Experimental design issues in clinical research of musculoskeletal pain disabilities

The purpose of the present article is to provide a basic overview of important experimental design and statistical analysis issues/concepts that require the attention of clinical researchers in the area of musculoskeletal disorders who attempt to demonstrate the effectiveness of a particular treatment modality. We begin with a discussion of important design issues in clinical treatment-outcome research, specifically those related to internal validity, external validity, and statistical conclusion validity (which involves concepts such as statistical power, assumptions of statistical tests, error rates of a statistical test, the operationalization of the key outcome measures, and the reliability of such outcome measures). We next discuss the importance of attempting to conduct the highest quality type of clinical outcome study to infer cause-effect relationships - the randomized controlled trial (RCT). Even though an RCT is often viewed as an important benchmark to use in considering the validity of treatment-outcome results, we also point out that RCTs can vary greatly in the degree of internal and external validity that may make them less than 'perfect' in nature. Fortunately, there are a host of other experimental designs that may be appropriately employed that can yield important scientific data to help in delineating cause-effect relationships (so-called quasi-experimental designs). A hierarchy of experimental designs is also presented in a descending order of rigor: the controlled group outcome study; the single-subject experiment and replicated single-subject experiments; the single-group outcome study; the systematic case study; and the anecdotal case report. We also review useful 'alternative' quantitative and qualitative methods for clinical outcome research that can be used when there are problems encountered with conventional research methodology. Finally, the use of meta-analysis techniques is reviewed as a means of coming up with a concordant summary statistic of effect size across many different experiments.