Background: Most discussions of statistical methods focus on accounting for measured confounders and random errors in the data-generating process. In observational epidemiology, however, controllable confounding and random error are sometimes only a fraction of the total error, and are rarely if ever the only important source of uncertainty. Potential biases due to unmeasured confounders, classification errors, and selection bias need to be addressed in any thorough discussion of study results.
Methods: This paper reviews basic methods for examining the sensitivity of study results to biases, with a focus on methods that can be implemented without computer programming.
Conclusion: Sensitivity analysis is helpful in obtaining a realistic picture of the potential impact of biases.