Summary
Surveys biases that even randomization cannot eliminate, including allocation concealment failures that Schulz showed inflate odds ratios by 30 to 40 percent, blinding failures that affect subjective outcomes, the placebo and Hawthorne effects, the healthy-adherer effect from the Coronary Drug Project, and contamination as seen in the COMMIT smoking-cessation trial. Develops time-related biases in pharmacoepidemiology and screening, including immortal-time bias, lead-time bias, length-time bias, and protopathic bias. Closes with the age-period-cohort identification problem and time-window biases that haunt longitudinal analyses.