Lateral epicondylosis (LE) (“tennis elbow”) is an important condition of the upper extremity with an incidence of up to 4–7/1000 patients per year,1^–3 having a substantial impact on athletes and workers.4 5 A subset of patients are refractory to non-surgical therapy including relative rest, eccentric exercise and corticosteroid injections and suffer long-term pain and disability on average lasting for 6 months to 2 years, regardless of therapy.6 7 Our understanding of the pathophysiology of lateral elbow overuse injury has changed in recent years.8^–11 The pathophysiological hallmark of tendinopathy is the presence of degenerative changes, including neovascularity and disorganised collagen fibres.9 12 The precise cause of degeneration and pain in patients with a tendinopathy is not clear; mechanical, vascular, neural and “failure of healing” aetiological models have all been proposed.13

Treatment approaches for LE vary widely and lack definitive evidence. Non-steroidal anti-inflammatory drugs and corticosteroid injections have traditionally been used but have not been shown to be more effective than watchful waiting in the long term.14 15 Eccentric exercise regimens have shown some efficacy compared with age–gender–activity-matched controls, though a subcohort of patients remain refractory.16 Other non-surgical therapies have been evaluated for LE refractory to such conservative measures; none have been shown to be consistently effective.17^–19 Polidocanol, prolotherapy, autologous whole blood and platelet-rich plasma (PRP) injection therapies have reported promising outcomes for LE and other sport-related tendinopathies.

Polidocanol is a vascular sclerosant. In treating tendinopathy, it is used to sclerose areas of high intratendinous blood flow, sometimes termed “neovessels”, which are seen histopathologically12 and in vivo under high resolution ultrasound with colour Doppler. Neovascularity is thought to be associated with the underlying mechanism of LE and other overuse tendinopathies,20 21 though whether it is a causal agent in the pathophysiology of tendinopathy is not clear.22 A recent study reported that sustained sclerosis of neovascularity in LE was a good predictor of positive clinical effect at 2 years.23 Several randomised controlled trials (RCTs) and prospective case series have reported positive effects of polidocanol therapy for patellar, epicondylar and Achilles tendinopathies.24^–26

The use of prolotherapy dates to the 1930s,27 when it was developed for pain associated with presumed ligament laxity. Although several injection agents have been used, hyperosmolar dextrose and morrhuate sodium (also a vascular sclerosant) are the most popular28 and best studied agents. A recent systematic review identified 42 studies, over 50% of which evaluated prolotherapy for back pain, the remainder largely for painful conditions such as osteoarthritis and injuries associated with ligament laxity.29 Prolotherapy has also been used to treat tendinopathy of elite athletes30 and LE.31

Autologous whole blood and the blood product PRP have been used as injectants for tendinopathy with the aim of providing cellular and humoral mediators to induce healing in areas of degeneration. Autologous whole blood injections have been used for medial32 and lateral epicondylosis33 and plantar fasciitis.34

Platelet-rich plasma is prepared from autologous whole blood, which is centrifuged to concentrate platelets in plasma. The intention is to augment the native healing process at the site of pain through the action of platelet-derived growth factors (PDGFs). Platelets contain at least six PDGFs vital to bone and soft tissue healing (table 1). The basic and clinical science of PRP has been reviewed.35 Since the early 1990s PRP has been used for its purported ability to improve soft tissue healing and bone regeneration. The use of PRP is being intensely studied and reports suggest that clinical use is increasing rapidly for LE,36 rotator cuff repair,37 38 acute and chronic muscle strain, muscle fibrosis, ligamentous sprains, and joint capsular laxity (David Crane, MD, personal communication).

These therapies, which target the diseased tendon tissue directly, may interrupt the degenerative cycle associated with tendinopathy and allow the return of the native healing process, ultimately leading to improvement in clinical outcomes. None have been directly compared in any trial setting, but each has been assessed for LE. Because of the potential for prolotherapy, polidocanol, whole blood and PRP injection therapies to be effective for tendinopathies, we undertook a systematic review of the literature for each technique for the treatment of refractory LE.

METHODS

Inclusion criteria included human clinical trials of any design involving pre- and post-treatment assessment evaluating any of the four injection therapies for LE. A literature search was performed by the lead author (DR) and library staff of the following electronic databases: Medline (Ovid Web, 1950–2008 and Medline In-process & Other Non-Indexed Citations), Embase (1974–2008), CINAHL (1982–2008), the Cochrane Central Register of Controlled Trials (through third quarter 2008) and Allied and Complementary Medicine (1985–2008). Search strategies utilised the name of each intervention and the names of the injectants (table 2). Reports not cited in the databases above were searched using the Google search engine and the National Institutes of Health (NIH) CRISP electronic database using relevant anatomical descriptors and therapy names. The reference lists of identified studies were reviewed to identify potentially eligible studies. E-mail or phone contact was attempted with relevant author(s) or principal investigator(s) of included articles or abstracts when additional information was needed.

