How to Reduce False Positive Study Results Due to Bias

Mon 3:30 PM - Track 2: Computational, Statistical, and Epidemiological Considerations for Mass Spectrometry in the Clinical Laboratory

David Ransohoff University of North Carolina School of Medicine

David Ransohoff, University of North Carolina School of Medicine, Chapel Hill, NC 27514

A major cause of problems in cancer marker research is when bias occurs -- a systematic difference in the compared groups that is not due to what you are trying to measure. For example, your goal is examine markers in blood in a comparison of cancers vs controls, but cancer bloods happen to come from older men while controls from younger women. Bias may lead to erroneously 'positive' study results. Bias is a particular problem in the observational or non-experimental research design used to study diagnostic tests. Sources of bias will be reviewed, along with 'what a biochemist ought to know' about how to address the problem.

Email: ransohof@med.unc.edu

---