Injury, Illness, and Training Load in a Professional Contemporary Dance Company: A Prospective Study

Active professional contemporary dancers from the same dance company were eligible to participate in the study on a voluntary basis. Sixteen dancers, 7 men and 9 women (see baseline characteristics in Table 1), agreed to participate in this study. Ethical approval was granted by the University of Technology Sydney Human Research Ethics Committee (HREC ETH17-1082) before the study began. Each participant provided written informed consent and was familiarized with all procedures before the start of the study.

We used a prospective cohort research design in which injury, illness, and TL data were collected during a 1-year period (2018–2019) in Sydney, Australia. The study location was the dance company's training base or domestic and international performing venues. All injuries sustained during the surveillance period were recorded by the company physiotherapist, and no participants' data were excluded from the analysis. Injury reporting and methods were based on guidelines from the International Association of Dance Medicine and Science¹¹  and international consensus statements.^12,13  The reporting of this study follows the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) statement.¹³ 

Injuries were defined as a (1) medical-attention injury (Med-Inj): assessment of a participant's medical condition by a qualified physiotherapist¹²  or (2) time-loss injury (Time-Inj): a participant was unable to dance for 1 or more days after the event.¹⁴  Information was collected regarding injury status (new, no previous history; reinjury, repeat episode of a fully recovered injury; exacerbation, worsening of a nonrecovered injury),¹²  mechanism of injury, location, injury occurrence, type of onset,¹²  report of cause, contact or noncontact, detailed injury assessment, and exposure status.

Injuries were classified into 3 categories of severity according to the length of absence from full training, with the day on which the injury occurred being counted as day zero: minor (1–7 days), moderate (8–28 days), and severe (>28 days). Dance exposure was defined as the time during which individuals were at risk of injury (hours spent participating in dance class, rehearsal, or performance). Injury incidence rate was calculated as all new Med-Injs per 1000 h and all new Time-Injs per 1000 hours. The Med-Inj and Time-Inj incidence rates were calculated for each training mode. Additionally, the Med-Inj and Time-Inj incidence rates were calculated for group and individual daily and weekly TL percentiles (low, medium, and high).

The Wisconsin Upper Respiratory Tract Infection Survey contains 10 items that assess symptoms, 9 items that assess functional impairments, and 1 item that assesses global severity and global change. For a URTI to be recorded, participants must have had upper respiratory signs and symptoms for ≥48 hours. For calculation of the illness incidence rate, participants were considered at risk of a new illness during dance training exposure minus both the duration of each illness episode and the 7 days after each episode. Illness incidence rates were also calculated based on group and individual daily and weekly TL percentiles (low, medium, and high).

Sessions were classified by the following training modes: ballet and contemporary class, rehearsals, and performance. The internal TL for each session was calculated using the session rating of perceived exertion (sRPE) method for each dancer during the study period.¹⁵  The sRPE was calculated as the product of exercise intensity and exercise duration. Exercise intensity was determined by the sRPE according to the Borg CR-10 scale¹⁶  and was collected within 10 minutes after each class. The sRPE method has been validated¹⁷  as a measure of internal TL in preprofessional dancers. In addition, the sum of the daily and weekly TLs was calculated for individual and group percentiles (low, medium, and high) and each type of session (ballet, contemporary, rehearsal, and performance).

All injury, illness, and TL data were prospectively collected and recorded in a centralized database (Edge10, London, UK) using a custom-designed app. Each injury was assigned a 4-character injury diagnosis code (Orchard Sports Injury Classification System 10.1).¹⁸  Injuries were time recorded according to the date of injury, and all data were linked by a unique identification code.¹⁹  This study was part of a larger 2-year project involving data from the surveillance system.

Descriptive analyses were performed on injuries and illnesses sustained during the 2018 surveillance period. We calculated injury and illness incidence rates using the following formula: number of new injuries per exposure in hours (individual hours collected from duration time calculated in sRPE) × 1000 and 95% confidence intervals (CIs). Kaplan-Meier injury survival analysis was conducted to determine the proportion of injury-free participants at monthly intervals for both Med-Injs and Time-Injs. All analyses were performed in Excel (version 2016; Microsoft Corp, Redmond, WA).

A STROBE flowchart of participants illustrates the sequential recruitment (Figure 1).

