Musculoskeletal injuries remain a global problem for the Thoroughbred racing industry and there is conflicting evidence regarding the effect of age on the incidence of injuries. The ideal time to commence race training is strongly debated, with limited supporting literature. There is also conflicting evidence regarding the effect of high-speed exercise on musculoskeletal injuries. There is a strong interest in developing training and management strategies to reduce the frequency of injuries. The types of musculoskeletal injuries vary between 2-year-old and older horses, with dorsal metacarpal disease the most common injury in 2-year-old horses. It is likely that risk factors for injury in 2-year-old horses are different than those for older horses. It is also likely that the risk factors may vary between types of injury. This study aimed to determine the risk factors for musculoskeletal injuries and dorsal metacarpal disease. We report the findings of a large scale, prospective observational study of 2-year-old horses in Queensland, Australia. Data were collected weekly for 56-weeks, from 26 trainers, involving 535 2-year-old Thoroughbred racehorses, 1, 258 training preparations and 7, 512-weeks of exercise data. A causal approach was used to develop our statistical models, to build on the existing literature surrounding injury risk, by incorporating the previously established causal links into our analyses. Where previous data were not available, industry experts were consulted. Survival analyses were performed using Cox proportional hazards or Weibull regression models. Analysis of musculoskeletal injuries overall revealed the hazard was reduced with increased exposure to high-speed exercise [Hazard ratio (HR) 0.89, 95% Confidence Interval (CI) 0.84, 0.94, p < 0.001], increased number of training preparations (HR 0.58, 95% CI 0.50, 0.67, p < 0.001), increased rest before the training preparation (HR 0.89, 95% CI 0.83, 0.96, p = 0.003) and increased dam parity (HR 0.86, 95% CI 0.77, 0.97, p = 0.01). The hazard of injury was increased with increasing age that training commenced (HR 1.13, 95% CI 1.06, 1.19, p < 0.001). Analyses were then repeated with the outcome of interest dorsal metacarpal disease. Factors that were protective against dorsal metacarpal disease and musculoskeletal injuries overall included: increased total cumulative distance (HR 0.89, 95% CI 0.82, 0.97, p = 0.001) and total cumulative days exercised as a gallop (HR 0.96, 95% CI 0.92, 0.99, p = 0.03), the number of the training preparations (HR 0.43, 95% CI 0.30, 0.61, p < 0.001). The age that training commenced was harmful for both dorsal metacarpal disease (HR 1.17, 95% CI 1.07, 1.28, p < 0.001 and overall musculoskeletal injuries.). The use of non-ridden training modalities was protective for dorsal metacarpal disease (HR 0.89, 95% CI 0.81, 0.97, p = 0.008), but not musculoskeletal injuries overall. The male sex increased the hazard of DMD compared to females (HR 2.58, 95% CI 1.20, 5.56, p = 0.02), but not MSI overall. In summary, the hazard of musculoskeletal injury is greatest for 2-year-old horses that are born from uniparous mares, commence training at a later age, are in their first training preparation, have undertaken little high-speed exercise or had limited rest before their training preparation. The hazard of dorsal metacarpal disease is greatest for 2-year-old horses that are males, commence training at a later age, are in their first training preparation, have undertaken little high-speed exercise or had limited use of non-ridden training modalities. Close monitoring of these high-risk horses during their training program could substantially reduce the impact of MSI. Furthermore, an understanding of how training methodologies affect the hazard of MSI facilitates modification of training programs to mitigate the risk impact of injury. The strengths of this study include a large sample size, a well-defined study protocol and direct trainer interviews. The main limitation is the inherent susceptibility to survival bias.