OBJECTIVES: Treatment of patients with acute myeloid leukemia (AML) is based upon stratification into risk (prognosis) groups. New diagnostic methods are in development to improve this stratification. Economic evaluations of these methods require knowledge of what happens when the current stratification methods are used. We examined whether the literature can provide valid estimates for the outcomes of complete remission rates for patients with primary AML aged 16-60 years. METHODS: A systematic literature review was performed using Pubmed and Embase. Inclusion criteria were: ≥100 AML patients and detailed outcomes per risk group (favorable, intermediate, unfavorable). Excluded were: Phase I/II studies, studies not containing any patients aged 16-60 years or with primary AML. We compared various study characteristics such as patient population, treatment given, risk group definitions and complete remission (CR) rates as outcome. A chi-square test for homogeneity of CR rates was performed. RESULTS: Twelve studies fulfilled the eligibility criteria. Great variation was found between study populations. While treatment varied between the studies, all patients received cytarabine and an anthracycline. Definitions of risk groups varied greatly except for the favorable risk group. There was no homogeneity in overall CR rate (range: 52-85% p<0.001). After excluding studies with many patients other than the target population, heterogeneity between the remaining studies decreased (N=7, p=0.083). CR rates were homogeneous in the favorable group (p=0.223), but heterogeneous in the intermediate and unfavorable groups (p=0.044 and p=0.096 respectively). CONCLUSIONS: Differences in patient population and risk group definitions lead to heterogeneity in CR rates. Only a small number of studies will provide valid estimates of the CR rates in patients with primary AML aged 16-60 years. However, this restriction may reduce the reliability of the estimates, because the estimates will be based on fewer patients. This will thereby increase the uncertainty around the ICER of new methods.