Aim The present study evaluated whether the MRI parameters of hamstring injuries in male professional football players correlate with time to return to play (RTP).
Methods 46 elite European football teams were followed prospectively for hamstring injuries between 2007 and 2014. Club medical staff recorded individual player exposure and time-loss after hamstring injury. MRI parameters were evaluated by two independent radiologists and correlated with the RTP data.
Results A total of 255 grade 1 and 2 injuries were evaluated in this study. RTP was longer for grade 2 than grade 1 injuries (24±13, 95% CI 21 to 26 days vs 18±15, 95% CI 16 to 20 days; mean difference: 6, 95% CI 2 to 9 days, p=0.004, d=0.39). 84% of injuries affected the biceps femoris (BF) muscle, whereas 12% and 4% affected the semimembranosus (SM) and semitendinosus (ST), respectively. No difference in lay-off time was found for injuries to the three different muscles (BF 20±15 days, SM 18±11 days, ST 23±14 days; p=0.83). The recurrence rate was higher for BF injuries than for SM and ST injuries combined (18% vs 2%, p=0.009). The size of the oedema weakly correlated with time to RTP (r2=6–12%). No correlation was found between location of injury and time to RTP. The majority of the intramuscular injuries affected the MT junction (56% in grade 1 and 2 injuries), but no difference in lay-off time was found between the different types of injuries.
Conclusions The radiological grade and size of the oedema correlate with time to RTP for both, grade 1 and 2 injuries. No correlations were found between time to RTP and the location and type of injury.
- Cohort study
- Elite performance
- Muscle injury
Statistics from Altmetric.com
Hamstring injury is the most common elite football injury, representing 12% of all injuries. A team with a 25-player squad typically suffers five to six hamstring injuries each season, resulting in 80 football days lost.1 Return to play (RTP) is the key question from the player, the coach, the media and agents.2 ,3 The team's medical staff need as much information as possible to provide an accurate response.4 ,5 The clinical diagnosis is key in predicting RTP.4 ,6 ,7 Clinical examination strongly correlates with the time to RTP.6 ,8 At the elite level, imaging of muscle injuries confirms clinical diagnosis and may help inform the decision for safe RTP.5 In elite European football, we have previously shown that 87% of injuries are examined with MRI or ultrasound.4 However, the additional value of MRI to patient history and clinical examination is debatable.7 Several studies have demonstrated the value of MRI as a prognosticator for lay-off time after hamstring injury.2–6 ,9–19 Schneider-Kolsky et al6 showed that both clinical examination and MRI strongly correlate with RTP, but the correlation between clinical examination and MRI was only moderate. These findings indicate that MRI and clinical examination provide different information that, together, may predict RTP. Recently, however, Wangensteen et al7 found that the additional predictive value of MRI was negligible compared to baseline patient history and clinical examinations alone.
The Elite Club Injury Study (ECIS) is an on-going prospective cohort study supported by the UEFA and a large source of evidence-based data (6000 muscle injuries between 2001 and 2014) in male elite footballers.12 ,20 We previously reported a good correlation between radiological grading and RTP.4 ,12 The objective of this study was to analyse whether the correlation between grading and lay-off time persisted in this extended material and whether further MRI parameters, such as muscles involved, location, type or size of hamstring injuries correlate with RTP. Our hypothesis was that a clear correlation exists between radiological grading of injuries and RTP, and that other imaging parameters may provide data to assist those working in professional football in estimating the time needed to RTP.