Identification of eligible studies

The titles and abstracts of all identified studies were screened by the study librarian and the lead author (DR). Studies whose title and abstract clearly indicated that the paper met criteria were reviewed. A description of excluded studies follows.

Data extraction

The data collection strategy for the methods and results sections of identified papers was determined a priori. Data from prospective case series was extracted by the lead author (DR). Two unblinded authors (DR, TB) assessed each controlled study. Data were extracted using a pre-existing technique.39 The RCT strength was assessed using the Delphi40 controlled trial internal validity assessment. Disagreements were resolved by consensus. Using these instruments, the methods and results of each study were described and an overall quality score assigned. Where possible, several measures of “effect size” at various follow-up time points were calculated by the authors. (1) “Per cent improvement” was calculated for all intervention arms as the change in pain score divided by the baseline score multiplied by 100%. (2) Cohen’s d was used as published or, if not provided, was calculated according to one of two formulas: (a) for controlled trials, d  =  [M₁ − M₂]/SD_pooled, where M₁, M₂ were the means in the two groups, and SD_pooled  =  √ [(SD₁2 + SD₂2)/2]; (b) for pre–post assessments, d  =  [mean of the pre–post difference]/[SD of the mean]; if data for formula “b” was not available, then formula “a” was applied, when possible. 3) In one study,33 effect size was estimated based on the Z value of the Wilcoxon Signed Rank Test according to the formula: Z/√N, where N equals the number of observations over the two time points.41 An overall evidence grade for each technique was assigned by the lead author (DR) based on the Strength of Recommendation Taxonomy criteria.42

RESULTS

The search identified 21 reports as possibly relevant; the abstract of each was reviewed. Nine studies met the eligibility criteria. Excluded studies were: general reviews (three), surgical (one), basic science (one), for different indications (three) or editorials (two), or assessed a non-reviewed therapy (one). Differences in the methods of the four therapies and in outcome measures used prevented pooling of data. Heterogeneity and inter-rater agreement of controlled study quality scoring were not formally assessed. First authors were queried about ambiguous elements of their studies. There was no significant disagreement between the raters.

Study outcomes

We identified nine eligible reports, four controlled studies and five prospective case series evaluating effects of prolotherapy (three), polidocanol (two), autologous whole blood (three) and PRP (one) therapies for LE assessing 208 adult subjects (tables 3,4). The common primary inclusion criteria were elbow pain for a minimum of 2 months and being refractory to one or more conservative therapies. The subjects were 19–66 years old, diagnosed with LE due to a variety of work and recreational activities. No study reported the inclusion of elite athletes. Subjects had an average pain duration ranging from 2 to 102 months. Subjects in three studies25 33 43 underwent ultrasound evaluation as part of the diagnosis and injection protocol and had areas of structural change and neovascularisation within the common extensor tendon. The controlled studies25 31 36 44 were of moderate to high quality, scoring 5–9/9 on the Delphi assessment. All controlled studies compared active solution with a comparison solution with either vasoactive control agents (lidocaine/adrenaline or bupivacaine/adrenaline)25 36 or saline.31 The primary outcome of each study was pain on a visual analogue scale (VAS) or pain questionnaire,44 though the denominator and specific issue addressed varied slightly. Improvement ranged from 51%25 to 94%45 for the active groups compared with baseline status; Cohen’s d effect size ranged from 0.6833 to 6.68,31 indicating a strong effect. Cohen’s d was not calculated for some studies due to insufficient data.25 36 45 46

In a prospective case series43 of subjects receiving ultrasound-guided polidocanol treatment, VAS scores improved by 37% at 3 months (p<0.05) and by 55% at 8 months (p<0.01). Grip strength significantly improved at 3 and 8 months. Structural defects and vascularity on ultrasound were improved at 8 months. In a subsequent double-blind RCT, Zeisig et al25 compared polidocanol with vasoconstrictive lidocaine/adrenaline injections. Both groups improved their VAS pain scores at 3 months, without significant differences between the groups. Three months following the first treatment session, subjects in both study groups who were unsatisfied with clinical results were offered an additional injection session with polidocanol. Follow-up at 12 months after enrolment showed that additional polidocanol injections improved VAS scores by 51% and 47% compared with baseline in subjects receiving initial polidocanol and adrenaline injection respectively.

