An electronic search was conducted up to May 3rd, 2023 on MEDLINE, Embase, CINAHL, AMED, PEDro, Cochrane Library, SPORTDiscus, and PsycINFO without language or date restrictions. Descriptors were related to “randomized controlled trial” and “lateral elbow tendinopathy.” To maximize the sensitivity of the search strategy and consequently avoid exclusions of potential therapies of which we are not aware, descriptors related to non-invasive therapies were not used. Detailed information on the search strategy is provided in Supplementary material 1. In addition, reference lists of previous systematic reviews and the clinical trials registers (www.clinicaltrials.gov, www.anzctr.org.au) were manually searched to maximize the identification of all eligible trials.

Randomized or quasi-randomized controlled trials (e.g., trials with allocation by hospital record number, date of birth, or alternation) that investigated the effectiveness of any non-invasive therapy on the outcomes of pain intensity, disability, maximum grip strength, and quality of life in patients with lateral elbow tendinopathy were included. The intervention of interest was defined as any non-invasive intervention, including oral medications. Lateral elbow tendinopathy was defined as constant tendon pain and reduced function related to mechanical loading of the lateral elbow tendons,23 epicondylar tenderness, and pain with resisted wrist extension and stretching of the radial wrist extensors.24 To be included, trials had to compare any non-invasive therapy with control (i.e., placebo, sham, waiting list, or no intervention) or invasive interventions (i.e., any type of injection or surgery). In this meta-analysis, placebo and sham were combined as a control group, and no intervention and waiting list were combined as another control group. Crossover randomized controlled trials were included if data from the first phase were reported separately. Trials comparing different modalities of non-invasive therapies were excluded. Pain intensity, maximum grip strength, disability, and quality of life were assessed using any valid instrument: for pain intensity (e.g., Numerical Rating Scale - NRS25 or Visual Analogue Scale - VAS25); maximum grip strength (e.g., Hand Grip Dynamometer26); disability (e.g., Disabilities of the Arm, Shoulder and Hand - DASH27 or Patient-Rated Tennis Elbow Evaluation - PRTEE28); and quality of life (e.g., Short Form Health Survey - SF-3629).

After searches, the identified references were exported to an Endnote® file and duplicates were removed. Then, two independent reviewers screened titles and abstracts, and evaluated potential full texts using the eligibility criteria outlined above. Between-reviewer discrepancies were resolved by a third reviewer.

When available, we extracted scores directly from the PEDro database (http://www.pedro.org.au/). When not available, the risk of bias of included trials was assessed by two independent reviewers using the 0–10 Physiotherapy Evidence Database (PEDro) scale.30 Discrepancies were resolved by a third reviewer. The PEDro scale is a reliable and valid tool to evaluate risk of bias of trials investigating non-invasive therapies.30-32

Two independent reviewers extracted data from included trials: source of participants; sex; type and dosage for non-invasive therapy and comparator; outcome data; and time points. The extracted outcome data included means and standard deviations (SDs) post-intervention and sample sizes for all groups of interest to investigate the short- and long-term effectiveness. The short-term effect was considered a follow-up of up to 12 weeks after randomization, and the long-term effect was considered a follow-up more than 12 weeks after randomization. When more than one time point was available in the same follow-up period, the one closer to the end of the intervention was considered. Following Cochrane's recommendations,22 when trials evaluated more than one similar non-invasive therapy or more than one form of similar comparator, outcome data of similar trial arms were combined.33 When data were not reported, attempts to contact the authors was made up to three times within a one week interval to obtain further information. In trials in which SD was not reported, missing data were calculated from 95% confidence interval (CI), standard error, p-value, baseline change, graphic representation, medians and interquartile ranges, or SD from baseline.22 When imputations were not possible, trials were excluded from the quantitative analysis. Between-reviewer discrepancies during the data extraction process were resolved by a third reviewer.

