Acute Myeloid Leukemia (AML) in the Elderly. To Treat or Not To Treat: A Decision Analysis.

Abstract

AML is a disease frequently observed in elderly patients (pts) Treating this population with intensive chemotherapy or palliative treatment is a controversial decision. We analysed the available evidence with a decision model tree considering prognosis and comorbidities.Method: A decision tree was designed to model significant events and prognostic variables that have an impact in survival at 5 years. The probabilities of this items were obtained from a literature review and for those items without literature support, assumptions were made by an expert opinion based on investigator's consensus. Utility of each strategy was defined in an ordinal scale and the preference for each was calculated as the expected utility (utility x probability of the result)Results: Probabilities were: unfavourable prognosis 95%, incidence of comorbidities 48%, death in induction without comorbidities 25%, response to induction in unfavourable prognosis 42%, response to induction in favourable prognosis 63%, adequate performance status after induction without comorbidities 43%, probability of finishing consolidation 64%, survival at 5 years in pts with unfavourable AML 11%, survival at 5 ys in pts with favourable AML 27%. The effect of comorbidities on survival to induction and adequate PS post induction was estimated as 50% higher than in pts without comorbidities. The survival at 5 ys in pts with no response or those who did not complete induction was considered equal to palliative treatment, and S at 5 ys in pts with comorbidities who finished consolidation 30% less than in those with no comorbidities. Utility for “not to treat” was 11, and for “to treat” was 6. The preference for “not to treat” was maintained for any modification of the survival to the induction chemotherapy independently of comorbidities. In sensitivity analysis, the only modification of the observed preference was that “To treat” strategy would be preferable only if the good PS after induction chemotherapy would be observed in 65% of pts or more. In conclusion, the preferred strategy as suggested by the model is “not to treat”, however many relevant questions are still unanswered, reinforcing the need to include these pts in clinical trials for better evidence to base the decisions.