Statistical considerations and modeling of clinical utility of tumor markers.

The proper assessment of potential tumor markers in the screening, diagnosis, prognosis, and monitoring of neoplastic diseases depends on careful attention to general principles of good design and analysis of scientific studies, as well as to some specific statistical principles to avoid erroneous conclusions. Published reports on tumor markers should be evaluated critically with these principles in mind. It is well known that statistical problems exist in the medical literature despite the peer review process. In addition, attention to these principles in the design of studies of tumor markers can help ensure their accuracy and validity and avoid the common pitfalls.

Duke Scholars

Author Stephen L. George Biostatistics & Bioinformatics, Division of Biostatistics