Meta-analysis was conducted using a random-effects model (DerSimonian and Laird method), when possible. Mean differences (MDs) and 95% CIs were presented for each specific non-invasive therapy in forest plots. The effect was evaluated by the Z test, and a p-value <0.05 was considered statistically significant. The clinical importance of therapies were interpreted by comparing the estimated effect sizes and 95% CI in association with the minimal clinically important difference (MCID) of the outcome of interest.34 The MCIDs considered were: 19 points on the 0–100 points scale for pain intensity35,36; 17 kgs (kg) for maximum grip strength37; and 11 points on the 0–100 points scale for disability.15,36,38

Data were converted to a common scale before being combined in the meta-analysis. The common scale ranged from 0 to 100 points for pain intensity, disability, and quality of life.22 For maximum grip strength, different units of measurement were converted to kg.22

To reduce clinical heterogeneity, we chose to group only non-invasive interventions with similar protocols. Otherwise, studies were renamed according to the experimental group and analyzed individually.39,40

Searches identified 4810 references. Following removal of duplicates 2904 titles and abstracts were screened, 141 potential full texts were assessed using the eligibility criteria, and 52 original trials were included in the review. Of the 52 included trials, 47 trials were included in the quantitative analysis. Five trials were excluded from the quantitative analysis because outcome data were not reported, and imputations were not possible.45-49 The flow of studies through the review is in Figure 1.

Fig. 1.

PRISMA flowchart.

(0.35MB).

Fifty-two randomized controlled trials published between 1985 and 2022 were included in the qualitative analyses. Trials were conducted in Asia (n = 22, 42.3%),47-68 Europe (n = 22, 42.3%),33,39,45,46,69-86 North America (n = 3, 5.7%),87-89 Oceania (n = 3, 5.7%),40,90,91 South America (n = 1, 1.9%),92 and one multicenter trial conducted concurrently in Asia, Europe, and Oceania (n = 1, 1.9%).93 Sample sizes of included trials ranged from 12 to 199 participants.

Twenty-two different non-invasive therapies were investigated in the 47 trials: non-pharmacological therapies (n = 37)50-62,65-68,71-74,76-81,83-89,91-93; combination of two or more non-pharmacological therapies (n = 7)39,40,56,57,63,82,90; combination of non-pharmacological and pharmacological therapy (n = 1)56; and pharmacological therapies (n = 6).33,63,64,69,70,75 Thirty-four (72.3%) of the included trials compared non-invasive therapy with control (i.e., no intervention, waiting list, placebo, or sham) and 17 trials (36.2%) compared non-invasive with invasive intervention (i.e., autologous blood injection, corticosteroid injection, leech therapy, neural therapy, non-steroidal anti-inflammatory drugs injection, prolotherapy injection, and tenotomy). Pain intensity was investigated in 38 of 47 trials (80.9%),33,39,40,50,52-67,69-76,78,79,81,82,85,86,90,91,93,94 maximum grip strength was investigated in 21 trials (44.7%),39,50,52,54-59,63,65-67,69,72,73,80,81,84,88,89 disability was investigated in 31 trials (65.9%),33,39,40,50,51,53-55,58-62,66-70,72,73,77,81-84,86,88,90,91,93,94 and quality of life was investigated in five trials (10.6%).40,54,69,83,87 All trials included in the quantitative analysis investigated only short-term effectiveness (i.e., ≤12 weeks after randomization). Supplementary material 2 shows detailed characteristics of the included trials (n = 52).

The mean risk of bias of the 52 included trials was 7.0 points on the 0 to 10 points PEDro scale, with scores ranging from 3 to 9 points. Of the 47 trials included in the quantitative analysis, 37 trials (78.7%) were considered to have low risk of bias (i.e., PEDro score ≥6 out of 10). The main methodological issues of included trials were absence of concealed allocation (n = 22, 47.8%), absence of participant blinding (n = 31, 66.1%), and absence of therapist blinding (n = 43, 91.5%). In addition, 31 trials (66.1%) did not report intention-to-treat analysis (Supplementary material 3).

Effect estimates are presented as MDs on a scale of 0–100 points (i.e., pain intensity, disability, and quality of life) or kilograms (i.e., maximum grip strength). We found no high certainty of evidence for the effectiveness estimates of non-invasive interventions for pain intensity, disability, maximum grip strength, or quality of life in this review. The main reasons for downgrading the certainty of the evidence were imprecision (88 of 88 comparisons, 100%), inconsistency (82 of 88 comparisons, 93.2%), and risk of bias (31 of 88 comparisons, 35.2%). None of the estimates were downgraded due to publication bias and indirectness. Supplementary material 4 shows complete forest plots of the analyses for interventions rated as moderate, low, and very low certainty of evidence.

