Agility competition, one of the fastest growing canine sports worldwide,1 is a performance sport in which dogs and their handlers work as a team to navigate a sequence of obstacles. Dogs perform tasks that include jumping, weaving, making tight turns in tunnels, climbing ramps and seesaws, and moving on and off of or across elevated surfaces while being timed for speed and scored for faults. The Agility Association of Canadaa and the American Kennel Club2 reported increases of approximately 10%/y in the number of participants involved in this sport from 2003 to 2010. With these increases in participation, there is also a growing interest in understanding the factors that influence the risk of injuries among dogs participating in the sport.
In a retrospective survey study3 to characterize agility-related injuries among dogs and examine associations between injury characteristics and perceived causes of injury, our group determined that 1,209 of 3,801 (32%) dogs incurred ≥ 1 injury during agility-related activities, with a total of 1,602 injuries reported. Nine hundred sixty nine (60.5%) of these injuries were examined by a veterinarian. The most common types of injury were sprains, strains, and contusions of the shoulder, back, phalanges, and neck. Levy et al4 found a similar pattern of injuries in another retrospective survey and identified several potential risk factors for agility-related injuries, including a higher than expected proportion of injuries incurred by Border Collies, compared with dogs of other breeds.
Other studies3,4 regarding this subject have largely involved descriptive analytic methods and only examined factors in isolation. Multivariable techniques enable the investigation of individual risk factors while controlling for all others, and the use of these techniques enable a more comprehensive study of the complex interactions among the many variables that may contribute to agility-related injuries.5 The purpose of the study reported here was to conduct a large-scale retrospective survey to investigate potential risk factors for agility-related injuries among dogs. To our knowledge, the study reported here is the first to investigate potential risk factors for agility-related injuries by use of multivariable techniques.
Materials and Methods
Sample—Survey participants were handlers of agility dogs who had Internet access and were willing and able to complete an electronic survey in English. Respondents from any geographic location were eligible to complete the survey.
Survey instrument and procedures—A retrospective cross-sectional survey3,b was used to collect information on agility-related injuries among dogs. The survey instrument and research protocol were reviewed and approved by the University of Guelph Research Ethics Board.
Data were collected between March 16 and September 30, 2009. Participants were recruited for the survey through member lists of the Agility Association of Canada and United Kingdom Agility and through promotion on an internationally recognized canine agility dog handler's blog site.6 The survey was promoted on these parties' websites and Internet discussion lists and at local competitions.
Collection restrictions on the basis of internet protocol address were used to ensure that only 1 survey/respondent could be completed. Respondents with multiple dogs were instructed to complete a separate form for each dog. Additionally, for dogs that incurred multiple injuries, respondents were instructed to fill out separate injury forms for each injury (up to 5 injuries/dog).
The 27-item electronic survey was used to collect specific data on agility-related injuries incurred by dogs as recollected by handlers (confirmation of reported injuries by a veterinarian was not a requirement to participate in this study). Variables of interest for the present study included geographic location, other demographic information for handlers (eg, gender, age, and years of agility experience) and their dogs (eg, breed, age, height from the ground to the highest point of the shoulders [ie, withers], weight, number of years of participation in agility activities, use of preventive measures intended to keep dogs fit for agility (warmup, cool-down, or conditioning exercises; alternative therapeutic treatments [eg, acupuncture, massage, or chiropractic care]; or dietary supplement products), and frequency of practice sessions and participation in agility events during the past year. Further details and other results of the survey are reported elsewhere.3
Data analysis—Data are presented as mean ± SD (for continuous variables) or number and percentage (for categorical variables). Variables were compared between dogs that incurred injuries and those that did not. Height and weight of dogs and the number of years of experience in agility competition (for the dog and the handler) were compared via independent Student t tests. Geographic region, breed, and variables for measures intended to keep dogs fit for agility activities (use of warmup or cooldown exercises for agility events; conditioning exercises, including aerobic, strengthening, and proprioceptive or balance training; alternative therapeutic treatments such as acupuncture, massage, and chiropractic care; and administration of dietary supplement products [all yes or no responses]) were compared by use of Pearson χ2 analyses with values of P < 0.05 considered significant.
