Length bias: Mammography detects a cancer while it is preclinical, and preclinical durations vary.

Screen-detected cancers have a more favorable prognosis than do interval cancers, even when matched for size and stage; this is an expression of length bias.

If the effect of length bias is relatively small, substituting for might not be an upper bound, although we believe it would be a reasonable approximation.)

Overdiagnosis bias: An extreme form of length bias; screening may find cancers that are very slow growing and that would never have become manifest clinically.

There was, however, no adjustment for self-selection bias, lead-time bias, overdiagnosis bias, or length bias.

Several observational studies have examined process measures such as sensitivity and case-survival data, but without appropriate controls and with no adjustment for lead-time and length biases.

Because awareness of a positive family history can lead to more frequent work-ups for cancer and result in apparently earlier prostate cancer detection, assessments of disease progression rates and survival after diagnosis are subject to selection, lead time, and length biases.

While improved survival rates after initiation of screening have been reported,[8,9] these observations should be viewed cautiously because improvements could be caused by lead-time bias, length bias, and identification of cases through screening that would have spontaneously regressed.

Length bias is the concept that slower growing, more indolent tumors are more likely to be diagnosed by screening tests, but improvements in diagnosing more cases of indolent cancer may not translate into better patient outcomes after the implementation of screening programs.