Abstract
Introduction: Risk stratification in acute promyelocytic leukemia (APL) is based on the easily accessible Sanz-Score, which combines leukocyte and platelet counts at initial diagnosis. This score showed significant differences in relapse-free survival (RFS) of APL patients in various studies and is currently used to determine whether a patient can be treated with ATRA and ATO alone or needs additional chemotherapy. However, to make therapeutic decisions based on a risk stratification system derived from the endpoint RFS bears the drawback that relapses are rare in APL and most events are deaths in complete remission (CR), which can be therapy related (e.g. toxicity). The cumulative incidence of relapse (CIR) therefore seems to be a better parameter for decision making with regard to therapy intensity. In this study, we optimized a risk score combining data on gene expression of BAALC (brain and acute leukemia, cytoplasmic), ERG (ets’ related gene) and WT1 (Wilms’ tumor 1) to retrospectively predict the CIR of APL patients.
Methods: Data on BAALC, ERG and WT1 expression levels of 79 patients with newly diagnosed APL were obtained from bone marrow mononuclear cells using quantitative real-time RT-PCR in preceding studies. The following gene expression levels were identified as negative risk factors: BAALC expression ≥25th percentile (BAALChigh), ERG expression >75th percentile (ERGhigh) and WT1 expression ≤25th percentile or ≥75th percentile (WT1low or high). As ERGhigh was the only independent predictor for relapse in multivariate analysis with a hazard ratio (HR) of 11.6, its predictive weight was regarded superior, respectively . Cut-off analyses were performed to determine the optimal ERG expression level cut-off for risk of relapse. Accordingly, the new cut-off for high ERG expression was set at ≥62nd percentile (optimized ERGhigh: optERGhigh; Sensitivity: 1.0, Specificity: 0.71). A combined risk score was developed as follows: For the presence of one of the mentioned risk factors, one scoring point was assigned to a respective patient, i.e. a maximum of 3 points (one point for BAALChigh, optERGhigh and WT1low or high, respectively) and a minimum of 0 points (i.e. presenting with none of the aforementioned risk factors) could be allocated to one patient. Accordingly, patients were divided into two risk groups: 34 patients scored 0-1 points and 45 patients scored 2-3 points. CIR, overall survival (OS) and RFS were calculated using the Kaplan-Meier method and a log-rank test was used to compare differences between the two risk groups (p<0.05).
Results: Patients with 2-3 points had a CIR of 18% at 10 years of follow-up whereas none of the patients with 0-1 points suffered a relapse (CIR: 0%; p=0.02; Fig. 1). All relapses occurred between 8.4 months and 3.5 years after first CR. Moreover, OS and RFS also differed significantly between the two risk groups: OS was 53% for patients with 2-3 points vs. 85% for patients with 0-1 points (p=0.004); RFS was 49% vs. 93%, respectively (p<0.0001). In multivariate analysis the optimized combined risk score was the strongest independent risk factor for every endpoint.
Conclusion: The combination of expression levels of BAALC, ERG and WT1 into a risk score identified a group of patients at high risk for relapse which could benefit from close monitoring resulting possibly in an early intervention when molecular relapse is detected. On the other hand, it identified a low risk group with very good outcome and no APL-related events after patients had achieved first CR. A molecular risk score focusing on relapse risk might be a promising approach to guide therapeutic decisions in the future.
No relevant conflicts of interest to declare.