Unadjusted ORs and 95% CIs were calculated for all variables of interest by means of univariable logistic regression. Each dog was included in the model only once. Dogs that had > 1 injury reported in the survey were classified as having a previous agility-related injury. For several variables, categories were collapsed on the basis of the distribution of responses. For geographic region, there were 3 primary regions where respondents competed: the United States, Canada, and Europe. Respondents not competing in these regions were classified as other. The breed variable was collapsed into 2 categories (Border Collie and non-Border Collie) because a proportionally greater number of injuries was incurred by this breed, compared with all other breeds, in the preliminary analyses.
The final multiple logistic regression model was built with a purposeful selection of covariates.7,8 Variables that were significant at P < 0.25 in unadjusted analyses and remained significant in the multiple logistic regression model (adjusted OR) at P < 0.15 or for which removal from the final model resulted in > 20% change in the remaining variable estimates were chosen. Several interactions were tested, but none improved the overall fit of the model, and thus, none were included in the final multivariable model.
Box-Tidwell tests were used to determine whether the continuous variables in the model met the assumption of linearity.9 Significance and goodness of fit of the final model were evaluated via the likelihood ratio test, Akaike information criterion, and Z statistic.10 Because there was a mean of < 5 observations (dogs/respondent), the hierarchic nature of the data was ignored in the multivariable model.11 All test assumptions were met, and statistical analyses were calculated with statistical software.12,c In the final model, values of P < 0.05 were accepted as significant.
Results
Completed surveys were received from 1,669 handlers from 27 countries including Canada, the United States, Europe, South America, Africa, Asia, Australia, and New Zealand and included data for 3,801 dogs; not all respondents answered every question. In total, 1,209 (31.8%) dogs had incurred an injury, and many of these (334 [28%]) had multiple injuries reported. Most dogs (2,592 [68%]) had not had an injury while participating in agility activities. Additional demographic information and characteristics of injuries incurred were reported elsewhere.3
Injured dogs were significantly (P = 0.02) taller, had more years of agility experience (P < 0.001), and were more frequently described as having participated in warmup (P < 0.001) and cooldown exercises for agility events (P < 0. 001), compared with uninjured dogs (Table 1). The proportion of injured dogs that underwent alternative treatments (eg, acupuncture, massage, and chiropractic care) or were administered dietary supplement products was greater than that of uninjured dogs that received these treatments (P < 0.001 for both). There was no significant difference in weight between groups. Data on age at injury were not evaluated because relevant information was not available for dogs that were uninjured (2,592/3,801 [68.2%]).
Selected characteristics of dogs that did (n = 1,209) or did not (2,592) incur injuries during agility-related activities as reported by 1,669 agility dog handlers from 27 countries in a 2009 Internet-based survey.
| Characteristic | Uninjured | Injured | P value |
|---|---|---|---|
| Weight (kg; mean ± SD) | 17.7 ± 9.3 | 17.9 ± 18.7 | 0.33 |
| Height (cm; mean ± SD) | 47.5 ± 11.9 | 48.3 ± 11.4 | 0.02 |
| Canine agility experience (y; mean ± SD) | 4.2 ± 2.7 | 5.5 ± 2.5 | < 0.001 |
| Handler agility experience (No. [%]) | < 0.001 | ||
| < 5 y | 224 (30) | 513 (70) | NA |
| 5–10 y | 929 (43) | 1,238 (57) | NA |
| > 5 y | 607 (42) | 827 (58) | NA |
| Preventative care (No. [%]) | |||
| Warmup exercise | 2,069 (82) | 1,167 (91) | < 0.001 |
| Cooldown exercise | 1,396 (56) | 899 (70) | < 0.001 |
| Conditioning exercises* | 1,826 (73) | 1,062 (82) | < 0.001 |
| Alternative therapeutic treatments† | 1,392 (55) | 958 (74) | < 0.001 |
| Dietary supplement products | 1,388 (55) | 904 (70) | < 0.001 |
| Breed (No. [% of breed])‡ | |||
| Border Collie | 379 (59) | 260 (41) | < 0.001 |
| Mixed | 305 (71) | 126 (29) | NA |
| Shetland Sheepdog | 260 (72) | 100 (28) | NA |
| Australian Shepherd Dog | 171 (68) | 81 (32) | NA |
| Labrador Retriever | 99 (74) | 34 (26) | NA |
| Golden Retriever | 106 (81) | 25 (19) | NA |
| Cocker Spaniel | 63 (68) | 29 (32) | NA |
| Pembroke Welsh Corgi | 56 (69) | 25 (31) | NA |
| Standard Poodle | 52 (65) | 28 (35) | 0.62 |
| Jack Russell Terrier | 44 (73) | 16 (27) | NA |
Conditioning exercise included aerobic, strengthening, proprioceptive, or balance training.
