Predictive Factors for Latency Period and a Prognostic Model for Survival in Patients with Therapy-Related AML.

Abstract 2589

Therapy-related acute myeloid leukemia (t-AML) is an increasingly recognized sequela in patients (pts) receiving chemotherapy or radiotherapy for a primary malignancy or autoimmune disease. Factors that adversely affect treatment response and survival in t-AML pts include poor cytogenetics, type of antecedent disorder (AD) and type of preceding therapy. The goal of this study was to design a comprehensive prognostic model integrating pt disease- and treatment-related characteristics to predict clinical outcome and to assess factors related to the latency period (LP) between the AD and t-AML diagnosis (dx).

We evaluated a retrospective cohort of newly diagnosed t-AML pts treated at Cleveland Clinic from 2001 to 2011. Data on age at initial dx of AD, type of AD, preceding treatment, type of chemotherapy, age at t-AML dx, gender, peripheral blood counts at t-AML dx, peripheral and marrow blasts, pathologic classification, metaphase cytogenetics (per CALGB/Alliance 8461 criteria), LP, complete remission (CR) and overall survival (OS) measured from t-AML dx were collated from our IRB approved AML database. Multivariable log-linear, logistic, and proportional hazards models with step-wise variable selection were used to identify independent predictors of each outcome.

Of 730 patients treated with cytarabine-based induction chemotherapy over the 10-year period, 77 had t-AML. Most (68%) were female; median age at dx of antecedent disorder was 56 years (range, 16–75); median age at t-AML dx was 61 yrs (range, 19–79); and median latency period to t-AML dx was 4.6 yrs (range, 0.5–38.4). Most (71%) had an antecedent solid tumor [breast cancer (44%), prostate (10%); colon (6%), other organ sites (8%); 23% had a prior hematologic malignancy [non-Hodgkin lymphoma (16%), Hodgkin lymphoma (4%), and leukemia (4%)]; and 5% had autoimmune diseases. Previous treatments for AD included radiation (26%), chemotherapy (30%), and chemotherapy and radiation (44%). Of 57 pts previously treated with chemotherapy, 68% received alkylating agents, 65% anthracyclines, 51% mitotic inhibitors (MI) and 30% all three drug classes. Cytogenetic risk distribution at t-AML dx was: favorable (19%), intermediate-risk (52%), and unfavorable (29%). Overall, 48 pts (62%) achieved a CR with induction chemotherapy and median OS was 9.6 months, with 30% surviving >2 years.

Independent prognostic factors of shorter LP were age at AD >55 (p=.001) and prior treatment with MI (p=.001). Median LP for pts aged <55 at AD with no prior treatment with MI was 7.6 yrs, compared to 4.6 yrs for pts <55 exposed to MI, 4.6 years for pts age >55 but no prior MI, and 2.0 years for pts >55 and prior treatment with MI. Age at t-AML (p=.001) was the only independent predictor of CR. Independent predictors for inferior OS were unfavorable cytogenetics (p=.002), antecedent hematologic or autoimmune disease (p=0.007) and platelet counts <25000/μL at the time of t-AML dx (p=0.02). A prognostic model based on these factors categorized t-AML pts into two risk groups based on previous diagnosis type, cytogenetics, and platelet count at t-AML dx (Table 1). This score-based risk stratification used a cutoff of 2 points to categorize pts as favorable or unfavorable. Pts with a favorable profile had an estimated median OS of 28.4 months compared to 5.0 months for pts with an unfavorable profile (p=.0003).

In conclusion, multicomponent prognostic models that integrate well-established disease or treatment-related covariates can be clinically helpful in risk stratifying t-AML pts undergoing induction therapy, identifying those who might benefit from more intensive interventions.