Sixteen professional contemporary dancers (7 men, 9 women) were followed during a 1-year period that involved 14 689 participation hours. A total of 79 injuries were reported during the 1-year period (new = 86.1% [n = 68]; reinjury = 6.3% [n = 5]; exacerbation = 7.6% [n = 6]) by the 16 particpants (100%), with a Med-Inj incidence rate of 4.6 per 1000 hours (95% CI = 3.8, 5.8) and a Time-Inj incidence rate of 1.4 per 1000 hours (95% CI = 0.8, 2.1). A detailed summary of the injury distribution characteristics and incidence rates is presented in Table 2. Most of the reported injuries (70.6%, n = 59) resulted in no time loss from dance training or performance. The Med-Inj incidence rate for each session was as follows: ballet = 4.4 per 1000 hours (95% CI = 1.9, 9.1), contemporary = 4.7 per 1000 hours (95% CI = 2.3, 9.2), rehearsal = 4.9 per 1000 hours (95% CI = 3.6, 6.5), and performance = 3.1 per 1000 hours (95% CI = 1.1, 8.1). The greater risk difference was between performance and contemporary sessions: 2.1 (95% CI = −2.6, 6.7; P = .58). The Time-Inj incidence rate for each session was ballet = 0.0 per 1000 hours, contemporary = 1.2 per 1000 hours (95% CI = 0.3, 4.6), rehearsal = 1.4 per 1000 hours (95% CI = 0.8, 2.2), and performance = 2.1 per 1000 hours (95% CI = 0.7, 6.5).

The majority of Med-Injs were to the knee (16.5%, n = 13), upper leg (15.2%, n = 12), and torso (15.2%, n = 12), which resulted in median time losses of 13 days (range = 4–90), 3.5 days (range = 1–10), and 6.5 days (range = 2–14), respectively. The most common type of injury (69.6%, n = 55) was ligament-joint and muscle-tendon (26.6%, n = 21), with median time losses of 4 days and 6 days (Table 3), respectively. Acute injuries were the most frequent type of injury, accounting for 74.7% (n = 59), with gradual injuries responsible for only 25.3% (n = 20).

The majority of Time-Injs were to the ankle (25%, n = 5), upper leg (20%, n = 4), and torso (20%, n = 4), which resulted in median time losses of 21 days (range = 5–28), 5 days (range = 1–10), and 5 days (range = 2–14), respectively. Ligaments and joints were injured most often (65%, n = 13). Acute injuries were the most common type, accounting for 90% (n = 18), with gradual injuries accounting for only 10% (n = 2).

Time-loss injuries constituted 29.4% (n = 20) of all injuries, resulting in 261 days away from dance. Sixty percent (n = 12) of all Time-Injs required the participant to take ≤7 days off dance; 35% took 8 to 28 days off and 5% took >28 days off (Figure 2). Injuries that necessitated the greatest time away from dance were meniscal injuries, ankle sprains, and muscle strains.

The starting point for the survival analysis was the beginning of the study. Participants were followed until the first new Med-Inj and Time-Inj occurred. We calculated Kaplan-Meier curves to estimate the survival function for the time to the first injury in all injured participants (Figure 3). All participants were injury free at the onset of the study.

A total of 134 illness episodes were reported across the 1-year surveillance period, with no missing data. The illness incidence rate was 9.1 per 1000 hours (95% CI = 7.7, 10.7). The number of days participants experienced URTI symptoms was 39.9 ± 26.9 (mean ± standard deviation [SD]; range = 1–96).

The average weekly TL was 6685 ± 1605 (range = 4641–10 391) arbitrary units (AU). The average TLs for individual sessions were ballet = 407 ± 120 (range = 70–1200) AU, contemporary = 387 ± 153 (range = 30–1620) AU, rehearsal = 578 ± 303 (range = 20–1800) AU, and performance = 691 ± 350 (range = 40–1800) AU. The group average RPE was highest in performance (8.0 ± 1.4) and lowest for contemporary (5.0 ± 1.4) as shown in Appendix A.

For the analysis of the effect of TL on injury, complete data were collected for injury and less than 10% of data were missing for TL. The missing data for TL were all from 1 participant (participant No. S3, injuries = 4). The main analysis used single imputation (mean TL imputed for missing TL data), and the results are provided in Table 4. Sensitivity analysis was conducted as follows: (1) nonimputed data (n = 16); (2) imputation of maximum TL observed on that day or week from the remaining participants entered as the missing value for dancer S3; (3) imputation of minimum TL observed on that day or week from the remaining participants entered as the missing value for participant No. S3; and (4) nonimputed data excluding participant No. S3 (n = 15). The results of the sensitivity analysis were similar and did not change the interpretation of the data (data not shown). This suggests that the distribution of missing data is consistent with the assumption that the data are missing at random.

