Anaerobic assessment and training monitoring in elite swimmers

Abstract

Performance in elite swimming is multifactorial and is influenced by training load, anthropometry, technique and physiology among other factors. Quantifying and mathematically modelling the relationships between these variables and their impact on performance, talent identification and training individualisation is a key role of swimming researchers in assisting coaches and athletes. To model these relationships effectively one must first accurately measure each of these variables. While the majority of these variables have well established field and laboratory measurement protocols, there is one major exception in swimming physiology research - the anaerobic energy system. The anaerobic energy system plays an important role in elite swimming performance given that 27 of 35 Olympic events derive 25-73% of their energy from anaerobic sources. As such it is vitally important for coaches to have a standardised test which assesses metrics related to both the aerobic and anaerobic systems and can be used to evaluate the impact of training. This thesis examines a practical issue in elite swimming; the lack of a simple standardised test to provide useful information which on performance metrics related to both the aerobic and anaerobic energy systems. A pool-based methodology based on the 3-minute test in cycling was proposed and assessed for reliability and validity. The newly proposed methodology for a 3-minute swimming test (12x25m) was then examined using a within-swimmer model to assess changes over a season, cross-sectionally between-swimmers, and correlated with competition performances. This methodology may provide coaches with a practical means for monitoring or assessing aspects of both the aerobic and anaerobic energy system which relate to performance using only a stopwatch and a spreadsheet. Initially the impact of training load on performance is examined within this thesis. The relationship between training load and performance is widely regarded as a dose-response relationship. However, there are a number of methodological procedures which have not been scientifically interrogated beyond their initial proposal and proof of concept investigations. To assess both the method of quantification, and the method of modelling, the training load-performance relationship competition performances over 50 and 100 m efforts were matched to training load over >600 days for three elite swimmers. Two different methods of training quantification (5- and 7-zone methods) and three different methods of modelling the load-performance relationship, the Banister, Busso and rolling averages models, were assessed. The rolling averages approach, as it is much simpler mathematically, was deemed preferable if it could provide a similar level of accuracy when compared with the two alternatives which are considered industry standards. Both five- (mean r-value, 90% CI; 0.65, 0.28 – 1.00) and seven-zone (0.62, 0.24 – 1.00) training load methods correlated similarly with performance. When comparing modelling methodologies the rolling average method was less accurate (standard error range 1.32-1.36 s) than either the Banister model (standard error range 0.62-0.64 s) or the Busso model (standard error range 0.70-0.73 s). This investigation highlighted that coaches use 50 m performance to nominally assess the anaerobic system, demonstrating the need for a practical anaerobic test for swimmers. The critical speed (CS) and supra-CS distance capacity (D') framework was identified as an option that includes the more anaerobic aspects of performance. A time-trial based CS-D' test was assessed together with other variables including anthropometry, lower limb power and maturity which could modulate the relationship between this test and competition performance. Across 48 national age group swimmers (males 16.5 ± 1.2 y, females 15.5 ± 1.1 y; mean ± SD) the results of the CS-D' test correlated highly with performance. When combined with anthropometric measures, lower limb power and maturity measurements CS and D' explained 93% and 82% of the variation in performance for male and female 100 m swimmers respectively. When assessing 200 m performance these measures explained 84% (males) and 44% (females) of the variance. The CS-D' test appears useful, however the multi-day nature of the time-trial methodology was deemed too arduous for wide adoption by coaches. A new methodology was formulated based on the cycling three-minute test. This protocol consisted of twelve 25 m efforts at maximal unpaced effort, with five seconds rest (12x25m test). Metrics calculated from this protocol were peak speed, CS, D' and drop off %. To assess reliability this test was performed by 14 national and international-level swimmers. Test re-test reliability was assessed for all metrics and CS and D' values were compared to the values derived using the time-trial method for 23 swimmers to evaluate the degree of criterion validity. These variables were deemed reliable, all ICC values were >0.97 and coefficient of variation values were: CS 1.2%, D' 5.7%, peak speed 0.5% and drop off % 4.5%. This protocol also compared favourably to the time-trial method and produced estimates of CS which correlated very well (r = 0.96, p<0.05) with the time-trial method. Estimates of D' also correlated strongly (r = 0.79, p<0.05) between methods, although correlations were not strong enough to conclude the time trial and new protocol values are fully interchangeable. While reliability values are important, their relative size to the magnitude of their target signal has more bearing on whether they can be deemed sensitive enough to be useful or not. To assess this relationship Guyatt’s responsiveness index was calculated using data collected throughout multiple competitive seasons for a group of 27 national and international-level swimmers. For each of the four metrics the mean test to test difference was analysed and made relative to the standard deviation of the test-retest differences of a ‘stable’ population; in this case seven of the fourteen swimmer population used to calculate typical error values. The responsiveness index, which can be interpreted as a modified effect size, were computed as follows: CS 1.4, D' 2.2, peak speed 2.7 and drop off % 1.6. All these are large effects and indicate that the magnitude of mean changes between testing sessions compare favourably to test-retest reliability values. While the 12x25m test appears responsive enough within an athlete, it is also important to ascertain whether it can distinguish between athletes, and also whether a swimmer with faster or larger measures will swim faster in a competition environment. Metrics from the 12x25m test were compared to competition performance in a group of 34 national and international-level swimmers. When stroke and sex were taken as covariates, critical speed correlated well with 50 m (r = -0.54), 100 m (r = -0.57) and 200 m (r = -0.78) performances. Similarly, peak speed exhibited moderate to high relationships with performance especially in the shorter events (50m r = -0.75; 100m r = -0.79; 200m r = -0.59). Values for D' had a significant parabolic relationship with 50m (r = 0.61) and 100m (r = 0.57), but not 200 m (r = 0.32) performances. The 12x25m is likely a practical option to estimate anaerobic capacity in highly trained swimmers. The metrics calculated using the 12x25m test are reliable and have adequate criterion validity. These metrics are also responsive to changes within athletes and show a similar ability to distinguish between performances across swimmers as other analogous variables calculated using alternative methods. The 12x25m test is a standardised assessment of a swimmer’s ability to swim at high intensities and can be completed with minimal technological support.

Date of Award	2019
Original language	English
Supervisor	Ben RATTRAY (Supervisor) & Philo Saunders (Supervisor)