Training Load and Injury in Professional Ballet

In professional ballet, training load has frequently been suggested to be associated with the risk of musculoskeletal injury. Despite a recent surge in the number of training load research studies in high performance sport, relatively little research has been conducted investigating training load in ballet. The aim of this thesis was, therefore, to describe the training loads undertaken by professional ballet dancers, explore the load-injury relationship in ballet, and provide valid methods and recommendations for load management in professional ballet. Two five-season cohort studies were conducted, investigating scheduling and medical data at an elite professional ballet company. Shared frailty models were used to investigate relationships between individual risk factors, accumulated dance volume, and hazard ratios for injury risk. Greater week-to-week changes in dance volume and smaller seven-day dance volume were associated with increase rates of overuse injury, whilst age (traumatic injury), previous injury, and company rank (overuse injury) were also associated with increases in hazard ratios for injury. Analyses of scheduling data were consistent with previous research regarding the large rehearsal and performance volumes completed by ballet dancers. For the first time, however, the present research revealed the large variation in dance hours occurring from week-to-week, across the season, and between company ranks. In professional ballet, there is great scope to optimise training loads from increased emphasis on periodisation of the repertoire and rehearsal schedule alone. Three methodological studies explored the development and validation of training load measures in professional ballet. Firstly, the validity of session rating of perceived exertion in professional ballet dancers was investigated, revealing very large positive linear relationships with Edwards and Banister training impulse scores. Correlation coefficients were comparable across men and women, though were larger in rehearsals compared with ballet class. Secondly, a rule-based classifier for measuring jump frequency and height from accelerometer data was developed and validated, demonstrating a high degree of accuracy, and providing a simple means of managing jump load. Finally, a series of recurrent neural networks were developed to facilitate the measurement of tissue-specific forces outside of a laboratory using inertial measurement units, outperforming single variable linear regression approaches for the measurement of Achilles tendon, patellar tendon, and tibial force. Open-source software was developed and presented to house these algorithms, and database and visualize longitudinal training load data. This thesis demonstrates the importance of managing a rehearsal and performance schedule throughout a professional ballet season. Where more in-depth understanding of training load is required for managing high-risk dancers, this thesis provides practical, valid, and open-source methods for quantifying load.