Modelling of intrinsic and extrinsic injury risk factors in elite tennis

Elite tennis places high physical, physiological and psychological stresses on its players which can amplify injury risk and limit player development and performance (1, 2). Therefore, understanding what and why injuries occur in both elite junior and professional tennis is critical in mitigating the risk. Unfortunately, current research on elite tennis injury is sparse. What is available is limited by methodological design and conflicting findings. Consequently, this thesis aimed to address these gaps through a more complete examination of the injury epidemiology and aetiology of elite tennis injuries throughout the entire player pathway. Specifically, Study 1 examined the epidemiology and in-event treatment frequency of injury at the 2011-2016 Australian Open tournaments. Injury incidence was defined as a medical consultation by a tournament physician, and in-event treatment frequency as the mean total number of follow-up medical/physiotherapy consultations (2013-2016 tournaments only). Data were collated by sex, injury region and type and reported as frequencies per 10,000 game exposures. Incidence rates ± 95% confidence intervals (CI) and rate ratios (RR) were used to test effects for injury, sex and year. Female players were found to experience more injuries than male players over the 6-years (201.7 vs 148.6). The shoulder (5.1 ± 1.1 injuries per year) and knee (3.5 ± 1.6) were the most commonly injured region among females and males respectively. The torso region attracted high in-event treatment frequencies in both sexes. Muscle injuries were the most common type of injury, yet stress fractures more than doubled over the 6-year period. Overall, there were sex differences in injury incidence at the Australian Open, however the in-event treatment frequency of injuries was similar for both males and females. Where Study 1 detailed the injury epidemiology of professionals at the Australian Open, Study 2 focused on the incidence and severity of injuries of elite junior tennis players from a national program over multiple years. Injury data were collated by sex, age and body-region for all nationally-supported Australian junior players (58 males, 43 females, 13-18 y) between 2012-2016. Injury was defined as a physical complaint from training/matchplay determined by presiding physiotherapists and doctors that resulted in interrupted training/matchplay. Severity represented the days of interrupted training/matchplay per injury. Injury incidence was reported per 1,000 exposure hours. Incidence rate change and RR ± 95% CI assessed changes over time. No difference in male and female injury incidence existed (2.7 ± 0.0 v 2.8 ± 0.0), however male injuries were more severe (3.6 ± 0.6 v 1.1 ± 0.9 days). The lumbar spine was the most commonly and severely injured region in both sexes (4.3 ± 0.2, 9.9 ± 1.4 days) with shoulder injuries the second most common in both sexes (3.1 ± 0.2) and registering the second highest severity in males (7.3 ± 1.4 days). The body-region findings are relatively consistent with the body-region findings discovered in Study 1. Independent of sex, the injury incidence increased with age from 2.0 ± 0.1 (13 y) to 2.9 ± 0.1 (18 y). This study concluded that despite no sex-based difference in injury incidence, male injuries resulted in more interrupted days of training/matchplay. The profiling of injuries across the elite tennis pathway highlighted high risk sexes, ages and body-regions of injury. An understanding of why these injuries occur is important. As the volume and intensity of training is commonly associated with injuries in sport, the aim of Study 3 was to determine the best internal workload model and timeframe to predict injury in elite tennis. Daily training loads, recorded as session-RPE (sRPE)(3), and injury incidence data (2012-2016) from nationally ranked tennis players (n = 101, 19.1 ± 2.8 y, 91 ± 112 peak national ranking) were obtained. Injuries were defined as per Study 2. Multiple workload metrics, including variations of acute:chronic workload ratios (ACWR) and exponentially weighted moving averages (EWMA) as well as daily loads, monotony and strain were assessed over numerous timeframes (8 time points between 1-60 days) to predict subsequent injury. The predictive performance of the models was assessed via area under the curve (AUC) of receiver operator characteristics (ROC) curves and reported as AUC ± 95% confidence limits, sensitivity and specificity. It was found that the daily rolling average load and EWMA models performed best (³0.76 ± 0.04 AUC), largely owing to predicting non-cases rather than injuries (³0.16 sensitivity, ³0.74 specificity). There were no differences between timeframes. All other models performed relatively poorly (0.50-0.66 ± 0.0). Non-injured players experienced higher loads compared to injured players (mean ± SD; 714 ± 521 au, 565 ± 426 au). Overall, the best predictive workload-injury models used daily rolling averages and EWMA regardless of the tested timeframes. This suggests load model selection is more important than timeframe selection for predicting injuries in tennis. The applied impact of Study 3 may be limited by the fact that only internal loads were related to injury outcomes, without the consideration of other risk factors. Therefore, Study 4 examined the association of intrinsic and extrinsic risk factors with injuries in elite junior tennis players over a 500-day period (26 males, 23 females, 15.6 ± 2.0 y, 143 ± 128 peak national ranking). Daily training loads, training types, perceptual wellbeing and soreness as well as baseline musculoskeletal function and physical capacity measures were collected from junior players in a national program. Training loads included serve counts and sRPE calculated as daily load and 21-day rolling average in the lead up to injury. Seventy-eight injuries occurred which were defined in the same way as Studies 2 and 3. Univariate Cox Proportional Hazard models determined the first factor in the forward step selection for the multivariate Cox Proportional Hazard model. The multivariate model then determined the aggregated factors with the strongest association to injury, with results reported as hazard ratios ± 95% CI. The multivariate analysis revealed that lower serve counts (-0.02 ± 0.00) and a higher number of sore body-regions on the day of injury (0.84 ± 0.48) combined with lower multistage fitness test scores (-0.46 ± 0.23) and slower change of direction (COD) speed (3.47 ± 2.88) were best at detecting injury hazard. Internal loads quantified as daily rolling averages (as per Study 3) did not feature in the multivariate model outcome in Study 4. Perhaps the inclusion of an external load measure, or the combination of other risk factors, limited the influence of internal load in explaining injury hazard. In conclusion, the interplay of serve loads, regions of body soreness, aerobic capacity and COD speed provided the highest association with injury. These risk factors have currency as injury prevention and training monitoring tools in high-performance tennis. In summary, this thesis provides insight into the epidemiology and aetiology of tennis injuries throughout the elite pathway. The findings highlight that there are consistencies in the body-region of injuries in both elite junior and professional players, however sex-based differences are apparent. Study 3 showed that there is some injury predictive power in univariate internal training loads, yet it appears limited to detecting when injuries do not occur. Consequently, the multivariate analysis of injury risk conducted in Study 4 found that the collection of serve loads, soreness and physical capacities should be monitored as part of an injury prevention program for elite tennis players.