Underpowered studies in muscle metabolism research: Determinants and considerations

Biomedical research frequently employs null hypothesis testing to determine whether an observed difference in a sample is likely to exist in the broader population. Null hypothesis testing generally assumes that differences between groups or interventions are non-existent, unless proven otherwise. Because biomedical studies with human subjects are often limited by financial and logistical resources, they tend to have low statistical power, i.e. a low probability of statistically confirming a true difference. As a result, small but potentially clinically important differences may be overseen or ignored simply due to the absence of a statistically significant difference. This absence is often misinterpreted as 'equivalence' of treatments. In this educational paper, we will use practical examples related to the effects of exercise and nutrition on muscle protein metabolism to illustrate the most important determinants of statistical power, as well as their implications for both investigators and readers of scientific articles. Changes in muscle mass occur at a relatively slow rate, making it practically challenging to detect differences between treatment groups in a long-term setting. One way to make it 'easier' to differentiate between groups and hence increase statistical power is to have a sufficiently long study duration to allow treatment effects to become apparent. This is especially relevant when comparing treatments with relatively small expected differences such as the effect of modest changes in daily protein intake. Secondly, one could try to minimize the variance and response heterogeneity within groups, for example by using strict inclusion criteria and standardization protocols (e.g., meal provision), by using cross-over designs, or even within-subject designs where two interventions are compared simultaneously (e.g., studying an exercised limb vs a contralateral control limb) although this might limit the generalizability of the findings (e.g. such single-limb exercise training is not common in practice). In terms of data interpretation, investigators should obviously refrain from drawing strong conclusions from underpowered studies. Yet, such studies still provide valuable data for meta-analyses. Finally, because muscle protein synthesis rates are highly responsive to anabolic stimuli, acute metabolic studies are more sensitive to detect potentially clinically relevant differences in the anabolic response between treatments. Apart from further elaborating on these topics, this educational article encourages readers to more critically question null findings and scientists to more clearly discuss limitations that may have compromised statistical power.