Objective To systematically review the evidence base related to hamstring strength and flexibility in previously injured hamstrings.
Design Systematic review and meta-analysis.
Data sources A systematic literature search was conducted of PubMed, CINAHL, SPORTDiscus, Cochrane Library, Web of Science and EMBASE from inception to August 2015.
Inclusion criteria Full-text English articles which included studies which assessed at least one measure of hamstring strength or flexibility in men and women with prior hamstring strain injury within 24 months of the testing date.
Results Twenty-eight studies were included in the review. Previously injured legs demonstrated deficits across several variables. Lower isometric strength was found <7 days postinjury (d=−1.72), but this did not persist beyond 7 days after injury. The passive straight leg raise was restricted at multiple time points after injury (<10 days, d=−1.12; 10–20 days, d=−0.74; 20–30 days, d=−0.40), but not after 40–50 days postinjury. Deficits remained after return to play in isokinetically measured concentric (60°/s, d=−0.33) and Nordic eccentric knee flexor strength (d=−0.39). The conventional hamstring to quadricep strength ratios were also reduced well after return to play (60:60°/s, d=−0.32; 240:240°/s, d=−0.43) and functional (30:240°/s, d=−0.88), but these effects were inconsistent across measurement methods.
Conclusions After hamstring strain, acute isometric and passive straight leg raise deficits resolve within 20–50 days. Deficits in eccentric and concentric strength and strength ratios persist after return to play, but this effect was inconsistent across measurement methods. Flexibility and isometric strength should be monitored throughout rehabilitation, but dynamic strength should be assessed at and following return to play.
Statistics from Altmetric.com
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.
Hamstring strain injuries (HSIs) are the most common non-contact injuries in Australian rules football,1–5 soccer,6–10 rugby union,11–14 track and field15–17 and American football.18 HSIs result in time away from competition,9 financial burden9 ,19 and impaired performance on return to competition.20
Further to this, recurrent HSI often leads to a greater severity of injury than the initial insult.10 ,14 The most commonly cited risk factor for future HSI is a previous HSI.21–24 The high recurrence rates of HSI10 ,14 are proposed to result from incomplete recovery and/or inadequate rehabilitation25 ,26 because of pressure for early return to play at the expense of convalescence.27 Consequently, there has been much interest recently in observations of hamstring structure and function in previously injured legs compared with control data.28–34 Despite the possible limitation of this approach, it is often agreed that deficits that exist in previously injured hamstrings could be a maladaptive response to injury.35 As such, these deficits that persist beyond return to play could provide markers to better monitor athletes during and/or at the completion of rehabilitation.35
Which parameters are the best markers to monitor an athlete's progress during rehabilitation? Conventional clinical practice focuses on measures of strength and flexibility; however, the evidence is based on predominantly retrospective observations of strength,28 ,29 ,36–42 strength ratios36 ,37 ,39 ,40 ,43 ,44 and flexibility26 ,28 ,42 ,45–49 in previously injured athletes. These studies are limited in reporting single or isolated measures with methodologies and populations that differ from study to study. To advance knowledge, we aimed to systematically review the evidence base related to hamstring strength and flexibility in previously injured hamstrings.
A systematic literature search was conducted of PubMed, CINAHL, SPORTDiscus, Cochrane Library, Web of Science and EMBASE from inception to August 2015. Keywords (table 1) were chosen in accordance with the aims of the research. Retrieved references were imported into EndNote X7 (Thomson Reuters, New York City, New York, USA), with duplicates subsequently deleted. To ensure all recent and relevant references were retrieved, citation tracking was performed via Google Scholar and reference list searches were also conducted.
Selection criteria were developed prior to searching to maintain objectivity when identifying studies for inclusion. To address the aims, included papers had to:
Assess at least one parameter of hamstring strength (maximum strength, associated strength ratios and angle of peak torque) or flexibility in humans with a prior HSI within 24 months from the time of testing;
Have control data for comparison (whether it was a contralateral uninjured leg or an uninjured group);
Have the full-text journal article in English available (excluding reviews, conference abstracts, case studies/series);
Not include hamstring tendon or avulsion injuries.