Alternative therapeutic treatments included acupuncture, massage, and chiropractic care.
P values shown for Border Collie and Standard Poodle breeds indicate comparison with all other breeds combined. Only these breeds were analyzed separately because the proportions of Border Collies and Standard Poodles injured (260/639 [40.7%] and 28/80 [35.0%], respectively) were greater than that of all breeds combined (1,209/3,801 [31.8%]).
NA = Not applicable.
Dogs in the study were of 162 different breeds; data for injured and uninjured dogs of the 10 breeds most highly represented in the survey were summarized (Table 1). The proportions of Border Collies and Standard Poodles injured (260/639 [41%] and 28/80 [35%], respectively) were greater than those of all breeds combined (1,209/3,801 [31.8%]); however, when tested by use of χ2 analyses, the difference was significant (P < 0.001) only for the Border Collie breed. The proportion of Standard Poodles that incurred injury was not significantly different from that of all breeds combined, with (P = 0.62) or without (P = 0.39) Border Collies included in the sample. Given these findings, for our logistic regression models, we categorized injured dogs on the basis of breed as Border Collie or non-Border Collie (Table 2).
Results of univariable (unadjusted ORs) and multivariable (adjusted ORs) analysis of risk factors potentially associated with injury among the same dogs as in Table 1.
| Variable | Injury (n = 940) | No injury (n = 2,464) | Unadjusted OR (95% CI) | Adjusted OR [95% CI]) | P value* |
|---|---|---|---|---|---|
| Previous agility-related injury | 316 | 0 | 132.3 (72.04–278.7) | 100.5 (54.4–212.4) | < 0.001 |
| Preventative care | |||||
| Warmup exercise | 845 | 2,053 | 1.8 (1.4–2.3) | 1.3 (0.9–1.7) | 0.13 |
| Cooldown exercise | 630 | 1,379 | 1.6 (1.4–1.9) | 1.0 (0.8–1.3) | 0.74 |
| Conditioning exercises† | 762 | 1,811 | 1.5 (1.3–1.9) | 1.1 (0.9–1.4) | 0.401 |
| Alternative therapeutic treatments | 671 | 1,366 | 2.0 (1.7–2.4) | 1.5 (1.2–1.8) | < 0.001 |
| Dietary supplement products | 638 | 1,385 | 1.6 (1.4–1.9) | 1.2 (1.0–1.5) | 0.088 |
| Height (cm) | 940 | 2,464 | 1.0 (1.0–1.03) | — | — |
| Weight (kg) | 940 | 2,464 | 1.0 (1.0–1.01) | — | — |
| Region§ | |||||
| Canada | 269 | 600 | Reference | Reference | |
| United States | 565 | 1,580 | 0.8 (0.7–0.9) | 0.8 (0.7–0.9) | 0.124 |
| Europe | 85 | 219 | 0.9 (0.7–1.2) | 0.9 (0.6–1.2) | 0.53 |
| Other | 21 | 65 | 0.7 (0.4–1.2) | 0.7 (0.4–1.2) | 0.746 |
| Canine agility experience (y) | |||||
| Dogs‖ | |||||
| 1–4 | 343 | 1,122 | 1.6 (1.5–1.8) | 1.5 (1.3–1.7) | < 0.001 |
| > 4–10 | 535 | 906 | 0.6 (0.5–0.7) | 0.6 (0.5–0.8) | < 0.001 |
| Handlers | |||||
| < 5 | 133 | 485 | Reference | Reference | |
| 5–10 | 506 | 1,200 | 1.5 (1.2–1.9) | 0.8 (0.6–0.9) | 0.041 |
| > 10 | 301 | 779 | 1.4 (1.1–1.8) | 0.6 (0.4–0.8) | < 0.001 |
| Competitions entered in past year (No. of events/mo) | |||||
| > 1 | 217 | 1,050 | Reference | Reference | |
| 1 | 271 | 740 | 0.9 (0.7–1.02) | 0.9 (0.8–1.2) | 0.635 |
| < 1 | 217 | 674 | 0.7 (0.6–0.9) | 0.9 (0.7–1.1) | 0.404 |
| Frequency of agility practice in past year (No. of times/wk) | |||||
| ≤ 1 | 217 | 681 | Reference | Reference | |
| > 1 | 723 | 1,783 | 1.3 (1.1–1.52) | 1.1 (0.9–1.4) | 0.342 |
| Breed | |||||
| All other breeds | 727 | 2,137 | Reference | Reference | |
| Border Collie | 213 | 327 | 1.9 (1.6–2.3) | 1.7 (1.4–2.2) | < 0.001 |
Data are shown for 3,404 of 3,801 dogs; those with missing data were excluded from analysis.