Based on group TL percentiles, the Med-Inj incidence in the high TL group was 5.7 per 1000 hours; medium, 3.6 per 1000 hours (P = .08); and low, 3.2 per 1000 hours (P = .17). Based on participants' individual TL percentiles, the greater daily Med-Inj incidence rate was in the low TL group (6.6 per 1000 hours, 95% CI = 4.4, 9.7), and the greater weekly injury incidence rate was in the medium group (5.9 per 1000 hours, 95% CI = 4.2, 8.3) as shown in Table 4. Additionally, we calculated the injury incidence rates for Med-Injs and Time-Injs during each type of session based on participants' individual TL percentiles. For Med-Injs, ballet in the low TL percentile group had the greater injury rate at 2.7 per 1000 hours (95% CI = 1.0, 7.31) and the smallest injury incidence rate of 0.0 per 1000 hours in the medium TL group. For Time-Injs, performance had the greater incidence rate for all 3 categories of TL at 0.7 per 1000 h (95% CI = 0.09, 4.9) as displayed in Figure 4.

The number of days participants experienced URTI symptoms in the high TL group was 52.2 ± 26.6 (range = 19–96), in the medium TL group was 47.8 ± 19.7 (range = 18–69), and in the low TL group was 19.6 ± 21.3 (range = 1–60). The illness rates were 3.9 per 1000 hours (95% CI = 2.9, 5.0) for the high TL group, 3.4 per 1000 hours (95% CI = 2.5, 4.4) for the medium TL group, and 1.8 per 1000 hours (95% CI = 1.2, 2.6) for the low TL group.

The purpose of our study was to provide a detailed description of injury, illness, and TL in professional contemporary dancers. We are the first to comprehensively examine the combination of injury, illness, and TL in the same population over an extended, continuous surveillance period. The key findings of our investigation were (1) high injury rates, particularly Time-Injs, (2) high illness rates and lengthy durations of URTI symptoms, (3) a median time to injury of 3 months, and (4) injury and illness rates in relation to TL depended on the categorization and reference method. Although a number of previous researchers^1,2,20  explored injury rates in professional dancers, only 1 group¹⁰  concurrently assessed injury and TL. However, that study lasted 7 weeks, and no time-loss or medical-attention injuries were reported.¹⁰  Additionally, no authors have examined illness in dancers.

Our entire cohort of participants (100%) acquired at least 2 Med-Injs during the surveillance period, and several participants experienced >7 new injuries. The Med-Inj incidence rate (4.6 per 1000 hours) was similar to that reported in a prospective epidemiologic study of professional ballet dancers (4.44 per 1000 hours).³  However, the Time-Inj incidence rate was higher (1.4 per 1000 hours) than in several evaluations of similar contemporary dancers (0.16–0.22 per 1000 hours).^1,14  Several factors may have contributed to these differences, including the repertoire, demographics, and study durations. For example, one investigation¹  spanned a 15-year period, compared with our 1-year duration. In addition, our injury incidence rates for individual sessions ranged from 4.9 per 1000 hours for rehearsal to 3.1 per 1000 hours for performance. In contrast, a previous group³  who studied professional ballet dancers reported lower injury incidence rates for rehearsal (2.43–2.99 per 1000 hours) and higher rates for performance (4.45–5.19 per 1000 hours). This difference might be due to variations in dance style and the intensity of rehearsal sessions. We reported 261 days for Time-Injs and 447 days of training (in a modifed form) for Med-Injs. Furthermore, 4 dancers missed performances due to new injuries and 1 due to reinjury during the surveillance period. This shows that dancers persisted in training and performance despite the presence of injury or a medical complaint.

In the current study, ligaments and joints were injured most commonly, predominantly in the lower extremities, with hip impingement accounting for the majority. In contrast, ligament and joint injuries accounted for only 35% of all injury types in a study of a similar cohort.¹⁴  Additionally, despite overuse injuries being cited as the most frequent type of injury in dancers,²  we found that overuse injuries were responsible for only 25.3%. This variance may be due in part to differences in definitions of overuse injuries. For example, no current consensus exists on the definition of overuse injury; some researchers²¹  considered overuse a mechanism of injury, whereas others²⁰  used a diagnosis-based definition. Overall, we identified a wide range of injury types that should be considered when injury-prevention programs are being developed or casual associations are being examined in future etiologic studies.

