Eighty percent of acute myelogenous leukemias (AML) occur beyond the age of 60, 35% occur beyond the age of 75, and 10% beyond the age of 85. Between 1975 and 2001, the 5-year survival of younger AML patients has more than doubled. Yet the survival of patients above the age of 65 remains dismal, with no progress over that period. Even pending the emergence of new strategies to target the biological challenges of AML in the elderly, significant progress could be made by optimizing the choices of treatment strategies. For example, in the Swedish population, AML patients aged 70-79 years who live in an area where a higher proportion are treated with intensive approach have a 2-year survival of 20% instead of 7%[1]. Also, for poor risk patients, new alternate strategies are emerging, such as hypomethylating agents, that might offer better outcomes than intensive induction chemotherapy. Therefore, we hypothesize that decision models which allow comparing treatment options more objectively might result in improved treatment decisions and outcomes for this population of patients who are underrepresented in clinical trials. Such an approach has been successful in breast cancer, with programs such as Adjuvant!Online. Aims: In this project we plan to design decision models to compare effectiveness and early risks of different treatment approaches for AML in patients older than age 70, and to test them on a sample of patients from the Moffitt Total Cancer Care/Malignant hematology database. Approach: We will first conduct a systematic review of the literature for articles focusing on the treatment of AML i the elderly, and conduct meta-analyses of the results. After that, we will develop decision models using TreeAge Pro. This software creates micro simulation models using Markov processes to model the outcomes of AML and its treatment over time. We will focus on the following outcomes: 1-year survival, 2-year survival, 1- month mortality, and complete response rates. The approaches compared will be: intensive induction chemotherapy, low-dose chemotherapy, hypomethylating agents, and supportive care. We will control for the impact of age, cytogenetics, functional status and comorbidity level. Once the models are developed, we will test them in Moffitt's Total Cancer Care/malignanthematology database, which is a large prospective database that will provide about 900 AML patients above the age of 70 for this study. We hypothesize that our models will correlate well with the outcome of patients in the database and accurately order the benefits of the treatment choices for various prognostic subgroups. Significance: If successful, this project will provide the first available decision mode for evaluating the outcomes of AML in patients aged 70 and above. Such a model could then be tested as a clinical decision aid, as a tool for comparative effectiveness, quality of life, and cot-effectiveness studies and might contribute in the longer term to improved outcomes for this underserved population.