Most multicenter randomized AML studies use randomizations of patients early or later in complete remission which necessarily occur with exclusions and positive selection of patients included. Since the exclusion criterias are regimen related and do not follow common standards, incomparabilities between treatment results across different studies are produced by these late randomizations. In order to overcome this problem, we here propose a cooperation of studies on the basis of a general up-front randomization with attribution of 10% patients from each study to a common standard arm. A validation of complete treatment strategies according to intent-to-treat against the standard arm and thus also across the studies is provided by this inter-group model which may contribute to accelerate the therapeutic progress in AML.