SPECIAL COMMUNICATIONS: Letters to the Editor-in-Chief
A recent study by Ryan et al. (7) published in Medicine & Science in Sports & Exercise® addressed the dose-response of stretch-induced muscular force deficits. However, their findings need to be interpreted in the context of several previous studies on this topic that they seem to have overlooked. In addition, weaknesses in their sample size and statistical analysis also affect the strength of their conclusions.
Despite their claim that "to date, however, no studies have examined the potential dose-response relationship that may govern the stretching-induced force deficit" (p. 1530), several studies have reported on just that (e.g., [1-3,5,6]). Three years ago, Knudson and Noffal (3) first reported a logarithmic decline in isometric strength after stretching. The threshold of 20-40 s they reported (depending on the statistical or functional standard of significance) has been confirmed by several studies reporting strength reductions after only 30-60 s of stretching (1,2,5). Ryan et al. are likely incorrect in concluding that the doses of 2-8 min of stretching do not significantly reduce peak ankle plantarflexion torque. It is more likely that their results represent a Type II error resulting from a small sample size and inadequate statistical power.
Ryan et al. performed a repeated-measures design examining six dependent variables with 13 subjects. Given this small sample and the variation in their data (CV = 10%), it is not clear that they have the statistical power they claim, although they did not make it clear what variation and meaningful change they expected. Power calculations for special statistical procedures for multiple tests of multiple dependent variables are not the same as power calculations for a single repeated-measures analysis. The repeated testing of subjects across days also increases the unexplained variance relative to repeated-measures testing of subjects at one testing session. The across-days design with a small sample decreases the ability to detect statistically significant differences in the 5-10% range common in stretch-induced strength deficits.
The inability of the study of Ryan et al. to detect a reduction in peak torque after stretching is all the more surprising because of another error in analysis. Their use of multiple repeated-measures ANOVA on the same data set will inflate the experimentwise Type I error rate or false discovery rate (4,8). The six univariate tests (c = 6) on six dependent variables (assuming the typical 0.05 error rate) of Ryan et al. correspond to an experimentwise Type I error rate of 0.26 (EER = 1 − (1 − α)c) if the six variables were independent. If some of the dependent variables were correlated, the EER would be between 0.05 and 0.26.
There is no perfect design for studies of this important topic. There are too many confounding variables to easily address this issue in a single study. However, the design, analysis, and literature integration weaknesses noted above suggest that the observations and conclusions by Ryan et al. (7) should be interpreted with caution.
Duane Knudson, PhD, FACSM
Department of Kinesiology
California State University
1. Brandenburg JP. Duration of stretch does not influence the degree of force loss following static stretching. J Sport Med Phys Fitness
2. Kay AD, Blazevich AJ. Reductions in active plantarflexor moment are significantly correlated with static stretch duration. Eur J Sport Sci
3. Knudson D, Noffal G. Time course of stretch-induced isometric strength deficits. Eur J Appl Physiol
4. Lundbrook J. Multiple comparison procedures updated. Clin Exp Pharmacol Physiol
5. Nelson AG, Winchester JB, Kokkonen J. A single thirty second stretch is sufficient in inhibiting maximal voluntary strength. Med Sci Sports Exerc
. 2006;38(Suppl 5):S294.
6. Robbins JW, Scheuermann BW. Varying amounts of acute static stretching and its effects on vertical jump performance. J Strength Cond Res
7. Ryan ED, Beck TW, Herda DJ, et al. Do practical durations of stretching alter muscle strength? A dose-response study. Med Sci Sports Exerc
. 2008; 40(8):1529-37.
8. Sankoh AJ, Huque MF, Dubey SD. Some comments on frequently used multiple endpoint adjustments methods in clinical trials. Stat Med