P values are shown for the final multivariable logistic regression model. In the univariable models, all variables were significant except for height (P = 0.10) and weight (P = 0.44).
Conditioning exercise included aerobic, strengthening, proprioceptive, or balance training.
Alternative therapeutic treatments included acupuncture, massage, and chiropractic care.
Data from handlers living in South America, Africa, Asia, Australia, and New Zealand were classified as other in this analysis.
Data for agility experience among dogs were modeled with restricted cubic splines with 3 knots to account for nonlinearity.9
— = Not applicable (variable was removed from the final model).
All of the variables of interest were significant predictors of injury in univariable logistic regression analyses, except for height (P = 0.10) and weight (P = 0.44). Height (P = 0.606) and weight (P = 0.722) met the assumption of linearity as evaluated via Box-Tidwell tests, but dogs' years of experience in agility competition did not (P < 0.001). This variable was modeled with restricted cubic splines with 3 knots to account for nonlinearity.9 The unadjusted (univariable models) and adjusted (multivariable model) ORs for all variables of interest were summarized (Table 2).
The adjusted ORs from the final multiple regression model indicated that previous agility-related injury, ≤ 4 years of agility experience for dogs, use of alternative therapeutic treatments, and Border Collie breed were significantly associated with increased odds of injury. Dogs having > 4 years of agility experience and handlers having ≥ 5 years of experience were associated with decreased odds of injury in the final multiple regression model (Table 2).
After controlling for all other variables in the model, dogs with a previous agility-related injury were 100.5 times as likely to incur an injury, compared with dogs without such injuries. The relationship between dogs' years of agility experience and the outcome (injury vs no injury) was nonlinear. Use of restricted cubic splines to model these factors and evaluation of plotted data (not shown) indicated that the odds of injury were increased for dogs with the fewest years of experience in the sport (≤ 4 years, [OR, 1.5]). For dogs with > 4 years to 10 years of experience, the odds of injury were decreased (OR, 0.6).
The final adjusted model indicated a protective effect of increased handler experience. The odds of a dog incurring an injury were lower when its handler had 5 to 10 years (OR, 0.8) or > 10 years (OR, 0.6) of experience, compared with the odds for dogs that had handlers with < 5 years of experience in the sport.
Dogs in the United States had lower odds of injury than did dogs in Canada (reference category) in the univariable analysis. However, in the final model, there was no significant difference for dogs from any evaluated region (the United States, Europe, or other [which included South America, Africa, Asia, Australia, and New Zealand]), compared with dogs from Canada.
After controlling for all other variables in the model, the odds of injury for dogs that received alternative therapeutic treatments such as acupuncture, massage, and chiropractic care were higher (OR, 1.5) than those of dogs that did not. The use of warmup and cooldown exercises, participation in some form of conditioning exercise (eg, aerobic, strengthening, or proprioceptive or balance training) and providing dogs with dietary supplement products were not significantly associated with injury in the final model, but their inclusion improved the fit of the final model and so these variables were retained.
Compared with competition in > 1 agility event/mo (the reference category), participating in both 1 event/mo and < 1 event/mo were associated with lower odds of injury in the univariable analyses; however neither of these variables remained significant in the final model. Although practicing > 1 time/wk was associated with higher odds of injury in the univariable analysis, this variable was also not significant when added to the final multivariable model.
After controlling for all other factors, the odds of injury were higher for Border Collies than for dogs of other breeds (OR, 1.7).
