Altered Strength Profile in Achilles Tendinopathy: A Systematic Review and Meta-Analysis

Search Strategy and Study Selection

The review was registered on the PROSPERO database (CRD42015025386) and has been reported in accordance with the PRISMA statement for systematic reviews.²² The following databases were searched between February and April 2016 by 2 authors (S.M., J.H.) independently using an agreed-upon set of key words: Academic Search Complete, AMED, Biomedical Reference Collection, MEDLINE, CINAHL, SPORTDiscus, Web of Science, Embase, and Scopus. The search strings are shown in Table 1. The 2 reviewers conducted the database searches independently using prespecified inclusion and exclusion criteria. They independently compiled “short lists” of suitable abstracts and compared their respective short lists before reaching agreement on potentially relevant abstracts. A third reviewer (K.O.) reviewed the short-listed abstracts, and any disagreement was discussed among the 3 reviewers. The primary author (S.M.) screened the agreed-upon abstract list and obtained full texts of the studies that met the inclusion criteria to create a final list. The final list was confirmed by one of the authors (K.O.) to ensure that the studies met the inclusion criteria and did not meet the exclusion criteria.

Table 1 Key Words for Search Strategy

View Fullscreen

Risk of Bias Assessment

The Critical Appraisal Skills Programme Case-Control Study Checklist was used to assess the risk of bias in the included studies.²³ This checklist contains 12 questions; questions 7 and 12 are guidance questions and were not rated. Therefore, the included studies were appraised using 10 questions. A list of criteria for each question and the justification for providing the indicated score are outlined in Supplemental Tables 1 and 2 (available online at http://dx.doi.org/10.4085/1062-6050-43-18.S1). Two authors (S.M., M.O.) scored the studies independently using the criteria outlined, with any disagreements in scoring mediated by a third reviewer (K.O.). Given that the Critical Appraisal Skills Programme checklist was originally designed as an educational tool in a workshop setting, no overall quality score was awarded to the included studies. Instead, the strengths and weaknesses of each study were noted based on these specific criteria.

Data Extraction

The following data were extracted from each study by 2 reviewers (S.M., J.H.): characteristics of the participants (sample size, sex, and age) and details of the comparative assessment (groups compared, strength variable investigated, and summary of the results). Study results were extracted according to the subcategories for strength as outlined previously. This included data on (1) maximal strength (peak torque [PT], maximal voluntary isometric contraction, or peak force [in newtons]), (2) reactive strength (distance and height during hopping or jumping movements), and (3) explosive strength (RFD and ground reaction force [GRF]). Data relating to the mechanical properties of the tendon (eg, tendon stiffness, leg stiffness, tendon strain) or data on endurance (electromyography studies) or related concepts were not extracted.

Statistical Analysis

For all studies, where possible we computed the Hedges g effect size as a summary measure that is comparable across independent- and matched-groups study designs.²⁴ If the standard deviation of the differences was not reported in studies examining differences between the injured and uninjured limbs, it was estimated using the following formula²⁵:

where r = 0.7 was used as a conservative estimate of the within-participants correlation as recommended by Rosenthal.²⁶ Hedges g and associated 95% confidence intervals (CIs) were computed manually using the formulae provided by Borenstein et al.²⁴ A positive effect size indicated the investigated strength measure was greater for the uninjured limb or asymptomatic group. The magnitude of Hedges g effect sizes was interpreted using the Cohen²⁵ convention as small (0.2), medium (0.5), or large (0.8).