This is the first study, to our knowledge, to examine the time-to-event analysis of injury in professional contemporary dancers. The Kaplan-Meier analysis involves calculation of the probability of injury at the time of occurence. We determined that the median survival time without injury for both Med-Inj and Time-Inj was 3 months. Within the first 5 months of the surveillance period, all participants except 1 experienced their first Med-Inj. Furthermore, the most severe injuries (meniscal tear requiring surgical repair, grade 3 lateral ankle sprain) occurred in the first quarter of the year. Based on limited numbers, these results suggest that dance is a very high-risk discipline, underlining the need for further research to support the implementation of injury-prevention programs to protect dancers' health.

Currently no investigations of illness incidence (in particular, URTI) have been conducted in professional dancers. Upper respiratory tract illness has been reported²²  as the most common reason for noninjury-related presentations to sports medicine clinics, accounting for 35% to 65% of illnesses. A major finding of our work was the high incidence of URTI, with an illness incidence rate of 9.1 per 1000 hours and 134 illness episodes. Additionally, the average duration of URTI symptoms was 39.9 ± 26.9 days, ranging from 1 day to approximately 3 months. Potential causes for high illness rates include TL management,^4,6,7  international travel,²³  and lifestyle factors.²⁴  However, these measures were beyond the scope of our study; further research is required to understand if these factors influence the reduction of pathogen exposure and potential immunosuppression. It should be noted that no modifications to training or time loss were reported while the dancers experienced URTIs. As such, persistent illness can have a negative effect on health and performance, particularly when undertaking high levels of strenuous exercise.²² 

As expected, we showed that professional contemporary dancers undertook high TLs, especially in comparison with many other high-level athletes.^25,26  For example, weekly accumulated TLs in elite soccer were 2994 AU²⁶  and average TLs in Australian football were 4261 AU.²⁵  Also, the average weekly accumulated TLs (6685 ± 1605 AU) were much higher than those reported in preprofessional contemporary dancers (4283 ± 2442 AU)¹⁷  and adolescent ballet dancers (596 ± 153 AU).²⁷  Collectively, these variations may reflect the increased training duration of professional dancers compared with preprofessional dancers and football players. Although the average TLs for individual sessions (contemporary, ballet, and rehearsal) were similar to those in previous reports of preprofessionals,¹⁷  our dancers undertook more sessions per day.

In recent years, many authors^9,28,29  have examined the association between TL and injury in various sport disciplines. Yet most of these studies were underpowered because the number of injuries required to indicate an association with TL was much higher than typically used.³⁰  The number of injury events we described was similar to those in previous publications but still inadequate to reliably determine associations; hence, we did not perform any predictive or associative analyses.^28,29  Rather, we accurately described the TL to help develop a conceptual framework and a causal structure that can be used in future epidemiologic studies that may require international and multicenter initiatives to obtain appropriate sample sizes.

An interesting finding of our investigation was that participants in the high TL group experienced the highest injury incidence rate (5.7 per 1000 hours) compared with the medium and low groups (3.6 and 3.2, respectively, per 1000 hours). This result is consistent with previous findings in a variety of sports^31,32  but in contrast with researchers^25,33  who showed the opposite (lower injury risk corresponding to higher TL). These conflicting results may be due to a variety of methodologic factors, including low sample sizes and different injury definitions and statistical approaches. However, we also demonstrated that the injury incidence rate in relation to the TL depended on the calculation method used. Most previous authors have used group data to create TL categories. For example, a study participant was placed in the high, medium, or low category based on the TL completed by the whole group. Nevertheless, no rationale exists for using the group TL as a reference. Given the individual nature of injuries, it would be reasonable to use individual data for categorization: that is, placing the individuals in the high, medium, or low category based on their TL range. To our knowledge, only 1 group³⁴  categorized TL using individual data (from cricket players). Interestingly, when we used individual TLs, the Med-Inj rate was opposite in direction to the group TL injury rate, with dancers in the highest TL category experiencing lower injury rates. Therefore, a clear association of TL level and injury rate was not present, which is consistent with earlier findings of both higher and lower injury risks associated with a high TL.²⁶  Differences in categorization methods, low sample size, and sparse data bias may explain the inconsistencies among these works⁹  and support the need for larger, hypothesis-driven studies to avoid results that simply reflect “statistical noise” and potentially false discoveries.

