Matched Studies

Matched studies

Matched designs are for studies where the outcome is observed on the same individual on two separate occasions, under different exposure or treatment, or where two different methods applied. In case-control studies, we have binary outcome observations that follow matched or paired design in selecting individuals. Each case is then matched with one or more controls that are chosen to have the same values for major confounding variables. Having similar age or living in a same area are two examples of controls. However case-controls studies show that matched designs often have few advantages, and many have serious disadvantages. Unless the matching factor is strongly associated with both the outcome and the exposure the increase in efficiency may not be large. In some case control studies it is difficult to define th e population that gave rise to the cases. It is essential to note that if matching was used in the design, then the analysis must always take this into account. Stratifying on the case control sets are used to estimate exposure odds ratios but they are severely limited because they do not allow for further control of the effects of confounding variables that were not also matching variables. This is because each stratum is a single case and its matched controls, so that further stratification is not possible. For example, it cases were individually matched with neighbourhood controls then it would not be possible to stratify additionally on age group. Stratification can be used to control of additional confounders only by restricting attention to those case-control sets that are homogenous with respect to the confounders of interest. (Kirkwood, 214-223, 410-412)