Ransohoff David F
University of North Carolina at Chapel Hill, 27599-7360, USA.
Recent Results Cancer Res. 2009;181:55-9. doi: 10.1007/978-3-540-69297-3_6.
In many "omics" fields, extraordinary promises have been made about the ability of biomarkers for detecting early cancer, for predicting prognosis, and for predicting response to therapy. Yet very few markers are brought to clinical practice, and many are not even found to be reproducible. This essay discusses chance and bias as threats to validity that can explain the huge disconnect between promise and product, along with approaches to address those threats.
在许多“组学”领域，人们对生物标志物在检测早期癌症、预测预后以及预测治疗反应方面的能力做出了非凡的承诺。然而，很少有标志物被应用于临床实践，而且许多标志物甚至被发现无法重复。本文讨论了机遇和偏倚作为对有效性的威胁，它们可以解释承诺与实际成果之间的巨大脱节，同时还讨论了应对这些威胁的方法。