Materials and methods
This is a substudy of the on-going UEFA ECIS. A total of 46 teams from 12 countries have been followed over several seasons, from 2007 to 2014, including a total of 175 club-seasons. The study design followed the consensus statement on injury definitions and data collection procedures in football,21 and the general methodology has been reported elsewhere.22
Inclusion criteria and definitions
All first team squad players were eligible for inclusion each season. The general operational definitions are provided in table 1. Hamstring injury was defined as a traumatic distraction or overuse injury to the hamstring muscle group, including both first-time and recurrent lesions.1 ,23 Structural disorders, such as total and partial muscle ruptures and functional disorders, such as fatigue-induced, spine-related or neuromuscular muscle disorders, were included, whereas contusions, haematomas, tendon ruptures and chronic tendinopathies were excluded.23 ,24 Players were considered injured until the club medical staff allowed full participation in training and availability for match selection.22 Re-injuries were defined according to the concept of early recurrence.21
Baseline player data and consent were collected yearly at the time of player inclusion. During the season, medical staff registered individual player exposure in minutes during all club and national team training sessions and matches. Injury cards and attendance records were sent to the study group once a month. Each injury was coded according to a modified version of the Orchard Sports Injury Classification System (OSICS) 2.0.25
Magnetic resonance imaging
Twenty-one of the 46 teams (46%) agreed to provide MRI scans of thigh muscle injuries. MRI examination was performed within 24–48 h of injury. Minimum inclusion criteria for MRI included machine <5 years old and field strength of 1.5 T. Minimum MR sequences included axial and coronal T1 and T2 with fat saturation/ short τ inversion recovery sequences. The MRI Thigh Injury Report Form included the date of imaging, radiologist, MR sequences, muscles involved and injury severity. If more than one injured muscle was present, the parameters for the muscle with the most extensive pathology were recorded.
For severity classification, a modified Peetrons classification26 with the following grading system was used: grade 0—negative MRI without any pathology, grade 1—oedema but no architectural distortion, grade 2—architectural disruption indicating partial tear and grade 3—total muscle or tendon rupture. The location of injury within the muscle was defined as proximal, middle or distal third. The extent of the oedema or tear was measured in three planes: maximal length, Z (craniocaudal); width, X (mediolateral); and depth, Y (anteroposterior). Injuries were classified as follows: MT junction, an injury affecting the intramuscular myotendinous junction; myofascial, an injury at the surface of the muscle; muscle, an intramuscular injury away from the intramuscular tendon; or fascial, fluid in the intermuscular space but no muscle injury.
Thirteen teams used PACSMail (http://www.sybermedica.com) to send MRI scans for on-line review by two independent radiologists (JCH and JCL) blinded to clinical details other than the clinical diagnosis of hamstring injury. Copies of the 212 scans and associated reports from the 13 teams using PACSMail were then sent electronically to the study group. The other eight teams used paper-based MRI forms (n=95) filled in by the club consultant radiologist, which were sent to the study group by mail.
ANOVA was used for between-group comparisons of lay-off time, oedema size and size of the tear. These variables were presented as mean±SD with 95% CI or range. Effect size, using Cohen’s d, was calculated using the mean difference of all pairwise comparisons. The association between continuous variables was measured using Pearson's correlation coefficient (r). Multiple linear regression with a backward elimination procedure was used to estimate the effects the Z, X and Y planes of the oedema have on lay-off time. Estimates were expressed as the days change in lay-off time per millimetre change in the size of the oedema. In addition, the coefficient of determination (r2) of the final models are presented. The association between categorical variables was measured using Pearson's χ2 test or Fisher's exact test.
For the two independent radiologists (JH and JL), the reliability of the evaluation of grading, muscles involved, location of injuries and injury types were assessed using κ statistics. The reliability of the assessment of the size of the oedema and/or tear was evaluated using the intraclass correlation coefficient (ICC(3,1)).
All analyses were two-sided and the significance level set at p<0.05. All statistical analyses were performed in SPSS (IBM SPSS Statistics for Windows, V.22.0. Armonk, New York, USA: IBM Corp). The study design underwent an ethical review and was approved by the UEFA Football Development Division and the Medical Committee.
Of the 1488 hamstring injuries recorded during the study period, 917 (62%) were examined by MRI. MRI forms were received for 307 (33%) of the examinations (figure 1). Since grade 0 injuries (n=35) have no pathology on MRI, grade 3 injuries (n=9) were too few for robust analyses and data were missing for eight examinations, only grade 1 (n=173) and grade 2 (n=82) injuries with completed MRI forms were included in the following results (figure 1).
Reliability of evaluations
An almost perfect agreement between the two independent radiologists was found for the description of radiological grade (κ value 0.94, 95% CI 0.85 to 1.00). Substantial agreement was found for assessing the location of injury (κ value 0.71, 95% CI 0.56 to 0.84) and type of injury (κ value 0.65, 95% CI 0.49 to 0.80). There was total agreement regarding the muscles involved. For the measurement of oedema size, the agreement was substantial to almost perfect (ICC(3,1) 0.88, 95% CI 0.81 to 0.92; 0.80, 95% CI 0.7 to 0.87; 0.76, 95% CI 0.64 to 0.84 for Z, X and Y planes, respectively). For the measurement of the size of the tear, the agreement was moderate to substantial (ICC(3,1) 0.69, 95% CI 0.36 to 0.87; 0.45, 95% CI 0.04 to 0.74; 0.70, 95% CI 0.38 to 0.87 for Z, X and Y planes, respectively).