The titles and abstracts of each article were scanned by one author (NM) and removed if information was clearly inappropriate. Selection criteria were then independently applied to the remaining articles by three authors (NM, RGT and DAO). Full text was obtained for remaining articles, with selection criteria reapplied by one author (NM) and cross-referenced by another author (DAO).
Assessing bias and methodological quality
Risk-of-bias assessment was performed independently by two examiners. We used a modified version of a checklist by Downs and Black.50 The original checklist contained 27 items, however many were relevant only to intervention studies. Since the majority of the papers in this review were of a retrospective nature, items 4, 8, 9 13, 14, 15, 17, 19, 22, 23, 24 and 26 were excluded as they were not relevant to the aims of the review.
Of the remaining items, 1, 2, 3, 5, 6, 7 and 10 assessed factors regarding the reporting of aims, methods, data and results, while items 16, 18, 20, 21 and 25 assessed internal validity and bias. Item 27 was not suitable to the context of the current review and was modified to address power calculations. Two new items (items 28 and 29) relating to injury diagnosis and rehabilitation/interventions were added to more appropriately assess the risk of bias; thus, the modified checklist contained 17 items (see online supplementary table S1).
Fourteen of the items were scored 0 if the criterion was not met or it was unable to be determined, while successfully met criteria were scored 1 point. The other three items (items 5, 28 and 29) were scored 0, 1 or 2 points, as dictated by the criteria presented in online supplementary table S1. This resulted in a total of 20 points available for each article.
Similarly, modified versions of this checklist have been used in previous systematic reviews investigating factors leading to heel pain51 and risk factors associated with hamstring injury.52 The risk-of-bias assessment was conducted by two authors (NM and DAO), with results expressed as a percentage. In the case of disagreement between assessors, an independent individual was consulted with consensus reached via discussion if necessary. In situations where one of the assessors (DAO) was a listed author on a study included for review, the independent individual completed the risk-of-bias assessment in their place.
Relevant data were extracted including the participant numbers, population and sampling details, diagnosis technique, severity of injury, time from injury to testing (in days assuming 30.4 days/month, 365 days/year), variables investigated and how these were tested, results including statistical analysis and, where appropriate, potential confounders that may affect strength or flexibility outcomes. The major confounders include other lower limb injuries likely to affect strength and flexibility, interventions and rehabilitation programmes performed. Furthermore, insufficient evidence exists regarding the interaction between gender and HSI; thus, mixed-gender cohorts were considered as a potential confounder.
Although objectively synthesising evidence via a meta-analysis is often desirable, this technique was not able to be applied to all the evidence retrieved in this review owing to insufficient reporting of data (ie, two or more studies or subgroups with mean, SD and participant numbers for contralateral leg comparisons) or methodological variations between studies.
When sufficient data were available, meta-analysis and graphical outputs were performed using selected packages53–55 on R (R Development Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing, 2010). Standardised mean differences (Cohen's d) facilitated the comparison of studies reporting variables in different units, with effect estimates and 95% CIs summarised in forest plots. A random-effects model was used to determine the overall effect estimate of all studies within the variable or subgroup as appropriate, with variance estimated through a restricted maximum likelihood (REML) method. The magnitudes of the effect size were interpreted as small (d=0.20), moderate (d=0.50) and large (d=0.80), according to thresholds proposed by Cohen.56 Where studies reported multiple types of data (eg, multiple isokinetic velocities, multiple subgroups or multiple time points), these data were analysed as subgroups to avoid biasing the weighting of the data. These time points were dictated by the data available. Where data were available in the acute stages (prior to return to play), time bands were kept at <10 days as it would be expected that deficits would change relatively rapidly during this time owing to ongoing rehabilitation and recovery.
Data presented for participants at or after return to play were pooled for two reasons: (1) no included study reported any ongoing rehabilitation after return to play and (2) many of these studies had variable time from injury until testing between individual participants. Where a study had multiple time points that fit within post return to play time band (eg, at return to play and follow-up), the earlier option was chosen as there was expected to be a lower chance of bias due to other uncontrolled or unmonitored activities. For the purposes of meta-regression (employed to assess the effects of time since injury), studies with multiple time points were pooled to provide the best assessment of the effect of time on the given variable. By considering each time point as a subgroup, sufficient data (>10 subgroups) were available for meta-regression analysis57 providing that time from injury until testing was reported. Funnel plots were visually inspected for asymmetry to assess publication bias. Heterogeneity was determined by the I2 statistic and can be interpreted via the following thresholds:57
0–40%: might not be important
30–60%: may represent moderate heterogeneity
50–90%: may represent substantial heterogeneity
75–100%: considerable heterogeneity.
