Tobacco Documents Online

General Definition of Bias: I I I I I I I i I I I I Ik , Deviation of results or inferences from the truth; or processes leading to such deviation. AnU trend in the collection, anaIysis, interpretation; publication or reuiezo of data that can lead to conclusions that are systernatically different from the truth.l Definition of Specific Biases: PubIication Biases: Reviewer Bias - Systematic error due to failure of journal editors to accept and'1 publish reports with negative, non-sigruficant; or contrary conclusions. File-drawer Bias - Systematic error due to failure of authors to submit reports on negative, non-significant, or contrary conclusions. Researcher Bias - Systematic error due to failure of authors to include negative, non- significant, or contrary conclusions in reports documenting multiple-endpoint studies. Subject Biases: Recall Bias - Systematic error due to differences between case and control subjects in accuracy or completeness of' recall of prior events or experiences that may be related to the medical endpoint of concern. Reporting Bias - Sy*st'ematic error due to selective suppression (or revealing) by the subject of informatioA such as past history of other disease that is related to the medical endpoint of concern. Misclassification Bias - Systematic error due to inclusion of subjects in case or controli groups who do not meet exposure aiteria. M'edical Biases: Detection Bias - Systematic error due to differing methods of ascert~ainment, diagnosis, or verification of cases between exposure gToups. N C Autopsy Bias - Systematic error resulting from the fact that autopsies represent a N . nonrandorn sample of' deaths. • ~ 21J 'r'r Dicioncni of tpiderniolbonJ, Sc:ond --di:ion. Ed: Last JTyi. Oxford University Press, New York 198& W ~ ~ 1 1_

Statistical Biases: I I I I I I I I I I i 3 I Design Bias - Systematic error due to faulty design of a study, including uncontrolled confounding, poorly defined~ populations, and nonsimultaneous comparisons using, historical controls. Sampling Bias - Systematic error due to nonrandom indusion of subjects from the reference population because of availability of subjects, willingness of subjects to participate, criteria for selection, use of hospital cases and/or controls, and subse4uent! follow-up failure, withdrawal or exclusion from the study.