Grading and lay-off time
The mean lay-off time for all 255 hamstring injuries was 20±15 days (95% CI 18 to 22). A total of 173 injuries (68%) were grade 1 and 82 (32%) were grade 2. The lay-off time was significantly longer for grade 2 injuries than grade 1 injuries (24±13, 95% CI 21 to 26 days vs 18±15, 95% CI 16 to 20 days; mean difference: 6, 95% CI 2 to 9 days, p=0.004, d=0.39).
Of the 253 injuries for which data on muscle involvement were available, 212 (84%) affected the biceps femoris (BF) muscle, 30 (12%) in the semimembranosus (SM) muscle and 11 (4%) in the semitendinosus (ST) muscle. There was no significant difference in lay-off time for injuries to the three different muscles (BF 20±15, 95% CI 18 to 22 days; SM 18±11, 95% CI 14 to 22 days; ST 23±14, 95% CI 13 to 32 days; p=0.68) or in the lay-off time between the muscles with grade 1 (p=0.72) or grade 2 injuries (p=0.98). However, the association between recurrence rate and affected muscle was significant (BF 39/212=18%; SM and ST combined: 1/41=2%; p=0.009).
The size of the injury and lay-off time
Size of oedema in grade 1 and 2 injuries
The average length, width and depth of the oedema in grade 1 injuries was 100±58 (range 8–325) mm, 29±17 (range 2–124) mm and 19±12 (range 3–66) mm, respectively. The size of oedema in grade 2 injuries was larger in the Z, X and Y planes: 131±67 (range 5–242) mm (mean difference 31, 95% CI 11 to 51 mm, p=0.002, d=0.51), 39±24 (range 5–108) mm (mean difference 10, 95% CI 4 to 17 mm, p=0.001, d=0.56) and 32±20 (range 5–100) mm (mean difference 14, 95% CI 9 to 19 mm, p<0.001, d=0.93). The correlation between lay-off time and oedema in the Z, X and Y planes was significant (r=0.26, p<0.001; r=0.26, p<0.001; and r=0.15, p=0.043, respectively). Linear regression with backward elimination revealed a 2.0 day (95% CI 1.0 to 3.1 days) change in lay-off time per 10 mm change in the X plane oedema (p<0.001) when modelling all planes of the oedema on lay-off time. However, as indicated by the coefficient of determination (r2), the longitudinal length of the oedema only explained 7% of the variation in lay-off time. No significant two-way interaction effects were found between the Z, X and Y planes.
There was a small to medium correlation between lay-off time and the length and width of the oedema for grade 1 injuries (r=0.24, p=0.003 and r=0.23, p=0.005, respectively) but no significant correlation with the depth of the oedema (r=0.16, p=0.06). Linear regression for grade 1 injuries revealed a 0.7 day (95% CI 0.2 to 1.1 days) change in lay-off time per 10 mm change in the Z plane of the oedema (p=0.003, r2=6%). No significant two-way interaction effects were found between the Z, X and Y planes (figure 2).
For grade 2 injuries, there was a medium correlation between lay-off time and the Z and X planes of the oedema (r=0.31, p=0.03 and r=0.35, p=0.01, respectively) but no significant correlation with the Y plane (r=0.12, p=0.43). Linear regression for grade 2 injuries revealed a 1.8 day (95% CI 0.4 to 3.3 days) change in lay-off time per 10 mm change in the X plane of the oedema (p=0.013, r2=12%). No two-way interaction contributed significantly to the model (figure 2).
Size of the tear in grade 2 injuries
The average length, width and depth of the tear in grade 2 injuries was 42±23 (range 3–134) mm, 16±8 (range 3–34) mm and 13±12 (range 0–80) mm, respectively. No significant correlation was found between the size of the tear and lay-off time in Z (r=0.19, p=0.22), X (r=0.22, p=0.15) or Y (r=0.13, p=0.40) plane.