In situations where it was deemed that reported data (ie, mean, SD, participant numbers for contralateral leg comparisons) were insufficient for meta-analysis and could not be obtained via supplementary material or from contacting the corresponding author, a best evidence synthesis58 was employed. The level of evidence was ranked according to criteria consistent with previously published systematic reviews59 ,60 as outlined below:
Strong: two or more studies of a high quality and generally consistent findings (≥75% of studies showing consistent results);
Moderate: one high-quality study and/or two or more low-quality studies and generally consistent findings (≥75% of studies showing consistent results);
Limited: one low-quality study;
Conflicting: inconsistent findings (<75% of studies showing consistent results);
None: no supportive findings in the literature.
A high-quality study was defined as a risk-of-bias assessment score of ≥70%, whereas a low-quality study had a risk of bias assessment score of <70%.57
The search strategy consisted of six steps (figure 1). The initial search yielded 7805 items (Cochrane Library=131; PubMed=2407, CINAHL=604; SPORTDiscus=640; Web of Science=1049; EMBASE=2974) from all databases. After duplicates were removed, 4306 items remained. Title and abstract screening resulted in 92 remaining articles, reference list hand searching and citation tracking resulted in the addition of six articles. Independent application of the selection criteria yielded 28 articles to be included in the review, 22 of which were included in meta-analysis.
Risk-of-bias assessment of each article is displayed in table 2. It is important to note that the risk-of-bias assessment was not the basis of exclusion. Included articles ranged from a score of 8 to 18 of a possible 20 (40–90%).
Description of studies
A sample of 898 participants (n=802 male, n=96 female; age range, 15–47 years) were examined across the included studies. A total of 17 studies included only male participants,29 ,34 ,36 ,37 ,39–43 ,45 ,46 ,48 ,49 ,61–64 10 studies had mixed gender,26 ,28 ,33 ,47 ,65–70 while only 1 exclusively studied females.71 Participants were generally considered recreationally active at a minimum.
Methods of diagnosis varied between studies, with some studies using multiple methods of diagnosis. A total of 12 studies used clinical criteria,26 ,28 ,33 ,34 ,36 ,37 ,42 ,48 ,66–69 10 used MRI,26 ,28 ,29 ,33 ,34 ,62 ,65 ,67–69 5 had medical or health practitioner diagnosis,39 ,41 ,43 ,48 7 used a questionnaire or self-report,40 ,46 ,47 ,49 ,58 ,63 ,71 2 used ultrasound36 ,37 and 2 had unclear methods of diagnosis.45 ,70 Description of severity of injury varied significantly between studies, with the most common being time to return to play26 ,28 ,29 ,40–43 ,48 ,49 ,63 ,67 and grade (I–III) of injury.29 ,31 ,33 ,39 ,62 ,66 ,68 ,69 ,70 Description of time from injury to testing varied significantly between studies (range, 2–690 days).
The strength variables examined were concentric, eccentric and isometric strength (absolute and normalised to body mass), strength ratios (usually hamstring to quadricep (H:Q)) and angle of peak torque. The five flexibility variables examined were passive straight leg raise, active straight leg raise, passive knee extension, active knee extension and the sit and reach. All five strength variables (concentric, eccentric, isometric, strength ratios and angle of peak torque) and three flexibility variables (passive straight leg raise, active knee extension and passive knee extension) were included for meta-analysis. Sufficient data were available to run meta-regression analysis for isometric strength and the passive straight leg raise. The best evidence synthesis method was applied to remaining variables for which insufficient data were available for meta-analysis. The best evidence synthesis is summarised in table 3.
Data for all studies which examined concentric strength can be found in online supplementary table S2.