Studies were pooled for the meta-analysis according to similarities in study characteristics and methods. Given the range of variables used in isokinetic dynamometry, the results were further pooled according to the mode of contraction (eccentric or concentric) and speed of contraction (fast [>120°/s] or slow [≤120°/s]). Hedges g effect sizes were pooled using random-effects models. We selected random-effects models a priori to account for expected differences in study characteristics. Heterogeneity between studies was assessed using the I² statistic, which summarizes the percentage of total variation across studies due to differences between studies rather than chance. An I² value of 30% or less indicates low heterogeneity, with cutoffs of I² greater than 30% and I² greater than 50% indicative of moderate and substantial heterogeneity, respectively.²⁷ For some outcomes of interest, the data could not be pooled because of heterogeneity in the measures, the methods used, or because only 1 study examined an outcome of interest (eg, eccentric PF PT as a percentage of BW [BW%]) at 180°/s). In these situations, such results are reported using weighted means. All meta-analyses were carried out using the Excel (version 2010; Microsoft Corp, Redmond, WA) spreadsheets of Neyeloff et al.²⁸

Identification of Studies

The electronic search yielded a total of 7133 potentially relevant studies. After the title and abstract of each study were screened, 29 full-text studies were identified as potentially relevant (Figure 1). A total of 11 studies were removed after screening of the full texts.^3,29–38 Searching the reference lists of these full-text studies led to the addition of 1 study.³⁹ The final number of studies reviewed was 19.

View Figure

• Download as Powerpoint
• Download Figure

Participants

A detailed description of the selected studies, listed alphabetically, is provided in Supplemental Table 3. The mean ages of the participants (range = 24–59 years) were similar among studies. A total of 13 studies^{11,19,40–50} included both male and female participants, whereas 6 studies^{13,39,51–54} included men only. The authors of 8 studies^{13,39,42,43,48–51} compared strength measures between symptomatic and asymptomatic participants, the authors of 8 studies^{11,40,41,44–47,52} compared strength values between injured and uninjured sides of the same participants, and the authors of 4 studies^48,53–55 compared strength variables between the injured and uninjured sides of the same participant and between the injured side and asymptomatic participants. Strength values between the “most” and “least” symptomatic sides were compared in 1 study.¹⁹ In relation to the characteristics of the included studies comparing the injured limb with the asymptomatic controls, the control groups were generally similar with regard to sex and age; only Firth et al⁴³ reported variability in sex between groups. The average duration of symptoms among participants (range = 5 weeks–37 months) was reported in 15 of 19 studies.

Outcome Measures

Maximal-strength values using isokinetic dynamometry were reported in 8 studies.^{11,39–41,46,47,49,54} Large variations were present in the speed and mode of contraction. Researchers investigated PF PT at slow speeds (≤120°/s) in 7 studies^{11,39–41,47,49,54} and at fast speeds (>120°/s) in 7 studies.^{11,39–41,46,47,54} Maximal isometric-strength measures were reported in 3 studies^13,45,48 using various force apparatuses. Silbernagel et al¹⁹ used a variety of heel-raise tests to quantify both concentric and eccentric PF strength and heel-raise–test outcomes using a customized spring-loaded linear encoder. Authors investigated explosive-strength variables in 7 studies,^{42,45,49–53} reported values for RFD in 2 studies,^45,52 and addressed GRF in 5 studies.^{42,49–51,53} For reactive-strength variables, various single-legged hop or jumping variables were compared in 4 studies.^19,43,44,52

Risk of Bias

The risk of bias of the included studies using the Critical Appraisal Skills Programme checklist is shown in Table 2. One of the main weaknesses involved the reporting of statistical findings. Whereas most authors provided actual values with accompanying P values, few provided 95% CI statistics. A further limitation was the considerable variation in the measurement techniques used by some researchers. Among the 4 studies^42,49,51,53 in which GRF was investigated, a large amount of variability was present in the methods used to measure GRF. For example, some authors used an instrumented treadmill to measure GRF, and others used force plates integrated in running tracks across a variety of capture distances. Furthermore, a large number of variables were reported for GRF across studies, such as horizontal braking force, vertical impact force, vertical loading rate, and vertical impulse. This variation in reported variables may have accounted for the variation in findings, leading to difficulty drawing conclusions from this outcome. Another limitation was the wide range of symptom durations. The variations in symptom duration suggested that considerable variation within the populations studied may have been present; however, this may also have had the advantage of increasing the generalizability of the findings. Whereas most researchers who compared strength between the injured limb and asymptomatic controls ensured matched control groups, Firth et al⁴³ included a control group that had more men than in the symptomatic group. Finally, several investigators^41,46,53 did not provide detailed inclusion criteria.

