Combined use of intersubject and intrasubject controls to distinguish treatment versus patient effect in retrospective studies

Clinical researchers seek to eliminate specific subject attributes from becoming confounding variables to distinguish a treatment effect from an underlying patient effect. In prospective studies, the random assignment of subjects into experimental groups results in the equal distribution of subject attributes, thereby eliminating them as confounding variables. Retrospective study designs, often used in prosthodontics research, preclude the random assignment of subjects. This article describes the application of a standard research design that combines the use of intersubject and intrasubject controls to compensate for the potential imbalance in subject attributes in retrospective clinical studies. This approach should help prevent clinical investigators from declaring differences between experimental groups when, in fact, no differences exist.