Meta-analysis. Concentric strength was measured isokinetically at 60°,29 ,40 ,48 ,61–63 ,66 ,67 ,71 180°29 ,40 ,61 ,71 and 300°/s.39 ,40 ,62 ,71 A statistically significant small effect for lower concentric strength at 60°/s was found in previously injured legs (effect size −0.33; 95% CI −0.53 to −0.13; I2 0%), but no significant effects were found at 180° or 300°/s (figure 2).
Best evidence synthesis. Of the dynamic strength variables which were not included in the meta-analysis, one (seated isokinetic at 240°/s)36 ,37 ,67 had moderate evidence for a decrease in strength in the previously injured hamstrings. Concentric strength at 270°/s in a seated position42 had limited evidence and concentric strength at 60°/s in a prone position49 had no supporting evidence.
Data for all studies which examined eccentric strength can be found in online supplementary table S3.
Meta-analysis. Eccentric strength measured during the Nordic hamstring exercise34 ,41 ,64 and isokinetically at 60°29 ,48 ,62 ,63 ,70 and 180°/s29 ,70 were included in the meta-analysis. Significant deficits in previously injured legs were found for eccentric strength measured via the Nordic hamstring exercise (effect size −0.39; 95% CI −0.77 to 0.00; I2 0%), but not for any other method (figure 3).
Best evidence synthesis. Eccentric isokinetic strength measured at 30°36 ,37 ,42 ,61 and 120°/s36 ,37 had moderate evidence, indicating lower strength in previously injured hamstrings, whereas measures at 230°42 and 300°/s39 had limited evidence. The measurement of eccentric strength at 60°/s in a prone position49 had no supporting evidence.
Data for all studies which examined isometric strength can be found in online supplementary table S4.
Meta-analysis. Isometric strength measured at long muscle lengths (hip 0°; knee 0–15°) was included in the meta-analysis.28 ,34 ,68 Measures were taken at multiple time points (<7, 7–14, 21, 42 and >180 days) postinjury; thus, subgroups were analysed (figure 4) and meta-regression was performed. A large effect for lower long-length isometric strength was statistically significant in previously injured legs compared with the uninjured contralateral legs <7 days postinjury (effect size −1.72; 95% CI −3.43 to 0.00; I2 91%), but not at any other time point. Meta-regression analysis (figure 5) revealed no significant effect for time since injury for isometric strength (intercept −0.92, p=0.002; coefficient 0.003, p=0.292).
Best evidence synthesis. One study68 assessed isometric strength in a short muscle length (hip 0°, knee 90°). This study did not statistically test for differences between muscles, but based on effect size and CIs, isometric strength was reduced at the initial evaluation (effect size −0.74; 95% CI −1.07 to −0.41) and at the 7-day follow-up (effect size −0.39; 95% CI −0.71 to −0.07), but not at the 26-week follow-up (effect size −0.12; 95% CI −0.45 to 0.20).
H:Q torque ratio
Meta-analysis. The conventional H:Q ratio, whereby peak torque of each muscle group is assessed during concentric isokinetic contraction, was assessed at 60:60,36 ,37 ,40 ,43 ,48 ,61 ,70 ,71 180:180,40 ,61 ,70 ,71 240:24036 ,37 and 300:300°/s39 ,40 ,71 (figure 6). A statistically significant small effect for a lower conventional H:Q ratio was found in previously injured legs compared with the uninjured contralateral legs at 60:60 (effect size −0.32; 95% CI −0.54 to −0.11; I2=0%) and 240:240°/s (effect size −0.43; 95% CI −0.83 to −0.03; I2 0%), but not at 180:180 and 300:300°/s. Meta-analysis of the functional H:Q (fH:Q), whereby the hamstring group is assessed eccentrically, but the quadricep group is assessed concentrically, included isokinetic velocities 30:24036 ,37 ,67 and 60:60°/s43 ,48 ,63 ,70 (figure 7). A large effect size for a lower fH:Q ratio was found in previously injured legs at 30:240°/s (effect size −0.88; 95% CI −1.27 to −0.48; I2 0%), but there were no significant differences between injured and uninjured legs at 60:60°/s.