Table 2 Risk of Bias Assessment of Included Studies

View Fullscreen

Maximal-Strength Profile

Isokinetic Dynamometry

The meta-analyses revealed moderate effect sizes for concentric PF PT fast (pooled Hedges g = 0.61; 95% CI = 0.43, 0.79; Figure 2) and concentric PF PT slow (pooled Hedges g = 0.43; 95% CI = 0.25, 0.62; Figure 3). However, effect sizes for eccentric PF PT slow were small (pooled Hedges g = 0.25; 95% CI = 0.09, 0.40; Figure 4). These differences reflected deficits of approximately 38%, 44%, and 18%, respectively, between the symptomatic and asymptomatic sides or between the symptomatic side and asymptomatic controls.

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

As outlined, the data are reported using weighted means when pooling was inappropriate because of heterogeneity in the measures, the methods used, or because only 1 study examined an outcome of interest.

Isokinetic Strength

Large effect sizes were observed for eccentric PF PT fast (Hedges g = 1.26; 95% CI = 0.34, 2.18),³⁹ moderate effect sizes were observed for eccentric PF PT BW% at 180°/s (Hedges g = 0.36; 95% CI = −0.07, 0.79),⁴⁹ and small effect sizes were observed for eccentric PF PT BW% at 60°/s (Hedges g = 0.24; 95% CI = −0.19, 0.67⁴⁹; Figure 5).

View Figure

• Download as Powerpoint
• Download Figure

Explosive-Strength Profile

Ground Reaction Force

Five studies^{42,49,50,52,53} reported inconsistent results of both increased and decreased GRF in those with AT (Hedges g = −0.73–0.66; Figure 6).

View Figure

• Download as Powerpoint
• Download Figure

Reactive-Strength Profile (Hopping)

Small to large effect sizes were reported^19,43,44,52 for reduced hop performance in those with AT using a variety of hop tests (Hedges g = 0.32–2.61; Figure 7). This reflected a difference of approximately 16% to 35% between the more and less symptomatic sides. The mean difference in distance hopped was 33% (43 cm) for the single-legged hop and 35% (151 cm) for the triple-legged hop. Average and maximal hop height during a single-legged hop demonstrated deficits of 18% (0.6 cm) and 16% (2.12 cm), respectively, on the symptomatic side. In addition, the hopping quotient (flight time/contact time) was reduced by 20% (0.1) on the symptomatic side during a single-legged hop.

View Figure

• Download as Powerpoint
• Download Figure

Heterogeneity

Statistical analysis revealed the percentage of total variation across studies was low when comparing injured and uninjured sides versus injured sides and asymptomatic controls (Figures 2–4).

Assessing Musculotendinous Function in AT

In clinical settings, the primary method used to assess PF function in individuals with AT has been the calf-raise or heel-raise test. The validity of the heel-raise test as an appropriate method for assessing PF strength in AT has been highlighted.¹⁷ However, relying on such unidimensional measures to quantify PF function may hinder the appropriate identification of functional deficits in AT, which may lead to inadequate rehabilitation and persistent symptoms. Our findings highlight the potential scope of additional methods of quantifying PF strength in AT. Deficits in maximal strength ranging from 16% to 28% were reported using weighted concentric and eccentric calf-raise tests, whereas deficits of 18% to 44% were observed when using eccentric and concentric isokinetic PT assessments. Therefore, these assessment techniques may be more appropriate for identifying strength deficits than a BW maximal-repetition heel-raise test, in which deficits appear to be much less obvious (deficits of only 8% between sides for the traditional BW heel-raise test). Unfortunately, research comparing these deficits in maximal strength, particularly isokinetic variables, with other pathologic tendon conditions is limited. We reported deficits of 5% to 12% for maximal isometric strength, which is broadly comparable with the isometric-strength deficits reported (9%–32%) for both patellar and gluteus medius tendinopathy.^55–57