High TL has been proposed as a risk factor for illness in athletes.^4–6  Changes in internal and external TL have been associated with an increased risk of illness.⁸  For example, increases in training volume have been related to an increased risk of illness in elite swimmers⁴  and elite junior tennis players.⁵  Additionally, in team sports, increased training intensity preceded the development of URTI symptoms.⁶  In our study, participants in the high and medium TL groups had the highest illness incidence rates (3.9 and 3.4, respectively, per 1000 hours) compared with the low TL group (1.8 per 1000 hours). The high TL group also experienced the longest duration of URTI symptoms compared with the medium and low groups. Still, it is difficult to compare these results because of possible variations in methods, participants' characteristics, and travel schedules. Also, many factors other than exercise affect the immune system. For example, poor nutrition, smoking, alcohol consumption, mental stress, and lack of sleep have all been associated with the impaired immune function and increased risk of infection.³⁵  Regardless of the underlying cause of the URTI, a major concern, other than the health effect, is the accompanying fatigue that can limit or prevent training.³⁶  Future researchers should adopt multifactorial preventive strategies that may include clinical, training, or lifestyle modifications.²⁴ 

Similar to our injury results, when individual TLs were used, the illness rate was in the opposite direction to the group TL illness rate, with dancers in the highest TL category experiencing lower illness rates. Therefore, depending on the categorization method, the findings in relation to TL will be different. Without a strong rationale or conceptual framework for deciding the best way to categorize TL, no conclusion can be drawn from the current study. Thus, future investigation is warranted to develop a conceptual framework that may explain particular causal phenomena and provide a basis for interpreting results.

This study had several limitations. The sample size was restricted to the number of participants in the dance company. Accordingly, our goal was descriptive, and associations between TL and illness and injury were explored to show the dependence of the results on the categorization criteria. Although previous authors^28,29  have used similar sample sizes and numbers of events, we believed this would provide unreliable results and would not satisfy the requirements for such a study⁹  that should also take into account the multifactorial nature of injuries and illnesses. However, these results increase the body of knowledge with which to create causal structures that can be verified in the future.

The main aim of our study was to contribute to the body of knowledge used to identify potential mechanisms and prognostic factors. Yet from a practical point of view, our results suggest that a possible area of intervention is TL distribution by applying training principles. Indeed, few recovery periods occurred within and between macrocycles (weeks) as shown in Appendixes B and C (mean daily and weekly TLs). Considering that professional dancers may need to undertake high TLs to achieve optimal performance, modifying the TL distribution and including recovery days and short periods of reduced load are potential immediate interventions. Furthermore, understanding the TL distribution may require monitoring strategies,³⁷  as well as injury and illness surveillance systems. Dancers' stoic natures and strong work ethics (eg, the desire to continue to work despite illness) may mean that practitioners need to carefully monitor the illness incidence and actively modify TLs accordingly.

Our aim was to provide details on the nature and incidence of injury and preliminary information on illness occurrence. These results together with the TLs experienced by the dancers can be used to develop and hypothesize an etiologic framework and hence a causal structure (eg, directed acyclic graph).³⁸  This is necessary if the goal of future observational studies, as is common in sport and clinical settings, is to manipulate a variable to alter the likelihood of an event (eg, injury or illness). Although causal inference from observational studies inevitably does not ensure causation (ie, experimental studies are needed), it would provide some support to a causal link and, if strong enough, may support interventions. The variety of injury types and natures in this cohort also indicates that even though the final mechanical causes are likely similar (stress and strain superior to the strength of the structure), the antecedent causes may be different. Combining all injuries to determine associations with TL is not a reasonable approach because it would assume the same causal path for all injuries experienced by the dancers. No relation between TL and injury and illnesses could be demonstrated, but this association is worth exploring using appropriate statistical methods and sample sizes. Therefore, other than a conceptual framework to develop a causal structure, multicenter studies are necessary. Furthermore, prediction models for risk stratification (that do not require causation) are also warranted.

Professional dancers experienced high TLs relative to other athletes and concomitantly high injury and illness incidences and risks. Interestingly, these dancers missed very few performances, despite injury and illness. In addition, they continued training, albeit with modifications, even when affected by Med-Inj or illness. Collectively, this suggests that dancers persist with training and performance despite injury and illness, which may reflect their high level of commitment. However, the health of dancers should be a priority and the development of preventive interventions and educational initiatives encouraged.

We thank Edge10 for their generous support of this research project.