Best evidence synthesis. One study which examined H:Q (60:60°/s)49 was not included in the meta-analysis owing to the prone and supine position in which knee flexor and quadricep strength, respectively, were assessed. This study found no significant difference between injured and uninjured legs. No supporting evidence was found for the fH:Q strength ratio at 180:180,70 30:60 and 30:180°/s,61 and limited evidence was found for 300:300°/s.39 The eccentric H:Q, whereby both knee flexor and quadricep strength are assessed via eccentric contractions, was assessed isokinetically in prone/supine49 position. This study found no differences between previously injured and uninjured legs. Limited evidence was found for eccentric knee flexor torque to concentric hip flexor torque ratio deficits in previously injured legs (effect size −0.9) compared with uninjured contralateral legs.39
Angle of peak torque
Data for all studies which examined optimal angle of peak torque can be found in online supplementary table S7.
Meta-analysis. The optimal angle of peak torque (concentric 60°/s) had sufficient data61 ,66 ,67 for meta-analysis. No significant differences between injured and uninjured legs were found (figure 8).
Best evidence synthesis. Limited evidence was found for the eccentric angle of peak torque to occur at significantly shorter muscle lengths in the injured legs compared with the uninjured contralateral legs at 30°/s.61 No differences were found for angle of peak torque between legs/groups at 240°67 and 300°/s39 concentrically or 300°/s39 eccentrically measured angle of peak torque.
Passive straight leg raise
Data for all studies which examined the passive straight leg raise can be found in online supplementary table S8.
Meta-analysis. Quantitative analysis of the passive straight leg raise26 ,28 ,62 ,68 revealed significantly reduced range of motion in previously injured legs compared with the uninjured contralateral leg. A large effect was found within 10 days (effect size −1.12; 95% CI −1.76 to −0.48; I2 81%), a moderate effect between 10 and 20 days (effect size −0.74; 95% CI −1.38 to −0.09; I2 76%) and a small effect between 20 and 30 days (effect size −0.40; 95% CI −0.78 to −0.03; I2 4%) since the time of injury, with no significant effect found after 40 days since the time of injury (figure 9). Meta-regression analysis (figure 10) revealed a significant effect for time since injury (intercept −0.81, p<0.0001; coefficient 0.006, p=0.019), indicating that the magnitude of the range of motion deficit decreases with increasing time from injury.
Passive knee extension
Data for all studies which examined the passive knee extension can be found in online supplementary table S9.
Meta-analysis. No significant differences were found for the passive knee extension measure at either time point subgroup analysed (<10 and 20–30 days postinjury; figure 11A,B).
Best evidence synthesis. A subset of the passive knee extension (insufficient data for subgroup meta-analysis, unable to be pooled with acute data) showed conflicting evidence across the three studies46 ,47 ,49 that conducted this assessment post return to play.
Active knee extension
Data for all studies which examined the active knee extension can be found in online supplementary table S9.
Meta-analysis. No significant differences were found for the active knee extension measure at either time point subgroup analysed (<10, 10–30 and >100 days postinjury; figure 11C–E).
Active straight leg raise
Data for all studies which examined the active straight leg raise can be found in online supplementary table S8.
Best evidence synthesis. Conflicting evidence was found for deficits in the active straight leg raise.45 ,65 Of note, the one study65 which did find deficits in previously injured legs performed the active straight leg raise in a rapid manner (Askling-H test), and as such this study could not be appropriately pooled with the other data for meta-analysis purposes.
Our systematic review revealed that after HSI, isometric strength and passive straight leg raise deficits normalised within 20–50 days. Deficits at or after return to play, if they did exist, manifested during dynamic strength measures (eccentric and concentric strength and their associated H:Q strength ratios).
We only included research articles that contained data from participants who had previously sustained a HSI (between 2 and 690 days prior). As a result, we cannot determine whether the reported deficits were the cause of injury or the result of injury. Given the increased risk of future HSI in those with an injury history,21–24 the characteristics that exist in these legs should be given consideration by the clinicians responsible for rehabilitation and clearance to return to play.