Maximal-strength variables represent only 1 aspect of an individual's overall strength profile and crucially may not sufficiently assess the explosive-strength capabilities that are fundamental during sport movements, such as running or jumping. Our results also indicated a consistent trend toward reduced explosive strength (10%–21% deficit) in individuals with AT assessed by quantifying the RFD. Given that the RFD quantifies the ability to quickly produce muscle force,⁵⁸ identifying deficits may be important when assessing individuals with AT to help guide rehabilitation.

In addition to explosive strength, many athletic movements, such as running and jumping, also require substantial amounts of reactive strength to store and release energy and force. In AT, quantifying hopping ability (eg, hop distance, hop height) appears to be the most popular method of assessing reactive strength. Our results indicated deficits of 16% to 35% in individuals with AT, suggesting that these tests may be an appropriate way of assessing reactive-strength deficits in AT. Various hopping tasks have the advantage of being inexpensive, quick, and reliable in research and clinical settings⁵⁹; however, in isolation, they may not discriminate between people using very different movement patterns to achieve hop distances, and hopping does not isolate the PF muscle group. Similar magnitudes of deficits in reactive strength have also been reported after ankle⁶⁰ and anterior cruciate ligament^58,61 injury. However, reactive strength requires considerable effort and functional demands, so assessing reactive strength may not always be appropriate, depending on an individual's loading demands and current capacity. The gross nature and variability of movement strategies possible during such tasks highlight the need to combine these assessment techniques with other functional measures.⁶²

A further assessment technique that may complement the aforementioned functional outcome measures is evaluation of the tendon's mechanical properties. Achilles tendinopathy leads to alterations in the mechanical properties of the tendon, typically reduced tendon stiffness and increased strain.^51,52,63 Researchers^13,64,65 have postulated that such altered mechanical properties in AT may lead to increased strain on the Achilles tendon, contributing to the ongoing recurrence of symptoms due to an inability to dissipate forces during SSC loading activities. An emerging body of literature^55,56,64 has indicated that the tendon's mechanical response to load may be quantified in individuals with AT using methods such as shear-wave elastography or ultrasound imaging combined with isokinetic dynamometry. Although it is beyond the scope of this review, correlating the tendon's mechanical properties using traditional functional measures may identify areas to be addressed during rehabilitation.

Appropriate Comparative Groups in Assessing Strength Variables

An area of debate in assessing strength outcomes among patients with tendinopathy centers on the suitability of the uninjured side as a comparison. In their systematic review, Heales et al⁶⁶ demonstrated motor deficits in the contralateral uninjured limb of patients with unilateral tendinopathy compared with asymptomatic controls, suggesting that the uninjured side may not be as “healthy” and unaffected as a pain-free matched control. In contrast, we did not find any differences in effect sizes when comparing the injured and uninjured sides or when comparing the injured side with asymptomatic matched controls. One potential explanation may relate to the characteristics of the studies included in the review by Heales et al⁶⁶; researchers in 18 of the 20 studies investigated motor deficits in upper limb tendons, which may limit the generalizability to the lower limb. A further consideration is the influence of limb dominance in strength comparisons. In sport environments that require unilateral-dominant movements (eg, jumping in volleyball or basketball), athletes may have a favored jumping limb, which can complicate comparisons. Further research is warranted to improve our understanding of appropriate comparative groups when assessing strength in patients with tendinopathy.