Location of injury and lay-off time
Grade 1 injuries were more common in the distal and middle thirds than in the proximal third (42% and 35% vs 23%). No significant difference in lay-off time was found between the different locations (p=0.58). The mean lay-off time for distal third injuries was 19±15 (95% CI 16 to 23) days compared to 17±10 (95% CI 14 to 19) and 19±21 (95% CI 13 to 26) days for middle and proximal third injuries, respectively.
Grade 2 injuries were more common in the middle third than in the proximal and distal thirds (44% vs 31% and 24%). Grade 2 injuries had a mean 23±11 to 15 days of lay-off time independent of location (p=0.99).
Re-injuries constituted 16% (41/255) of grade 1 and 2 injuries, with no significant difference in rates between the grades (p=0.95). For grade 1 injuries, the re-injury rates differed significantly between locations, as they were more common the more distal the injuries were located (proximal 5%, middle 14% and distal 21%, p=0.03). Recurrences of grade 2 injuries were too few (n=13) for robust analyses, but the tendency was opposite to grade 1 injuries with more recurrences in proximal injuries (27% vs 13% and 6% for middle and distal locations, p=0.06).
Type of injury and lay-off time
As shown in table 2, the majority of intramuscular injuries affected the MT junction (56% of both grade 1 and 2 injuries). No significant difference was found in lay-off time between the different types of injuries (MT junction 20±15, 95% CI 18 to 23; muscle: 20±10, 95% CI 16 to 25; fascial 15±6, 95% CI 10 to 20; myofascial 19±15, 95% CI 15 to 23; p=0.68). A separate analysis of grade 1 (p=0.87) and grade 2 (p=0.21) injuries indicated similar results.
Radiological grading is clearly related to time to RTP
The principal finding of our study of men's professional football is that radiological grading relates to lay-off duration. In our analysis of the ECIS material during the seasons in 2007–2010, we found that the lay-off time differed significantly between the four grades of muscle injury.4 In pairwise comparisons, the differences were significant between all pairs except between grade 1 and 2 (p=0.053).4 In the present extended study of seasons in 2007–2014, in which we evaluated only grade 1 and 2 injuries, we found significant differences in lay-off times for these two grades of injury, indicating that radiological grading of MRI is a helpful prognosticator for RTP. On average, grade 2 injuries were associated with 6 additional days to RTP.
Weak correlations between the size of the injury and RTP
The size of the oedema on MRI of the initial injury correlates slightly with lay-off time in grade 1 and grade 2 injuries. For grade 1 injuries, the longitudinal length of the oedema (Z-axis) had the strongest correlation with time to RTP. This correlation was weak and, as indicated by the coefficient of determination (r2), the longitudinal length of the oedema only explains 6% of the variation in lay-off time. For grade 2 injuries, the linear regression showed that the width of the oedema (X-axis) had the strongest correlation with time to RTP (explaining 12% of the variation in RTP). We found no correlation between the cross-sectional area of the injury (X×Y plane) and RTP.
Conflicting results concerning the correlation between oedema size and RTP have been reported. Three studies using multivariate analyses reported a moderate or strong correlation between both the longitudinal length and the cross-sectional area of the oedema and RTP,2 ,3 ,6 ,27 whereas others have found no correlation.8 ,28 The size of the tear (architectural distortion) did not correlate with lay-off time, which seems counter-intuitive. Oedema size was more significant, suggesting local reactive changes to insult may be more important than the primary injury.
No correlation between location of injury and RTP
The location of grade 1 and 2 injuries did not correlate with lay-off time. Conflicting results have previously been reported, suggesting that the distance from the muscle origin to the injury is associated with prolonged time to RTP.19 ,29 ,30 In a study of 18 sprinters, Askling et al8 reported that the closer the lesion is to the ischial tuberosity, the longer the interval to full activity. In contrast, our present data and the report from Slavotinek et al15 regarding Australian football revealed no such correlation. This difference probably reflects differences in patients. The time to RTP in Askling's paper was a median 16 weeks (112 days), indicating hospital-based material with severe injuries. The median absence was 3–4 weeks (27 days) in the Australian football study and 2 weeks (14 days) in the present study, indicating field-based material with less severe injuries. We conclude that, in elite football, the location of the injury (proximal, midsubstance or distal) in grade 1 and 2 hamstring injuries has no predictive value for RTP.
Is typing the injury meaningful?