Strength and flexibility deficits after hamstring injury
Conventional rehabilitation practice traditionally focuses on restoring isometric strength and range of motion.72 The meta-analysis revealed that deficits in long length (hip 0°; knee 0–15°) isometric strength and the passive straight leg raise are resolved 20–50 days postinjury. This provides support for the use of the passive straight leg raise and isometric strength measures during rehabilitation.72 Furthermore, deficits in isometric strength and range of motion (as measured by the active knee extension test) just after return to play are independent predictors of reinjury,73 suggesting that these variables likely also have value in criterion-based rehabilitation progressions. However, where evidence of deficits were found beyond return to play, these were during measures of dynamic strength.
The evidence supporting deficits in eccentric strength in those with prior HSI is mixed.29 ,34 ,36 ,37 ,39 ,41–43 ,48 ,63 ,64 ,70 Lower levels of eccentric hamstring strength are proposed to increase the likelihood that the demands of high-force musculotendinous lengthening, such as during the terminal swing phase of running, exceed the mechanical limits of the tissue.74 It may be that lower eccentric strength in previously injured hamstrings is at least partly responsible for the greater risk of recurrent hamstring strain.75
Other measures of dynamic strength, including concentric strength29 ,33 ,36 ,37 ,40 ,48 ,61–63 ,66 ,67 ,71 and both conventional33 ,36 ,37 ,39 ,40 ,43 ,48 ,61 ,66 ,70 ,71 and functional36 ,37 ,39 ,43 ,48 ,61 ,63 ,67 ,70 H:Q strength ratios, also show conflicting findings, with measures at some testing velocities showing lower strength in previously injured legs, but others showing no differences. The reasons for these discrepancies are unclear, but it may be due to inherent differences in groups studied and/or methodological issues. For example, studies which included females tended to observe slightly higher strength in previously injured legs.70 ,71 Insufficient data were available to assess this observation via regression analysis; thus, more research is needed to investigate any potential gender-specific responses to HSI. The particulars of the rehabilitation performed could also explain the disparity, as differing rehabilitation strategies would result in differing adaptations. Rehabilitation was rarely controlled in the included studies, suggesting that more studies should aim to control rehabilitation to limit this potential confounder.
Mechanisms that may explain long-term dynamic muscle strength deficits
There is the possibility that chronic deficit/s in dynamic strength in previously hamstring strain injured legs is a downstream outcome of prolonged neuromuscular inhibition.35 Reduced activation of previously injured hamstrings has been associated with maximal eccentric contractions,29 ,30 ,48 ,76 particularly at long muscle lengths.29 ,48 What remains to be seen, however, is whether or not these deficits are associated with increased risk of injury or reinjury, and what the most appropriate intervention is to ameliorate these deficits. However, activation deficits do not occur during concentric contractions;29 ,48 thus, further research is needed to understand why dynamic strength deficits tend to persist beyond return to play.
The data presented in this review have implications for practitioners who rehabilitate and return athletes to play following HSIs. The supplementary result tables provide practitioners a detailed resource of data for almost all strength and flexibility measures that have been assessed in athletes with a prior HSI. These data can be used to compare individual athlete/patient data. It should also enable practitioners to select measures to monitor in their injured athletes which are known to be in deficit despite ‘successful’ return to play. The presented evidence justifies the use of the passive straight leg raise and isometric strength measures to monitor progression through rehabilitation, while additional measures of dynamic strength may have more value at and after return to play.
In addition, the present review would also question the use of commonly recommended74 ,77 and employed markers for successful rehabilitation, such as knee flexor angle of peak torque. The use of angle of peak knee flexor torque, particularly during concentric contraction, in athletes with prior HSI has been popularised following the seminal paper;66 however, the ensuing evidence is generally conflicting,33 ,39 ,61 ,67 suggesting that the value of this measure should be questioned.
The primary limitation of this review is that the retrospective nature of the data makes it impossible to determine if deficits are the cause or result of injury. For example, eccentric strength deficits could be the result of uncorrected strength deficiency that may have caused injury, as higher levels of eccentric strength and eccentric training are associated with a reduction in new and recurrent HSI.73 ,78 ,79 Furthermore, the majority of the included studies did not control rehabilitation, and this introduces another potential source of bias. For example, a study in which participants focused heavily on eccentric exercise as part of rehabilitation may show no evidence of significant eccentric strength deficits post HSI. Consequently, the effect of these interventions on strength and flexibility outcomes remains an area for future research. Ideally, researchers should control rehabilitation to minimise confounding, and where this is not possible, collect and report details of rehabilitation protocols. Inconsistent time from injury until testing between studies also introduces bias. We analysed data in time bands and performed meta-regression analysis where possible to assess and adjust for this potential confounder, but we also acknowledge that this approach was limited by within-study variability, variability between studies within the time band subgroups and insufficient data for regression analysis. Future research should investigate the effect of time since injury on deficits, particularly prior to return to play, as strength and flexibility appear to change rapidly during this period.