Discussion
The primary purpose of the present study was to identify risk factors for agility-related injuries among dogs. The multivariable logistic regression analysis allowed us to consider the effect of each potential risk factor while controlling for others. We identified several variables that were associated with increased odds of injury in the population of dogs studied. Previous agility-related injury; Border Collie breed; use of alternative therapeutic treatments such as acupuncture, massage, and chiropractic care; and fewer years of agility-related experience (< 5 years for handlers and ≤ 4 years for dogs) were significant risk factors for injury.
Studies13–20 in humans and horses have consistently shown that previous injury is strongly predictive of future injuries. This is supported by results of the present study, which revealed significantly greater odds of injury in dogs that had previous agility-related injuries (OR, 100.5), compared with dogs that did not have such injuries.
In our study, Border Collies had greater odds of injury than did dogs of all other breeds. This finding is consistent with the limited literature available.4 The Border Collie breed is more prevalent in the sport than other breeds; 639 of 3,801 (16.8%) dogs in our sample population were Border Collies. Dogs of this breed are known for their athletic stamina and willingness to perform tasks,21 and this may allow handlers to work with Border Collies for longer durations during competition and practice sessions than with other breeds. We included 3 exposure variables in our multivariable model: dogs' years of experience in agility competition, number of agility events per month entered in the past year, and amount of practice per week. After controlling for these (and other predictor variables), Border Collies still had 1.7 times the odds of injury, compared with other breeds. Another possible explanation is that the speed at which dogs navigate an agility course may potentially be related to the increased risk of injury. Border Collies are known for drive, speed, and quickness in changing directions, compared with other breeds.21 They were originally bred to herd livestock, which requires great athletic prowess and stamina. Studies22–24 examining equine steeplechase, hurdle, and flat races have demonstrated that high rates of speed are associated with increased risk of injuries to the horses. Future research examining the biomechanical variables associated with agility-related activities may help to shed light on the observed differential risk for injuries among breeds.
The relationship between injury and the number of years of agility experience for dogs was nonlinear. Our results indicated that the odds of injury were increased for dogs with the least amount of experience in the sport (≤ 4 years; OR, 1.5). Interestingly, for dogs with > 4 years of experience, there was actually a decrease in the odds of sustaining injury (OR, 0.6). This measure could be an indication of skill acquisition. Helton25 suggested in an examination of skill automaticity and expertise in agility dogs that, with increased deliberate practice for agility events, dogs became more accurate and faster on course. These changes accompanying deliberate practice may include safer obstacle performance and better decision making; thus, more experienced dogs may put themselves at lower risk for injury. This increase in expertise with experience may be related to the changes in odds of injury over time.
Dog experience may also be a proxy for age, which we could not include in our model. In the present study, 2 additional exposure measures, amount of practice per week and number of competitions per month, were not significant in the multivariable analysis. This may have resulted from the imprecision of our survey instrument for these variables. Future prospective epidemiological studies should include age to better examine the dynamics among age, experience, exposure, and injury in dogs that participate in agility-related activities, because increasing age has been shown to have a greater influence than experience on injury related to sports in other species (eg, in equine jumping26,27 and various human sports).13,28
An interesting relationship was found between handler experience and the odds of agility-related injury in dogs. In the unadjusted analysis, longer participation in the sport by handlers (5 to 10 or > 10 years, compared with < 5 years) was associated with an increased risk of injury in dogs. Intuitively, it would make sense that having a longer career in the sport would increase the likelihood that a handler would, at some point in time, have a dog incur an agility-related injury. However, when previous agility-related injury was entered into the multivariable model, the direction of effect reversed, and the odds of injury were significantly decreased for dogs handled by individuals with 5 to 10 years (OR, 0.8) or > 10 years (OR, 0.6) of experience in the sport. Inexperience has been documented as a risk factor for injury in many situations, including equestrian rider injuries,29 workplace injuries,30 and automobile accidents.31
Use of alternative therapeutic treatments as preventive measures to keep dogs fit for agility was significantly associated with increased odds of injury (OR, 1.5). Additionally, while not significant in the final model, participation in conditioning activities including aerobic, strengthening, proprioceptive, or balance training and the administration of dietary supplement products appeared to be associated with increased odds of injury. It may be that these actions were implemented after an injury had already occurred in the dogs of this report or in the handler's experience with other dogs. Handlers typically become more aware of the possibility for future injury after such an incident has occurred and may take a more proactive role in managing the health of dogs in their care on an ongoing basis from that point forward.
