Interpreting results in 2 x 2 tables: part 9 of a series on evaluation of scientific publications.

BACKGROUND

The findings of epidemiological studies, diagnostic tests, and comparative therapeutic trials are often presented in 2 x 2 tables. These must be interpreted correctly for a proper understanding of the findings.

METHODS

The authors present basic statistical concepts required for the analysis of nominal data, referring to standard works in statistics.

RESULTS

The relative risk and odds ratio are defined to be indices for the relationship between two binary quantities (e.g., exposure--yes/no and disease--yes/no). The topics dealt with in this article include the effect of sample size on the length of the confidence interval and the p-value, and also inaccuracies caused by measuring error. Exposures are often expressed on a three-level scale (none, low, high). The authors also consider the 2 x 3 table as an extension of the 2 x 2 table and discuss the categorization of continuous measurements. Typically, more than one factor is involved in the development of a disease. The effect that a further factor can have on the observed relationship between the exposure and the disease is discussed.

CONCLUSIONS

Sample size, measurement error, categorization, and confounders influence the statistical interpretation of 2 x 2 tables in many ways. Readers of scientific publications should know the inherent problems in the interpretation of simple 2 x 2 tables and check that the authors have taken these into account adequately in analyzing and interpreting their data.