Bias as a threat to the validity of cancer molecular-marker research.

Abstract

Claims that molecular markers can accurately diagnose cancer have recently been disputed; some prominent results have not been reproduced and bias has been proposed to explain the original observations. As new '-omics' fields are explored to assess molecular markers for cancer, bias will increasingly be recognized as the most important 'threat to validity' that must be addressed in the design, conduct and interpretation of such research.