There was no relationship found between the use of warmup and cooldown exercises and injury in our study. It is a commonly held belief among individuals in the canine agility community that these are important factors for reducing the risk of injury in dogs.32 In contrast, evidence from the literature on the effects of warmup and cooldown exercises remains inconclusive. A recent systematic review33 of injuries in human athletes found equivocal results regarding whether these factors can modify the risk of injury.
Compared with dogs living in Canada, dogs in the United States had slightly lower odds of injury (although this was not significant in the final model). There was no significant difference observed for regions in Europe and the rest of the world when compared with Canada. There are differences in equipment standards or in styles of course design that exist among these regions. A more rigorous study design (eg, a prospective cohort or case-control study) would be necessary to investigate potential effects of this variable further.
The present study was not without limitations. Most notably, recall bias may have been a factor because data were collected retrospectively and reported by handlers and as such could not be easily verified. In addition, our sampling strategy may have led to selection bias. Handlers had to be able to communicate in English and have access to the Internet to participate. Additionally, it is possible that handlers with a greater interest or personal experience with dogs injured in agility-related activities were more likely to complete the survey. Nevertheless, our sample included a large number dogs and handlers from various regions worldwide, and in fact, there were a greater number of uninjured dogs than injured dogs, making this issue less of a concern.
Previous studies of agility-related injury risk factors have been smaller in scale and have included simple descriptive analyses.3,4 We believe that the present study has several advantages over these studies. First, we obtained data for large numbers of uninjured and injured dogs and we were able to use multiple logistic regression analyses to estimate the effect of individual risk factors while controlling for other variables in the model. This led to different interpretations of the data than the simple univariate analyses alone would have, especially in regard to the protective effect that emerged for increased handler experience when the model was controlled for previous agility-related injury. We also were able to recruit participants from 27 countries worldwide, which improved the generalizability of our findings.
The results of the study reported here have important implications to guide future research and prevention activities aimed at reducing agility-related injuries in dogs. We recommend that agility organizations, instructors, judges, and veterinary practitioners take measures to educate agility dog handlers about the risks for injury in the sport. This effort should target handlers that are new to the sport or working with inexperienced dogs and those that work with Border Collies, given the higher odds of injury associated with these factors. We also believe that agility associations should consider implementing more comprehensive injury surveillance systems. The primary goal of such systems would be to collect injury and activity data from a representative sample of canine athletes in various geographic regions. Relevant data could then be shared with the research community and appropriate policy committees to provide a foundation for evidence-based decision making with regard to health and safety issues.
ABBREVIATION
| CI | Confidence interval |
Agility Association of Canada, North Gower, ON, Canada: unpublished data, 2011.
Copies of the questionnaire are available from the corresponding author on request.
R, version 2.14.0, R Foundation for Statistical Computing, Vienna, Austria. Available at: www.r-project.org/. Accessed Nov 8, 2011.
References
1. United States Dog Agility Association. USDAA—dog agility. 2011. Available at: www.usdaa.com/se_agility.cfm. Accessed Nov 8, 2011.
2. American Kennel Club. American Kennel Club—annual statistics. 2011. Available at: www.akc.org/events/agility/statistics.cfm. Accessed Nov 3, 2011.
3. Cullen KL, Dickey JP, Bent LR, et al. Internet-based survey of the nature and perceived causes of injury to dogs participating in agility training and competition events. J Am Vet Med Assoc 2013; 242: 1010–1018.
4. Levy M, Hall C, Trentacosta N, et al. A preliminary retrospective survey of injuries occurring in dogs participating in canine agility. Vet Comp Orthop Traumatol 2009; 22: 321–324.
5. Norman GR, Streiner DL. Logistic regression. In: Biostatistics: the bare essentials. Shelton, Conn: PMPH-USA Ltd, 2008; 159–166.
6. Susan Garrett Agility Training. Say yes! dog training—by Susan Garrett. Available at: www.susangarrettdogagility.com/. Accessed Mar 16, 2009.
7. Hosmer DW & Lemeshow S. Applied logistic regression. New York: John Wiley & Sons, 2000.
8. Bursac Z, Gauss CH, Williams DK, et al. Purposeful selection of variables in logistic regression. Source Code Biol Med 2008; 3: 17.
9. Harrell FE. Regression modeling strategies: with applications to linear models, logistic regression, and survival analysis. New York: Springer Publishing, 2001.