A new grading system for muscle injuries was proposed recently (British Athletics Muscle Injury Classification).30 ,31 This classification defined injury according to the intramuscular site: myotendinous (termed MT junction in the present study), myofascial, perifascial (termed fascial in the present study) and intramuscular (termed muscle in the present). The classification also takes into account whether the tendon itself is involved. The authors suggest injury to the MT junction, and especially injury to the tendon, may be associated with prolonged time to RTP.32 However, this grading system is currently expert opinion and awaits validation.7 ,31 ,32
In our study, we found that the majority of grade 1 and 2 injuries affect the intramuscular myotendinous junction, whereas injuries affecting the myofascial tissue and not involving the tendon are less frequent. Pure muscle or fascial injuries were rare. Our data suggest that typing injuries according to location does not correlate with lay-off times and its utility is questionable. The results of our study agree with data from Connell et al.2
No difference in lay-off days between specific muscles but a higher re-injury rate for BF injuries
We found that the majority of injuries involved the BF muscle. Injuries to the SM and ST are much less common, which is in agreement with other studies.2–4 ,33 ,34 Furthermore, our study shows that the specific muscle involved does not correlate with the time to RTP, which is in agreement with earlier studies.3 ,15 ,34 However, we found that the recurrence rate was substantially higher for injuries affecting the BF compared to those affecting the SM and ST, but larger data sets are needed to validate this statement, as injuries to ST and SM are less frequent than BF injuries.
Strengths of the study
The main strength of this study is the prospective data set with 255 injuries from a homogenous group of male professional footballers. We used standardised methodology that complies with international consensus agreements on procedures for epidemiological studies including RTP in football injuries.21 ,22 Furthermore, our classification and measurements have excellent interobserver agreement and reliability. The radiologists were blinded to clinical findings and performed their analysis independent of each other but achieved excellent agreement for radiological grading (κ value 0.94). Similar excellent interobserver and intraobserver reliability for grading and prognostic MRI parameters in acute hamstring injuries were previously reported by Hamilton et al.35 Thus, the scoring of hamstring injuries by experienced radiologists is reproducible.35
Limitations of the study
The medical teams in the clubs who made the RTP decisions were not blinded to imaging results; they were blinded to the grading of the study radiologists but were informed of the MRI evaluations by their own radiologists. There were no specific criteria for RTP due to the lack of consensus regarding optimal treatment or uniform guidelines for RTP in hamstring injuries in football players.7 In addition, we did not investigate how risk taking in important matches or due to external pressure influences RTP decisions.7
What are the findings?
The radiological grade correlates with time to return to play (RTP) for both, grade 1 and 2 injuries.
The size of the oedema only weakly correlates with time to RTP.
For grade 1 and 2 hamstring injuries, the proximity to the ischial tuberosity has no predictive value for RTP in elite football.
There is no difference in lay-off days between specific hamstring muscles involved, but biceps femoris (BF) injuries have a higher re-injury rate.
The typing of muscle injuries according to intramuscular site can be questioned.
How might it impact on clinical practice in the future?
For practitioners working on the football field and frequently using imaging, information about MRI parameters that are and are not correlated with RTP will be of interest.
The difference in recurrence rates between injuries to BF compared to semimembranosus (SM) and semitendinosus (ST) could guide practitioners to be more conservative with BF injuries due to the high recurrence rate (better safe than quick) and to be more aggressive in the rehabilitation of injuries to the SM and ST muscles because re-injuries to these muscles are rare.
The authors wish to thank all the participating clubs, including coaches, players and medical staff. Biostatistician Henrik Magnusson is acknowledged for his assistance with the statistical analyses.
Contributors JE, JCL and JCH were responsible for the conception and design of the study. All the authors were involved in data collection over the study period. JE conducted the analyses with the biostatistician (HM). All the authors contributed to interpreting the findings. JE wrote the first draft of the paper, which was critically revised by JCL and JCH. The final manuscript was approved by all the authors. JE is the study guarantor.
Funding The Football Research Group was established in Linköping, Sweden, in collaboration with Linköping University and through grants from the Union of European Football Associations, the Swedish Football Association, the Football Association Premier League Limited and the Swedish National Centre for Research in Sports.
Competing interests None declared.
Ethics approval The study design was approved by the UEFA Medical Committee and the UEFA Football Development Division.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.