Comparison between 2,000 m and 3,000 m time trials to estimate the maximal aerobic speed for collegiate runners

Abstract

Considered to be a lesser resource burden, 2,000 and 3,000 m time trials (TTs) have been recognized as alternatives to accurately estimate the maximal aerobic speed (MAS) derived from laboratory-graded exercise testing (GXT). Previous studies have commonly used ordinary least squares linear regression and the Bland–Altman method to compare the agreement between MAS and TT performance. The agreement analysis aimed to identify the systematic bias between the results of the two methods, rather than to identify similarities. The model II regression technique (ordinary least product regression) is increasingly favored by researchers in the field of physiology. Thus, we aimed to 1) use the ordinary least product (OLP) and bootstrap methods to determine the agreement between the average speed of 2,000 m TT (S2000) and the average speed of 3,000 m TT (S3000) and 2) determine whether S2000 or S3000 can accurately approximate the GXT-derived MAS. It is used as an alternative to estimate the MAS and prescribe training intensity. Thirty-five Beijing Sport University recreational male runners completed an MAS test in laboratory settings, followed by 2,000 and 3,000 m TTs randomly, with a 7-day interval. OLP regression was used to analyze the agreement between the GXT-derived MAS and S2000 and S3000. The bootstrap method was used to calibrate the equations. Differences between the GXT-derived MAS and S2000 and S3000 were compared using a one-way repeated measure analysis of variance (ANOVA) and a post hoc analysis (Bonferroni). The significance level was p < 0.05. The results showed that before calibration, the 95% CI of the OLP regression intercept and slope between the GXT-derived MAS and S2000 and S3000 did not include 0 and 1.00, respectively. These values, after calibration, included 0 and 1.00, respectively. Post hoc analysis revealed that S3000 closely approximated the GXT-derived MAS and underestimated 0.46% (0.06 km h−1 and p > 0.05), and S2000 overestimated 5.49% (0.81 km h−1 and p < 0.05) by the MAS. It concluded that the 3,000 m TT performance approximated the GXT-derived MAS compared to the 2,000 m TT performance. There exist fixed bias and proportional bias between the GXT-derived MAS and TT performance. More attention should be applied to calibration when using the TT performance to estimate the MAS.