Risk Factors for Shoulder Injuries in Water Polo: a Cohort Study

INTRODUCTION

Water polo is a popular aquatic contact sport, and has the highest rates of injuries amongst other aquatic disciplines during competitions (16.2% to 19.4%).¹ Although the majority of observed traumatic injury incidence occur to the head and fingers during matches,² the most common overuse injury area is the shoulder.^3–5 In order to decrease shoulder injuries in water polo, a better understanding of their risk factors is necessary to target prevention measures.⁶

Lack of strength may be related to a higher risk of injury in this sport,³ as well as deficits in external rotation (ER) strength relative to internal rotation (IR) strength.⁷ Previous authors have shown that water polo players are stronger than healthy non-players in abduction, adduction, ER and IR (p<0.05)^8,9 and showed lower ratios of ER over IR rotation strength.^9,10 Recently, Hams et al¹¹ have shown that sub-elite players (national development group) who were weaker in isometric ER and IR at baseline testing were more likely to have new shoulder injuries occur over the following three seasons. Furthermore, no significant difference was found between injured and non-injured groups for ER to IR strength (ER:IR) ratios. However, as Hams et al. performed isometric tests, testing shoulder strength with isokinetic devices at higher speeds may replicate the muscle activity which occurs during the throwing action and may yield different findings.¹²

Lack of shoulder range of motion (ROM) has been shown to correlate strongly with shoulder injuries in swimming and overhead throwing sports.^13–16 Water polo players show greater ER, decreased IR, and increased total range of motion in their dominant shoulders compared to their contralateral side.^8,17 However, Elliott¹⁸ found no statistical correlations between shoulder pain and ROM in a group of 13 male national team water polo players. In contrast, Hams et al¹¹ found that players in the injured group showed significantly less total range of motion (ER plus IR) (p<0.05). Thus, more evidence is needed to correlate injuries with ROM measures of the shoulder in water polo players.

Altered scapular posture is related to shoulder pain in throwing sports,^19,20 and it is hypothesized that the “head up” swimming pattern typical during water polo can also lead to impingement syndromes.³ One group of authors found no differences between water polo players and healthy controls in scapular upward rotation (UR) using electromagnetic 3D kinematic measurements (frontal plane angle of the spine of the scapula vs a horizontal line).²¹ Two-dimensional measurements of UR have also shown good to excellent reliability,²² and have been implemented by other authors to assess water polo players. Mukhtyar et al²³ compared the scapular abduction position of healthy water polo players (n=16) to players with impingement symptoms (n=14) by measuring the distance between scapular angles and the spine after training. The group with shoulder impingement showed significantly decreased values for scapular abduction and UR (p˂0.05) at 45° or more of shoulder abduction.²³ However, Witwer et al⁸ did not observe these patterns of decreased upward rotation in a cohort of 31 collegiate water polo players (12 males and 19 females) in a rested state.

Previous researchers have investigated strength, ROM,^11,17 scapular alignment,²³ throwing variables,^24,25 and shooting volume²⁶ as potential risk factors for shoulder injuries. However, only one investigation was performed prospectively on sub-elite players, and none in other age groups. Therefore, the causal relationship between injuries and these variables remains unclear. Strength and ROM were the only variables measured in relation to shoulder injury incidence. Additional understanding of risk factors is necessary to inform effective injury prevention strategies in this sport. Therefore, the purpose of this study was to estimate whether previous injury, changes in strength, ROM or UR are related to shoulder injuries in water polo players. A secondary objective was to compare sex differences among these risk factors. Given previous findings, it was expected that weaker players with less ROM and less upward rotation of the scapulae would be at higher risk of injuries.

METHODS

Subjects

Nineteen male and twenty female water polo players from the Canadian senior national team were selected for this cohort study. Participants had to have a minimum of five years of experience, and be training full-time in a high-level competition environment (at least five practices per week). Subjects with a history of shoulder injury or surgery were included if they were able to participate fully in all team training sessions at the beginning of the study. A formal sample size calculation was not performed because all members of the senior national teams in Canada were recruited (n=39). Further recruitment would have required the addition of lower level players that did not represent the target population. Data were collected at the training center at the Institut National du Sport du Québec in Montreal, Canada. This study received ethics approval from McGill University Ethics Institutional Review Board, in compliance with the Helsinki Declaration. All participants signed informed consent to take part in the study.

Procedures

Demographic data were collected for age, body mass index (BMI), hand dominance, player position and training setting. Shoulder passive ROM was assessed in ER and IR using a standard goniometer. Shoulder strength was assessed with an isokinetic device for ER and IR. Scapular UR was assessed with a digital inclinometer.

