Acute myeloid leukemia (AML) remains difficult to treat: despite multiagent chemotherapy, allogeneic hematopoietic cell transplantation, and several newly approved agents, many patients will not be alive and in remission 3 years after diagnosis. However, with more agents available there are more options and a corresponding need to choose among them. Doing so is complicated by the molecular diversity of AML and the older age of many patients, predisposing them to both treatment-related mortality and, more commonly, resistance to treatment. There is no shortage of scoring systems to identify patients at high risk of early death or treatment resistance after conventional AML induction chemotherapy. As we point out here, their accuracy is limited. Furthermore, without periodic recalibration to account for new therapies and changes in supportive care, the accuracy of any prediction model will decrease over time. The limitations we describe here are important for clinicians to be aware of.

急性髓细胞性白血病（AML）仍然难以治疗：尽管进行了多药物化疗，同种异体造血细胞移植和几种新批准的药物，许多患者在诊断后3年仍无法存活并缓解。但是，可用的代理越多，选项就越多，并且需要在它们之间进行选择。AML的分子多样性和许多患者的年龄增加使这种情况变得复杂，这使他们容易面临与治疗相关的死亡率以及更常见的对治疗的耐药性。在常规AML诱导化疗后，不缺乏评分系统来识别处于早期死亡或治疗耐药性高风险的患者。正如我们在这里指出的那样，它们的准确性是有限的。此外，如果不进行定期重新校准以考虑新疗法和支持治疗的变化，任何预测模型的准确性都会随着时间而降低。我们在此描述的局限性对于临床医生来说很重要。