One of the difficulties of this review was the numerous methods employed by different studies to assess a given parameter. For strength testing, it appeared that lower isokinetic velocities (<60°/s) were the most sensitive to deficits; however, there are insufficient data at higher velocities to draw definitive conclusions. Similarly, a number of different measures of flexibility (passive26 ,28 ,42 ,65 and active45 ,65 straight leg raise, passive26 ,46 ,47 ,49 and active knee extension,26 ,48 sit and reach test48 ,63) have been assessed in previously injured athletes, with inconsistent findings among studies. Indeed, within each variable, the meta-analysis revealed significant heterogeneity as determined by the I2 statistic in certain measures, particularly in the initial days following injury.
To address these issues as far as possible, we performed sensitivity analysis (see online supplementary table S10) to examine the influence of individual studies on effect estimates and heterogeneity where moderate (≥30%) heterogeneity57 may have been present. While high heterogeneity often impairs the validity of synthesised data, the low number of studies in many of these subgroups precludes confidence in the precision in these I2 estimates, suggesting that more studies are needed to properly interpret heterogeneity estimates. These studies should also take care to accurately describe diagnostic procedures, injury severity and other lower limb injuries likely to confound results. The data reported in this review may also have limited application to female athletes, as majority of the data were obtained from male only or predominately male cohorts. We acknowledge that the search strategy may not have captured all relevant literature. However, reference list searching and citation tracking were also performed to enhance article retrieval.
In conclusion, the meta-analysis found that deficits in isometric strength and flexibility (as measured by the passive straight leg raise) resolve within 20–50 days following HSI. Deficits that were present beyond return to play were found for dynamic measures of strength (concentric and eccentric strength, and conventional and functional H:Q strength ratios). This evidence suggests that clinicians monitor isometric strength and the passive straight leg raise throughout rehabilitation, while dynamic measures of strength may hold more value at/after return to play. Furthermore, it may behove clinicians and patients to continue rehabilitation after return to play.
What are the findings?
Isometric strength returns to the level of the contralateral uninjured leg within 20 days.
Range of motion measured by the passive straight leg raise returns to the level of the contralateral uninjured leg within 50 days.
Lower dynamic strength (concentric, eccentric and associated strength ratios) in previously injured legs compared with the uninjured contralateral legs persists beyond return to play, but this is inconsistent across measurement technique.
How might it impact on clinical practice in the future?
Isometric strength and the passive straight leg raise provide a measure of progression during rehabilitation.
Dynamic strength (concentric/eccentric hamstring strength and associated hamstring to quadricep strength ratios) may also be helpful in monitoring progress through rehabilitation and return-to-play decisions.
This review adds weight to the argument that rehabilitation should continue after return to play if the goal is to achieve symmetry in strength and range of motion.
The primary author's position was supported through the Australian Government's Collaborative Research Networks (CRN) program. The authors would also like to sincerely thank Professor Geraldine Naughton for acting as an independent assessor for the risk-of-bias assessment.
Contributors NM conducted the search, risk-of-bias and criteria assessments, extracted the data, performed all analysis and drafted the manuscript. AJS and MDW contributed to interpretation of results and the manuscript. RGT conducted criteria assessments and contributed to the manuscript. DAO conducted risk-of-bias and criteria assessments and contributed to the interpretation of results and the manuscript.
Funding Australian Government's Collaborative Research Networks (CRN).
Competing interests DAO and AJS are listed as coinventors on an international patent application filled for the experimental device (PCT/AU2012/001041.2012) used in three of the included studies in this review. The authors declare no other competing interests.
Provenance and peer review Not commissioned; externally peer reviewed.