Range of Motion

Participants were positioned in supine, with the shoulder in 90° of flexion and abduction (Figure 1). A small lift was placed under the elbow to align the humerus parallel to the ground. The fulcrum was placed distally to the patient on the elbow, with the reference arm perpendicular to the arm and the measurement arm aligned with the styloid process of the ulna. The participant’s shoulder was then brought passively into the maximal tolerated ER, and a measure was taken at the end position. The shoulder was then brought back to the resting neutral position, and the procedure was repeated to take a second measurement. The evaluator then changed sides to measure the contralateral shoulder using the same procedure. Next, the evaluator returned to the starting side and measured shoulder IR twice using the same procedure, which was finally repeated on the contralateral shoulder.

Figure 1:Participant setup for shoulder IR ROM measurement.

Shoulder ER ROM was obtained by taking the average of the two measurements. This was repeated for IR. Shoulder total range of motion was calculated as the sum of both ER and IR for each shoulder. Internal rotation loss was defined as the difference between shoulder IR from the dominant side compared to the non-dominant side.²⁷ External rotation gain was defined as the difference between shoulder ER of the dominant side with the non-dominant side.²⁷ Similar methods for measuring shoulder ROM have demonstrated very good inter-rater (intra-class correlations of 0.97 (ICC); 95%CI=0.89,0.99) and intra-rater reliability (ICC=0.95; 95%CI=0.87,0.98).²⁸

Strength

Shoulder IR and ER strength was measured using a CON-TREX® isokinetic dynamometer (CON-TREX MJ; CMV AG, Dübendorf, Switzerland) with a protocol of 90°/s concentric/concentric contractions with a maximum torque tolerance of 250Nm sampled at 4000Hz. Participants were measured in supine with the shoulder placed in 90° of flexion and abduction to replicate the throwing position (Figure 2). All measurements were taken in the afternoon before practice to avoid testing in a fatigued state. Eccentric contractions were not employed to avoid muscle soreness prior to training. Participants were provided with an opportunity to perform 10 sub-maximal repetitions of IR and ER of the non-dominant side as a warm up. After a one minute break, participants were asked to “push against the machine as hard as [they] can” for five repetitions. Verbal encouragement was provided throughout the testing procedure. After a two minute break, the procedure was repeated on the dominant side.

Figure 2:Participant setup for shoulder ER and IR strength measurements.

Shoulder torque values provided by the CON-TREX® software were gravity-corrected. A custom RStudio²⁹ script was written to filter only the values measured at the target test speed of 90°± 0.5°/s. The peak value was identified as the maximum value recorded within this filtered subset and used for the rest of the analysis in the study. Measures of relative torque were calculated by dividing the absolute values by the participants’ body weight. Ratios were obtained by dividing the peak ER torques by the peak IR torques. Between-days repeatability of isokinetic dynamometers is very good to excellent for shoulder assessments (ICC = 0.85,0.97).³⁰

Scapular Alignment

Scapular UR was measured using a Halo™ digital inclinometer (model HG1, HALO Medical Devices, Australia) after performing the dynamometer testing and with the participant standing with their shoulder in a 90° of abduction position (Figure 3). Scapular orientation was measured in the frontal plane only, and measurement of upward rotation was estimated by placing the fulcrum on the superior angle of the scapula and estimating the angle between the tip of the acromion and the horizontal plane. The participants were given 30 seconds to bring their arms down to rest, and the measure was repeated after the participants performed another 90° abduction movement. This was then repeated for the contralateral shoulder. Scapular UR was calculated by taking the average of the two measurements. This method was described previously to be reliable (ICC 0.81-0.94),³¹ and the position of shoulder abduction at 90° was preferred to identify differences.²³

Figure 3:Participant setup for scapular UR measurements.

Injury Surveillance

Injuries were defined in accordance with established consensus statements as any musculoskeletal injury or concussion for which the athletes required a consultation with a health care practitioner.³² In order to establish previous injury counts at baseline, a database of medical records was reviewed with a focus on shoulder injuries that had occurred in the prior 12 months. This database is linked with the participants’ electronic medical record (EMR), where every consultation with a sports medicine doctor, physiotherapist, or other health care practitioner had been entered and labelled for the corresponding injury accordingly. The EMR is maintained on a secure server with password encryption according to standards established by the Collège des Médecins du Québec. For the new injury incidence, an online surveillance program Hexfit™ (Hexfit Solutions Inc, Canada) was used to collect daily information on training loads and overuse injuries longitudinally for nine months of normal training and competitions. The system automatically flagged athletes who reported pain during training, and they were then contacted by the lead researcher to confirm that the injury qualified as per the study inclusion criteria. This method has been shown to be reliable in the past with a population of water polo players.³³