10. Hosmer DW, Hosmer T, Le Cessie S, et al. A comparison of goodness-of-fit tests for the logistic regression model. Stat Med 1997; 16: 965–980.
11. Clarke P. When can group level clustering be ignored? Multilevel models versus single-level models with sparse data. J Epidemiol Community Health 2008; 62: 752–758.
12. R Development Core Team. The R manuals. Vienna: R Foundation for Statistical Computing; 2011. Available at: cran.r-project.org/manuals.html. Accessed Nov 8, 2011.
13. Emery CA, Meeuwisse WH. Risk factors for groin injuries in hockey. Med Sci Sports Exerc 2001; 33: 1423–1433.
14. Emery CA, Meeuwisse WH, Hartmann SE. Evaluation of risk factors for injury in adolescent soccer. Am J Sports Med 2005; 33: 1882–1891.
15. Hägglund M, Waldén M & Ekstrand J. Previous injury as a risk factor for injury in elite football: a prospective study over two consecutive seasons. Br J Sports Med 2006; 40: 767–772.
16. Dyson S. Lameness and poor performance in the sports horse: dressage, show jumping and horse trials (eventing), in Proceedings. 46th Annu Meet Am Assoc Equine Pract 2000; 308–315.
17. Bahr R & Krosshaug T. Understanding injury mechanisms: a key component of preventing injuries in sport. Br J Sports Med 2005; 39: 324–329.
18. Dowling BA, Dart AJ, Hodgson DR, et al. Superficial digital flexor tendonitis in the horse. Equine Vet J 2000; 32: 369–378.
19. Gibson KT, Burbidge HM, Pfeiffer DU. Superficial digital flexor tendonitis in Thoroughbred race horses: outcome following non-surgical treatment and superior check desmotomy. Aust Vet J 1997; 75: 631–635.
20. Perkins NR, Reid SW, Morris RS. Risk factors for injury to the superficial digital flexor tendon and suspensory apparatus in Thoroughbred racehorses in New Zealand. N Z Vet J 2005; 53: 184–192.
21. American Kennel Club. The complete dog book. New York: Random House Digital Inc, 2006.
22. Verheyen K, Price J, Lanyon L, et al. Exercise distance and speed affect the risk of fracture in racehorses. Bone 2006; 39: 1322–1330.
23. Pinchbeck GL, Clegg PD, Proudman CJ, et al. Case-control study to investigate risk factors for horse falls in hurdle racing in England and Wales. Vet Rec 2003; 152: 583–587.
24. Pinchbeck GL, Clegg PD, Proudman CJ, et al. Horse injuries and racing practices in National Hunt racehorses in the UK: the results of a prospective cohort study. Vet J 2004; 167: 45–52.
25. Helton WS. Skill in expert dogs. J Exp Psychol Appl 2007; 13: 171–178.
26. Bailey CJ, Reid SWJ, Hodgson DR, et al. Flat, hurdle and steeple racing: risk factors for musculoskeletal injury. Equine Vet J 1998; 30: 498–503.
27. Williams RB, Harkins LS, Hammond CJ, et al. Racehorse injuries, clinical problems and fatalities recorded on British racecourses from flat racing and National Hunt racing during 1996, 1997 and 1998. Equine Vet J 2001; 33: 478–486.
28. Van Mechelen W, Twisk JWR, Molendijk AJ, et al. Subject-related risk factors for sports injuries: a 1-yr prospective study in young adults. Med Sci Sports Exerc 1996; 28: 1171–1179.
29. Hitchens PL, Blizzard CL, Jones G, et al. Predictors of race-day jockey falls in flat racing in Australia. Occup Environ Med 2010; 67: 693–698.
30. Breslin FC & Smith P. Trial by fire: a multivariable examination of the relation between job tenure and work injuries. Occup Environ Med 2006; 63: 27–32.
31. Zhang J, Fraser S, Lindsay J, et al. Age-specific patterns of factors related to fatal motor vehicle traffic crashes: focus on young and elderly drivers. Public Health 1998; 112: 289–295.
32. Steiss JE. Muscle disorders and rehabilitation in canine athletes. Vet Clin North Am Small Anim Pract 2002; 32: 267–285.
33. Fradkin AJ, Gabbe BJ, Cameron PA. Does warming up prevent injury in sport? The evidence from randomised controlled trials? J Sci Med Sport 2006; 9: 214–220.
