The science of early detection.

Most clinical judgment and clinical intuition derives from observations made on patients who suffer from a disease or medical condition. However, the target population for cancer screening is healthy people who would not seek out a health professional unless convinced to do so by advertising or public messages. Extrapolation of clinical observations to the target population for screening can be very misleading and even harmful. This is because powerful screening biases and confounding effects--such as selection bias, lead-time bias, length-bias sampling, and overdiagnosis--can mislead even the most astute clinician. This article will discuss those biases, review methods to avoid them, and provide useful resources to the clinician or health scientist. Copyright 2004 Elsevier Inc.