Analysis

Given the small sample available for this study, groups were dichotomized at baseline by those who had sustained a previous shoulder injury and those who had not. An additional analysis was done after nine months follow-up to compare players with new injuries versus no new injuries. Most variables showed close to normal distributions, except for strength variables. Therefore, independent t-tests were applied to compare dominant shoulder ROM and UR variables between healthy and injured players. Range of motion comparisons were made for range into ER and IR, total range of motion, ER gain and IR loss compared to the non-throwing shoulder. Mean UR was compared for scapular alignment differences. Mann-Whitney U tests compared relative dominant shoulder strength and strength ratios between the healthy and injured groups. The variables compared were average relative peak torque in ER and IR as well as ER:IR ratios. Effect sizes were calculated to compare group means with a Hedges g correction approach given the sample size, with small effect described as values <0.2, medium effect <0.5 and large effects >0.8.³⁴ Male and female players were compared as groups using the same approach. A chi-square analysis compared the proportions of players with a new injury vs a previous injury.

A logistic regression was performed to estimate the relative impact of the risk factors on new injuries in an exploratory analysis. The dependent variable was the development of a new injury over the nine month follow-up (1=injury, 0= no injury). In the first step, a history of previous injury was entered as a confounding variable (1=previous injury, 0=no previous injury). Next, a strength, ROM or UR variable was entered to determine if they related to the development of injuries over the nine month follow-up. Separate models were created for each strength, ROM or UR variable. The optimal model was decided as that which included only significant coefficients, provided the highest pseudo-R² value, and minimized the residual deviance. Odds ratios with 95% confidence intervals (CI) were also calculated for the variables included in the model based on the logit of the coefficients.

RESULTS

Nearly half of the participants in the study (18/39) had sustained a previous shoulder injury at baseline. Demographic variables were similar for the previously injured vs previously healthy groups in terms of age, sex, BMI, hand dominance, and training setting (Table 1). However, there were no goalies with previous shoulder injuries.

Observations comparing dominant to non-dominant sides showed increased dominant shoulder ER ROM (105±11° vs 98±11°, p=0.01) and decreased IR (53±11° vs 59±10°, p<0.01). There was however no difference in total range of motion (p=0.98). Furthermore, there were no significant differences in strength (p=0.58-0.70) or UR (p=0.99). Findings for group comparisons of strength, ROM and UR can be found in Table 2 and Table 3.

The previously injured group showed no significant differences in shoulder ROM into ER, IR or in total range of motion. However, athletes with a previous injury showed greater IR loss on the dominant shoulder (moderate ES g=0.68, 95%CI=0.03, 1.34) and higher mean relative IR strength (moderate effect size (ES), g=0.60; 95%CI=-0.05, 1.25). The ER:IR ratios were not significantly different between groups (Table 2). No significant difference was observed in UR.

At the nine month follow-up, players were once again divided into two groups based on the presence of a new shoulder injury (Table 3). Three players from the men’s team quit the program during the study, but had already developed new shoulder injuries before they left. Therefore, they were classified into the group with new injuries (n=19). A chi-square test confirmed that the players that had a previous injury were significantly more likely to develop new injuries (71.4% vs 27.8%, p=0.02). Furthermore, dominant shoulder IR loss was significantly higher in the group with new injuries (p=0.04, ES=0.68). Relative strength values were not different between groups, but UR was significantly greater in the group with new injuries (p<0.01, ES=0.81).

Sex comparisons showed that female players demonstrated higher total range of motion in rotation (p=0.02, ES g=0.75). Males were much stronger than the female players in both ER and IR, respectively (p<0.01, large ES g=2.03, 2.04), but ER:IR ratios were not different (Supplemental Table 1). No other variables were significantly different between sexes.

The best model fit to explain new injuries included previous injuries and UR (Table 4). This model minimized residual deviance (37.04) and maximized the pseudo-R² value using the Nagelkerke method (R²=0.47). The odds ratios (OR) for history of previous injury are 6.5, (95%CI=1.6, 26.4), and increased UR was related to more likelihood of developing a new injury (OR=1.5, 95%CI=1.1, 2.0) after accounting for a previous injury. No other variables were significantly related to new injuries in the logistic regression analyses.

