Adequate statistical power in strength and conditioning may be achieved through longer interventions and high frequency outcome measurement.

Abstract

Pre-post randomised controlled trials (RCT) are the most common design used to build an evidence base in strength and conditioning but are limited by small effects, small samples sizes, and concomitant low statistical power. The purpose of this study was to explore the effects of manipulating a range of factors including intervention length, frequency, and pattern of outcome measurements on the sample size required to achieve adequate statistical power. A case-based approach was used to enhance applicability.

Realistic data generating patters were considered for hypothetical RCTs investigating resistance training interventions to improve maximum strength as measured by the 1RM bench press. Improvements for the ‘reference’ intervention and subsequent average treatment effect for the ‘testing’ intervention were matched to data summarised in recent large meta-analyses. Different measurement error magnitudes were added to the high frequency RCT data to recreate the use of 1RM prediction methods that could be used during training sessions. A closed form solution linking statistical power and sample size was used to explore different strategies with simulations performed as a final check.

The results showed large improvements in statistical power could be achieved when conducting interventions over a longer period (e.g. 18 weeks), and/or performing multiple outcome measurements. Efficient reductions in required sample sizes could be achieved by performing multiple measurements at baseline and post-intervention. This strategy, however, may be limited by induced fatigue or training effects. Similar reductions in sample size could be achieved by performing high-frequency measurement throughout the intervention. This reduction in sample size was demonstrated despite acute increases in measurement error (factor of 1.5 and 2) that would occur when using prediction methods.

In conclusion, very low statistical power is likely the norm in pre-post RCTs in strength and conditioning. Simply increasing sample size is unlikely to remedy the situation given the resource constraints that are common in the discipline. The results of this study suggest that researchers should consider other strategies including longer interventions and high frequency data collection to obtain adequate statistical power with feasible resources.