In two RCTs,31 44 and one prospective case series,45 subjects received either prolotherapy or normal saline. Active subjects in Scarpone et al reported improvement of 90% at 16 weeks compared with 22% for controls (p<0.001), with four prolotherapy subjects reporting complete pain resolution. In a small RCT, Glick et al, subjects reported 66% improvement on a disease-specific questionnaire compared with 11.5% for controls (p = 0.09). In a prospective case series, Lyftogt45 reported 94% improvement compared with baseline scores using a novel subcutaneous injection technique (p<0.05).

Three prospective case series assessing autologous whole blood were identified.33 46 47 Each study reported significant (p<0.05) improvement compared with baseline: Edwards et al reported 88%,47 Gani et al reported 64%46 and Connell et al33 reported a median score of 0.

In a non-randomised controlled trial36 comparing a single treatment session of PRP with control injections, PRP subjects improved by a mean of 81% by 27 weeks. PRP subjects were further followed to a mean of 25.6 months, at which point the authors reported 93% pain reduction compared with baseline. Controls reported 17% improvement at 4 weeks; three of five control subjects dropped out before the 8 week follow-up and the remaining two control subjects were not followed further.

Secondary outcome measures also improved in all eight studies. Mishra et al36 reported significant improvement on the Mayo Elbow-Performance Index. Each study assessing whole blood injections reported significant improvement in Nirschl scores.33 46 47 Zeisig et al25 and Scarpone et al31 reported significant improvement in maximal grip strength compared with baseline in the intervention groups. Scarpone et al also reported improved isometric strength in the prolotherapy group compared with controls. Zeisig et al43 and Connell et al33 reported decreased structural defects and neovascularity on ultrasound, though these were not reliably correlated with clinical gains. Treatment satisfaction on single-item assessments was reported by 78% of the “polidocanol only” subjects at 12 months,25 by 93% of the PRP subjects at 25.6 months,36 and by 100% of subjects in Lyftogt et al.45 Scarpone et al31 reported that prolotherapy subjects qualitatively reported maintenance of treatment effects at 12 months.

DISCUSSION

Prolotherapy, polidocanol, autologous whole blood and PRP injection therapies have received attention in the treatment of tendinopathies among elite athletes and primary care patients. This is the first systematic review to compare these techniques for a single condition, LE. Each of the studies reviewed is small, and their methodological limitations prevent a consensus recommendation on the use of any of the three therapies compared with another at this time. However, the large effect sizes reported by all studies are compelling and suggest several areas of clinical, theoretical and research interest.

Clinical implications

In a technique that places injectant on or near a degenerative area of the tendon–bone insertion, each injectant appears safe. Though not powered to detect rare local or systemic negative effects, no study evaluating any of these therapies for musculoskeletal conditions has reported serious adverse events. Two systematic reviews of prolotherapy29 48 and a study of negative consequences of prolotherapy28 reported only minor side effects consistent with injection trauma, suggesting that the prolotherapy injectants themselves are safe. Though vascular sclerosants can theoretically cause tissue necrosis, this was not reported in these studies. The transmission of bloodborne disease is a possibility in each therapy, and underscores the need for universal precautions, including the use of gloves and appropriate handling and disposal of medical waste.

With moderate-to-large effect sizes that far exceed minimal clinically relevant effect sizes for chronic pain,49 and which are sustained over 1225 31 to 25 months36 compared with baseline or comparison groups, each technique appears potentially effective for refractory LE, thus expanding treatment options for patients who have failed conservative care. In one author’s clinic (EZ) the stepwise treatment algorithm for LE is: (1) conservative measures including eccentric exercise, (2) polidocanol injections, and (3) surgery, which, in the author’s experience, is rarely required. A similar algorithm has been reported for Achilles tendinopathy.50

The ease of clinical application of these techniques varies. Each requires routine medical knowledge of diagnostic, anatomical and joint injection skills. However, the procurement and processing of the injectants, and the required assessment associated with each therapy, vary. For example, polidocanol therapy calls for ultrasound and colour Doppler exam and specialised skills to visualise increased vascularity. PRP therapy requires investment in a centrifuge and blood processing equipment, and corresponding skills. Individual characteristics of the platelet preparation differ slightly between companies based on a number of factors; several reviews are available.35 37 Prolotherapy has limited start-up costs and is the easiest to implement. However, it may be more labour-intensive; in the reviewed studies, it required three treatment sessions, whereas the other three therapies typically used one or two treatment sessions.

