Appropriateness of Sample Sizes in Published Research on Prosthetic Knee Componentry

    loading  Checking for direct PDF access through Ovid

Abstract

Introduction

A research study's sound approach to statistical analysis holds the key to its validity and, thereby, its credibility in the medical community. Previous research shows that the credibility of prosthetics research is often rated at comparably low levels, partly because of its reported small sample sizes. We analyzed how authors of studies on prosthetic knee componentry have reported limitations regarding sample size and sample composition. To that end, we calculated the observed statistical powers in the recent literature to investigate whether sample sizes were sufficient. We further investigated how representative sample demographics and limitations were of the overall target population.

Methods

Observed power was used as the main identifier of a study's appropriate (i.e., large enough) sample size. Powers of at least 0.80 were considered to indicate a sufficient sample size for the studies in this review. Reported study limitations were qualitatively analyzed and compared with observed powers.

Results

A comprehensive search of PubMed resulted in 32 articles meeting the inclusion criteria. Studies on average had around 20 participants and rarely more than 50 participants. Of the 32 articles, 19 either reported statistical power or contained enough information to calculate it. Only three of these articles achieved a power of at least 0.80. Regardless of specific sample size or power, most studies reported some sort of limitation in relation to the size or homogeneous nature of their sample.

Discussion

Our findings suggest that most research studies in the field of prosthetics are underpowered. Given the noted difficulty of increasing sample sizes, it seems important to choose comparison variables that promise a large effect size and/or to deliberately reduce the heterogeneity (i.e., the standard deviation) of the subject sample to address this problem.

Related Topics

    loading  Loading Related Articles