The use of age for selection of treatment protocols has been used as a surrogate for other factors that have been suggested to influence the health status of patients and likelihood of death after intensive chemotherapy, including neutrophil count and levels of bilirubin, albumin, hemoglobin and fibrinogen. To determine if a more accurate prediction of early death could be obtained by the inclusion of specific health-related factors, Drs. Walter and Estey and colleagues from the Clinical Research Division as well as the M.D. Anderson Cancer Center examined the correlation of early death with 17 different health factors among AML patients, and created predictive models using a subset of these factors.
Analysis of 3,365 newly diagnosed AML patients receiving intensive chemotherapy demonstrated the highest risk of early death among all age groups occurred within the first 28 days of treatment, which was used to empirically define TRM. The authors then modeled the effects of individual and combined health parameters on death within this period, wherein a value of 1.0 denotes a perfect prediction and a value of 0.5 is the rate of chance. Performance status (PS), a measure of patients’ general well-being and activity level, was the single most successful parameter to predict TRM, with a modeled value of 0.75. Some factors such as age, platelet and white blood cell counts (WBC), and albumin levels were individually less predictive, while several other factors had minimal ability to predict TRM. A simplified model that combined eight factors was highly able to predict TRM with a modeled value of 0.82. Factors included performance status, age, platelet count, serum albumin and creatine level, if the event was a secondary AML, WBC count and peripheral blast percentage. A simplified three-factor model was further developed for clinical use, consisting of age, PS and platelet count; however, this minimal model was much less accurate in predicting early death, emphasizing the need to integrate information from several factors for best results. Together these results demonstrate the superiority of using multiple health-related factors for accurate prediction of TRM risk.
RB Walter, M Othus, G Borthakur, F Ravandi, JE Cortes, SA Pierce, FR Appelbaum, HA Kantarjian, EH Estey. 2011. Prediction of early death after induction therapy for newly diagnosed acute myeloid leukemia with pretreatment risk scores: a novel paradigm for treatment assignment. Journal of Clinical Oncology. 29(33): 4417-4423.