Theoretical implications

The cause of pain in LE and the precise mechanism of action of the four injectants are unclear. The reviewed studies offer an opportunity to address aspects of both issues. Researchers have suggested that clinical effects of these therapies may in part result from the compressive effects of injected solutions, needle trauma, and irritant effects of blood.25 29 The studies, however, suggest that the “active” group injectants themselves provide the majority of therapeutic effects. No significant volume-related effect of comparison saline or dilute adrenaline injection was found in three RCTs.31 36 44 Similarity in outcomes in Zeisig et al25 may be explained by the actions of the two solutions: vasosclerosis may have a slightly greater effect on the neurovascular milieu than does temporary vasoconstriction. However, the study design did not allow additional adrenaline injections, so whether polidocanol is more effective than adrenaline for LE remains unclear. Mishra et al found no significant effect from mild preinjection fenestration. It is unclear whether the more active fenestration performed by Connell et al influenced outcomes, as no other study reported such a preinjection procedure.

The notion that these injectants exert a biological effect independent of needle trauma or volume-related effects is consistent with clinical, animal model and in vitro evidence. Recent clinical studies documented that areas of increased vascularity are associated with painful LE,51 have significant sensory innervation, and are linked with higher concentrations of the pain modulators glutamate and calcitonin gene-related peptide.52 Sclerosing of such structures in Achilles and patellar tendinopathy has also led to reduction in pain.24 26

Prolotherapy with dextrose with or without morrhuate sodium has been reported to decrease pain and improve function in a variety of tendinopathies.30 53 54 The historical hypothesis that prolotherapy causes an inflammatory response leading to reduced tendon and ligament laxity55 has not been confirmed. Two recent studies did not detect increased inflammation or decreased laxity following prolotherapy in a rat model.56 57 Morrhuate sodium is in the same chemical class as polidocanol and likely acts as a vascular sclerosant. Animal model data support a biological effect. Rabbit medial collateral ligaments injected with morrhuate sodium were significantly stronger (31%), larger (47%), thicker (28%) and had larger collagen fibre diameter (56%) than saline-injected controls.58 Rat patellar tendons injected with morrhuate sodium were able to withstand a mean maximal load of 136% of the uninjected control tendons.59 Hyperosmolar dextrose is also a mild vascular sclerosant, though its potential effect in tendinopathy is not well understood. A hyperosmolar glucose environment has been shown to increase platelet-derived growth factor expression and upregulate multiple mitogenic factors60^–62 that may act as signalling mechanisms in tendon repair. Lyftogt has suggested that neurogenic inflammation63 may contribute to pain in LE, and that subcutaneous injections of hyperosmolar dextrose target inflamed branches of the posterior antebrachial and medial antebrachial cutaneous nerves (Lyftogt personal communication, 2008). However, neither growth factor nor “neurosclerotic” effects have been confirmed in a tendinopathy model.

PRP injections make use of activated platelets which discharge bioactive signalling molecules including three adhesion molecules and seven growth factors. A total of 21 of 28 clinical reports, largely from the maxillo-facial and wound care fields, have reported positive PRP effects on bone and wound healing. However, many studies had a small sample size and used different methods for platelet processing, thereby preventing definitive conclusions.35 Most PRP-related in vitro and animal model science reports come from the orthopaedic literature on bone healing and report a variety of cellular and growth factor effects of potential importance to tendon healing.35 Studies assessing PRP effects for soft tissue healing showed increased anabolic gene expression in horse flexor tendons64 and proliferation of tendon cells and production of VEGF.65 66 Two large animal studies have recently reported improved healing of repaired dog and porcine cruciate ligaments following PRP therapy.67 68

Research implications

Basic science research is needed to elucidate the mechanism of action for each injection therapy. Sufficiently powered clinical trials should evaluate efficacy and effectiveness of each of the therapies compared with eccentric exercise and with each other. Research would benefit from the unification of outcome measures across these studies, which should include clinically relevant, patient-reported and objectively assessed outcomes such as pain, function and disability. Assessment of tissue injury/healing-sensitive biomarkers may enhance our understanding of the processes underlying treatment efficacy. Existing preliminary data suggest that these injection therapies have a disease-modifying potential; therefore, imaging studies, such as MRI and/or ultrasound with colour Doppler, may also be useful in addressing the mechanisms by which these agents promote healing. Whether results can be generalised across different patient populations (e.g. athletes and occupational workers) remains an important question to be answered. Larger studies assessing PRP, prolotherapy and autologous stem cell injection for LE are currently in progress (personal communication: Allan Mishra, Ron Glick and David Connell).

CONCLUSIONS

Existing data for prolotherapy, polidocanol, autologous whole blood and PRP injection therapies for refractory LE suggest effectiveness, but are limited by lack of large definitive trials. These therapies appear safe and effective when performed by an experienced clinician. Positive results have been reported in case series and non-randomised and randomised studies with LE from a variety of sport and work-related causes. Future studies using validated clinical measures, and radiological, biomechanical and tissue injury/healing-responsive biomarkers, as secondary outcome measures are needed to determine whether these injection techniques can play a definitive role in a cure for LE and other tendinopathies.