Further studies from California investigated risk factors for suspensory apparatus failure and fractures of the Mc3 condyles [43] and scapula [44]. A longer interval since the last 60+‐day period without a race and the distance exercised in the last month (suspensory apparatus failure) or two months (Mc3) were associated with an increased risk. For every extra day since the last 60+‐day lay‐up, the odds of condylar fracture increased by 0.3%. The odds of suspensory apparatus failure remained level for up to 120 days since the last 60+‐day lay‐up, but increased thereafter: 3.4 times for periods between 121 and 214 days since the last 60+‐day lay‐up and 5.9 times for periods greater than 320 days since the last 60+‐day lay‐up. For every extra furlong exercised at fast pace, the odds of both outcomes increased by 4% [43]. Most significantly, the work on scapular fractures, although limited somewhat by a lack of statistical power, demonstrated that a tapering off of the total distance in the month prior to fracture (compared with the preceding month) was seen more frequently in horses with fractures compared to control horses [44].
A range of exercise‐related risk factors were demonstrated by the same group when investigating proximal sesamoid bone fractures [45]. For example, compared with horses that died or were subject to euthanasia for other reasons, horses that had sustained proximal sesamoid bone fracture(s) were more likely to have spent a greater time in active training and racing, completed more exercise events, exercised further during their whole career and had higher exercise intensities in the 12 months prior to fracture.
Similar work conducted in the UK has identified associations between the risk of fracture in training or racing and exercise distance over relatively short time periods [28, 46]. Horses that trained over more than 220 furlongs (44 km) at canter speed and 30 furlongs (6 km) at gallop speed, in a 30‐day period were at the highest risk of fracture [46]. More specifically, the risk of pelvic or tibial stress fracture increased with increasing distance cantered up to a maximum at around 250 furlongs (50 km) per 30‐day period [28]. Importantly, the associations with the gallop distance correspond very closely with those reported in California (six furlongs per week) [42]. Even though the case definitions and the populations studied were quite different, the fact that these results concur is regarded as evidence of a true causal effect and provides greater confidence when offering advice on optimal training distances at different speeds.
The association between average fast‐pace distance and musculoskeletal injury has also been demonstrated in studies from Australia [47, 48]. In the first of these, two‐year‐old horses that had a greater percentage of fast work days during their first fast work preparation were more likely to sustain musculoskeletal injury that ended the training preparation. The average distance trained at speeds greater than or equal to 800 m/min was also positively associated with musculoskeletal injury [47]. The second study investigated fatalities in flat racing and demonstrated that the high‐speed distance accumulated during the period 31–60 days prior to a race start was most important in determining the likelihood of fatality [48]. In a parallel study of fatalities in jump racing, the total number of career starts and having started more than once in the period 14 days prior to the case race were both associated with an increased likelihood of fatality [49]. Although these studies used a broader case definition of ‘fatality’, it was previously reported that the majority were due to musculoskeletal injury [13]. It is therefore most likely that this result is due to the effect of exercise as identified in the studies conducted in the USA and the UK. The moderate differences in hazard periods between these studies may be due to the broader case definition or to local differences in the racing population and racing and training practices.
All of these studies provide good evidence of an association between increased amounts of high‐speed exercise and risk of severe musculoskeletal injury and/or fracture that support the hypothesis that horses doing large amounts of fast exercise accumulate (sub‐clinical or clinical) bone damage that can ultimately result in catastrophic failure [50–52].
At the other end of the ‘exercise scale’, a few studies have demonstrated the deleterious effect of a lack of fast training work on the risk of fracture during racing [53–55]. The ability of bone to adapt to its mechanical environment is well documented [56, 57], and in the racehorse this is principally influenced by the training programme to which the horse is exposed. Changes to the distal condyles of Mc/Mt3 of horses in race training have been observed [58, 59], and more specifically the subchondral bone of this region has been shown to undergo an adaptive response to high‐speed exercise [60]. The bones of the horses which were doing no high‐speed exercise in the observational epidemiological studies are therefore unlikely to have adapted to the loads that they would experience under racing conditions, thus exposing them to increased risk of fracture [54, 55].
In the final multivariable models produced for catastrophic distal limb fracture and lateral condylar fracture, the best‐fitting form of the variable relating to the distance galloped in training indicated that the risk was highest for horses doing no fast work. For horses doing between 4 and 10 furlongs of fast work per week, the risk was reduced, and thereafter the level of risk did not alter [54, 55]. Only a few high load cycles have been demonstrated as sufficient to induce an osteogenic response in avian ulnas [61]. By extrapolation, relatively short distances of gallop work during training may be adequate to stimulate adaptation and be protective against fracture during racing. Alongside the previous work suggesting an optimal six to seven furlongs of fast work per week, these findings may contribute to the formulation of training regimens specifically designed to reduce the risk of fracture.
An important caveat to these conclusions is that inferring causality is difficult, and it is possible that the associations between the absence of fast exercise and increased risk of fracture are an example of effect rather than cause. In other words, horses that are suffering sub‐clinical injury are unable to train to the same extent as the rest of the population, and it is the sub‐clinical injury itself and not the reduced exercise that increases the likelihood of fracture. At this point in epidemiological investigations, it becomes necessary to either design intervention studies that prospectively examine the impact of a training modification that is designed to minimize the risk of fracture (for example in this case to ensure all horses do at least some fast work) or to encourage more comprehensive recording of veterinary and treatment records of horses in training so that existing (sub‐) clinical injury can be accounted for during the model building process [62].
Another important aspect of an effective training regimen are rest periods and correct management of return from rest. In the racehorse, some work has demonstrated that inappropriate rest periods can be detrimental. In the California post‐mortem studies, a hazard period of up to 10 days following a 60+‐day period of rest was identified as being most significant with respect to the risk of fracture [63]. As previously intimated [64], the authors hypothesized that this may be due to the fact that osteoclastic resorption had taken place, but osteoblastic remodelling was not yet complete. The bones of horses returning to exercise before the remodelling process is complete are likely to be less able to withstand training load than before the period of rest. Based on the differing rates of osteoclastic and osteoblastic activities, the hazardous period of rest is hypothesized to be between 30 and 90 days [65]. Humeral fractures were found to have significant acute callus formation, indicating stress fracture remodelling prior to catastrophic failure [63]. It is plausible that it is only for fatigue/stress fractures that the length of rest period is important. Greater data resources will enable the refinement of case definitions to allow the investigation of risk factors for specific injuries. It is therefore likely that future studies will identify associations with periods of rest for other fracture types that have a similar pathogenesis to humeral fracture.
The Importance of Detailed Information About Horses Under Investigation
Current epidemiological work aimed at minimizing fracture incidence is moving towards prevention by way of accurately identifying horses at significant risk. In order for such predictive models to be useful, it is important that the overall accuracy of prediction is high. At present, the predictive ability of models (Section Predictability and Potential for Effective