Current Rehabilitation Programs and Adequately Addressing Deficits in Achilles Tendon Function

Exercise or strength interventions using repetitive concentric-eccentric PF muscle exercises have become the cornerstone of conservative treatment for AT.⁶⁷ The most popular exercise intervention has been the Alfredson heel-drop program, which is characterized by progressive, twice-daily, eccentric-only contractions over a 12-week period.¹¹ The popularity of other strengthening programs, most notably progressive mixed concentric-eccentric loading and mixed-contraction, heavy, slow resistance training, has grown.^{12,38,55,68,69} Regardless of the mode of contraction, strength interventions have been reasonably effective in improving pain and disability in those with midportion AT; the average reduction in pain was approximately 55%.^{34,68,70–78}

Despite the relative success of loading interventions for improving pain and disability, AT is associated with a high recurrence rate (27%).^20,35 One potential reason for the high recurrence rate and persistence of symptoms may be the nonresolution of the strength deficits associated with AT. The degree of improvement in PF strength (maximal-, explosive-, and reactive-strength outcomes) after loading interventions has been reported for only a few strengthening interventions for AT. Yu et al⁷⁹ and Alfredson et al⁴¹ noted improvements in concentric PF PT of 10% to 15%, respectively, after the completion of a 12-week strengthening intervention. However, such reported improvements were less than the deficits we observed in isokinetic maximal-strength variables (up to 44%). Such nonresolution of deficits after strength interventions was reiterated by Silbernagel et al,¹² who demonstrated that full symptomatic recovery did not ensure full recovery of muscle-tendon function in patients with AT. Comparisons of outcomes at baseline and 1 year after a loading intervention revealed that only 4 (25%) of the 16 patients (67%) who had fully recovered had achieved an acceptable level of muscle function, which was defined as having normal (≥90%) capability across the test battery. These findings suggested that individuals with AT continued to display strength deficits despite reduced pain and disability. The persistence of strength deficits could be speculated to result in the tendon's inability to withstand the desired load, potentially accounting for the high recurrence rate associated with AT. In fact, the “one-size-fits-all” method of assessing and rehabilitating Achilles tendon function may fail to adequately address deficits not only in maximal-strength variables but also throughout the entire strength spectrum.^12,80 Alfredson eccentric loading has become the mainstay of conservative treatment for chronic AT.⁶⁷ The preference for eccentric-only interventions reflects the suggestion that eccentric training is more specific and provides a greater load via the force-velocity curve than concentric loading.⁸¹ However, evidence to support these claims appears tenuous.⁸¹ Training only 1 aspect of an individual's strength profile may not optimally improve performance across the entire strength spectrum. Few researchers have attempted to address the entire strength spectrum in an intervention study. Notably, Silbernagel et al¹² used a loading intervention aimed at improving functional outcomes across the entire strength spectrum by integrating plyometric exercises for longer than the traditional 12 weeks, which resulted in improvements in pain and disability at 1-year follow-up. This finding highlights the need for further research in this area.

Mechanistic Effects of Strength Training in Tendinopathy

Numerous mechanistic theories have been proposed to explain improvements in pain and disability associated with strengthening interventions in AT. The term mechanotherapy has been used to describe how the body converts a mechanical stimulus into a cellular response,⁸² which may influence structural alterations in the tendon due to AT (eg, disorganized collagen architecture, thinner collagen fibers, increased water content in the extracellular matrix). These structural alterations may alter a tendon's capacity to store and produce kinetic energy, affecting strength and functional performance.^13,83 Investigators^84,85 have suggested that the loading forces applied to the tendon will help stimulate the remodeling of the abnormal tendon structure. However, alterations in tendon structure may not be associated with improved clinical outcomes. In a systematic review, Drew et al⁸⁶ concluded that alterations in tendon structure did not adequately explain the response to loading interventions in tendinopathy, which is consistent with little turnover of collagen after puberty.⁸⁷ Nevertheless, tendon structure can be altered in young populations (<25 years) after loading interventions.^88,89 Whereas the evidence for the effect of loading interventions in altering abnormal tendon structure is conflicting, loading interventions may effectively improve the mechanical properties of the tendon, albeit in asymptomatic populations.⁹⁰