DISCUSSION

Overall, this study showed that shoulder ER and IR ROM, strength, and UR are risk factors associated with shoulder injuries in water polo. At baseline, players with previous injuries demonstrated statistically significantly increased IR strength and loss of IR ROM on the dominant side. After nine months (and redistribution into injured/uninjured groups) strength measurements were not significantly different, but rather IR loss (greater in injured athletes) and UR showed a positive association. Largely, the most important predictor of new injury was the presence of a previous injury, with a 6.5 times increased odds of developing a new injury with this risk factor. Finally, male players showed higher strength values and less total ROM than their female counterparts.

Measures of relative IR strength were the only strength variable correlated with previous injury, and no strength variables were associated with new injury. In their group, Hams et al³⁵ found that high-level Australian water polo players with lower isometric strength had an association with new injuries. In the present study, relative IR strength was significantly higher for the group with previous injuries, but was not related to new injuries. The higher values of dominant shoulder strength for athletes with previous injuries may reflect that they may have been more likely to be performing targeted strengthening exercises to avoid new injuries, and thus demonstrated stronger test values. Consistent with Hams et al,¹¹ ER:IR strength ratios were not associated with new injuries, which suggests that asymmetries in rotator cuff strength may not be as widely present as was once suspected in this population.³

A greater loss of IR ROM was significant in the injured groups at baseline and after nine months. All other measures of ROM were otherwise similar between healthy and injured groups, and consistent with previous authors.^8,18 The loss of IR ROM may impact the players’ ability to decelerate the overhead throws, and put more mechanical stress on the rotator cuff muscles. Over time, this can lead to pathologies such as those observed in this population with MRI which affect the postero-superior area of the gleno-humeral joint.^36–38 A loss of shoulder IR ROM may also decrease the mechanical efficiency of the pulling motion of swimming, where players would need to increase scapular tilting to bring the arm in an optimal mechanical position. This in turn can lead to an increase in mechanical stress on the anterior structures of the shoulder such as the acromio-clavicular joint and the brachii.³⁹

The injured group at follow-up showed a significantly higher dominant shoulder mean UR. This variable was also a key factor in the logistic regression model, showing that increasing UR contributes to the risk of sustaining an injury. Based on previous studies, it would rather have been expected to find decreased values in the injured group.⁴⁰ These findings may be the result of limiting measurement to static positions where the range of values observed was narrow. Active movement measured with three-dimensional kinematic equipment would be more precise. Furthermore, Mukhtyar et al²³ found significant differences between injured and non-injured water polo players only when the players were in a fatigued state after training. The task of repeated shoulder rotations on the isokinetic dynamometer may not have stressed the scapulo-thoracic musculature sufficiently, and may not have induced the type of fatigue expected after water polo training.

The male players showed significantly higher relative strength compared to the female players in both ER and IR. This can be the result of different training methods, or a reflection of the more physical demands of the sport in the men’s style of play. Given that female players use a smaller and lighter ball, this may decrease the impact of lower strength on their ability to generate powerful overhead throws, but comparisons between sexes are lacking in the literature. The increased ROM that the female players demonstrated may be advantageous to accelerate the ball over a larger distance before throwing. However, this increased ROM may be an added risk factor for specific types of shoulder pathologies affecting joint stability.⁴¹

The study is limited in its generalizability given the small sample size. However, this sample included the entire population of international level water polo players in Canada, and the findings remain important for this group. A twelve-month follow-up was planned, but confinement due to COVID-19 pandemic interrupted all training activities after nine months. Secondly, a test of eccentric ER strength using the isokinetic dynamometer would allow to calculate a functional ratio of strength at the shoulder that resembles the throwing motion more closely (concentric IR to eccentric ER. In this study, this method was not chosen in order to limit fatigue before training sessions. Further studies investigating strength should consider this approach. Third, the methodology for measuring UR was optimal in the training setting, but it cannot yield information about active range of motion. In addition to taking all the measurements after training, future research should include a more substantial fatigue protocol to explore the conclusions of Mukhtyar et al.²³ Finally, other important risk factors were not considered, such as training volume and psychological factors.⁴²

CONCLUSION

In conclusion, the results of the current study indicate that a history of previous injury, as well as measures of shoulder IR and UR were most strongly associated with risk for sustaining a new injury in a sample of international level players of both sexes. This study adds to a small body of Level 2⁴³ literature on risk factors for shoulder injuries in water polo. These findings indicate that monitoring shoulder ROM, UR, and strength should be considered as core elements of an injury prevention program for water polo players. Additional studies which investigate the effectiveness of different protocols to optimize strength ratios and ROM are needed to guide these programs.

Conflicts of interest

None to declare

Funding

Research, Innovation and Dissemination of Information Program (PRIDI)