Exclusion Criteria

Exclusion criteria are a set of predefined definitions that is used to identify subjects who will not be included or who will have to withdraw from a research study after being included. Together with inclusion criteria, exclusion criteria make up the eligibility criteria that rule in or out the participants in a research study. Similar to inclusion criteria, exclusion criteria are guided by the scientific objective of the study and have important implications for the scientific rigor of a study as well as for assurance of ethical principles. Commonly used exclusion criteria seek to leave out subjects not complying with follow-up visits, those who are not able to provide biological specimens and data, and those whose safety and ethical protection cannot be assured.

Some definitions are needed to discuss exclusion criteria. Generalizability refers to the applicability of study findings in the sample population to the target population (representativeness) from which the sample was drawn; it requires an unbiased selection of the sample population, which is then said to be generalizable to, or representative of, the target population. Ascertaining exclusion criteria requires screening subjects using valid and reliable measurements to ensure that subjects who are said to meet those criteria really have them (sensitivity) and those who are said not to have them really do not have them (specificity). Such measurements should also be valid (i.e., should truly measure the exclusion criteria) and reliable (consistent and repeatable every time they are measured).

The precision of exclusion criteria will depend on how they are ascertained. For example, ascertaining an exclusion criterion as “self-reported smoking” will likely be less sensitive, specific, valid, and reliable than ascertaining it by means of testing for levels of cotinine in blood. On the other [Page 438]hand, cotinine in blood may measure exposure to secondhand smoking, thus excluding subjects who should not be excluded; therefore, a combination of self-reported smoking and cotinine in blood may increase the sensitivity, specificity, validity, and reliability of such measurement, but it will be more costly and time consuming.

A definition of exclusion criteria that requires several measurements may be just as good as one using fewer measurements. Good validity and reliability of exclusion criteria will help minimize random error, selection bias, and confounding, thus improving the likelihood of finding an association, if there is one, between the exposures or interventions and the outcomes; it will also decrease the required sample size and allow representativeness of the sample population. Using standardized exclusion criteria is necessary to accomplish consistency, replicability, and comparability of findings across similar studies on a research topic. Standardized disease-scoring definitions are available for mental and general diseases (Diagnostic and Statistical Manual of Mental Disorders and International Classification of Diseases, respectively). Study results on a given research topic should carefully compare the exclusion criteria to analyze consistency of findings and applicability to sample and target populations. Exclusion criteria must be as parsimonious in number as possible; each additional exclusion criterion may decrease sample size and result in selection bias, thus affecting the internal validity of a study and the external validity (generalizability) of results, in addition to increasing the cost, time, and complexity of recruiting study participants. Exclusion criteria must be selected carefully based upon a review of the literature on the research topic, in-depth knowledge of the theoretical framework, and their feasibility and logistic applicability.

...