ 60, not in low risk), high risk (WBC/platelet ≥0.2 or age > 60, not in low and ultra-high risk) and ultra-high risk (WBC > 50 × 109/L), the early death rates were 0, 0.6, 12.8, and 41.2%, respectively. In conclusion, we proposed a modified Sanz risk model as a useful predictor of early death risk in patients with APL.