Non-invasive therapy versus control (i.e., placebo, sham, waiting list or no intervention)

For pain intensity in the short term, moderate certainty evidence suggests that betamethasone valerate (BMV) medicated plaster likely results in a slight reduction in pain when compared with placebo (MD −10.7, 95% CI: −18.4, −3.0, two trials,33,70n = 261 participants). The estimated effect for BMV medicated plaster did not reach clinical relevance (i.e., 19 points on the 0–100 point scale). Low certainty evidence shows that extracorporeal shockwave therapy (ESWT) does not reduce pain intensity when compared with placebo or sham in the short-term (MD −3.3, 95% CI: −11.9, 5.1, three trials,72,81,85n = 194 participants) (Figure 2).

Fig. 2.

Summary of Moderate, Low, and Very Low-Certainty of Evidence on Pain Intensity in Lateral Elbow Tendinopathy.

Dashed line indicates the minimum clinically important difference. BMV, betamethasone valerate medicated plaster; ESWT, extracorporeal shockwave therapy; NSAIDs, nonsteroidal anti-inflammatory drugs; US, ultrasound therapy.

aDowngraded owing to imprecision: less than 400 participants included in the meta-analysis or 95% CI includes no effect and crosses SMD = 0.5 or MD > 10%. bDowngraded owing to inconsistency: I² statistic was higher than 50%, absence of overlap between CI or pooling was not possible. cDowngraded owing to risk of bias: more than 25% of the participants in the meta-analysis were from trials with a high risk of bias (i.e., PEDro score <6 out of 10).

(0.26MB).

For maximum grip strength in the short-term, moderate certainty evidence suggests that ESWT likely results in little increase in strength when compared with placebo or sham (MD 4.9 kg, 95% CI: 1.2, 8.6, four trials,72,81,84,88n = 311). The estimated effect for ESWT did not reach clinical relevance of 17 kg. Low certainty evidence shows that laser therapy does not increase strength (MD 5.0 kg, 95% CI: −2.3, 12.3, two trials,54,89n = 76 participants) when compared with placebo or sham (Figure 3).

Fig. 3.

Summary of Moderate, Low, and Very Low-Certainty of Evidence on Maximum Grip Strength in Lateral Elbow Tendinopathy.

Dashed line indicates the minimum clinically important difference. ESWT, extracorporeal shockwave therapy; NSAIDs, nonsteroidal anti-inflammatory drugs; US = ultrasound therapy.

aDowngraded owing to imprecision: less than 400 participants included in the meta-analysis or 95% CI includes no effect and crosses SMD = 0.5 or MD > 10%. bDowngraded owing to inconsistency: I² statistic was higher than 50%, absence of overlap between CI or pooling was not possible. cDowngraded owing to risk of bias: more than 25% of the participants in the meta-analysis were from trials with a high risk of bias (i.e., PEDro score <6 out of 10).

(0.16MB).

For disability, moderate certainty evidence shows that BMV medicated plaster likely reduces disability (MD −6.7, 95% CI: −11.4, −2.0, two trials,33,70n = 261 participants) when compared with placebo in the short-term. The estimated effect may reach clinical relevance of 11 points on the 0–100 point scale. Also, with an effect that may reach a clinical relevance, low certainty suggests that acupuncture reduces disability when compared with sham in the short-term (MD −9.1, 95% CI: −11.7, −6.4, three trials,62,77,93n = 178 participants) (Figure 4). Low certainty evidence shows that ESWT does not reduce disability (MD 2.9, 95% CI: −5.9, 11.7, four trials,72,81,84,88n = 311 participants) when compared with placebo in the short-term. Very low certainty evidence for quality of life is reported in Figure 5.

Fig. 4.

Summary of Moderate, Low, and Very Low-Certainty of Evidence on Disability in Lateral Elbow Tendinopathy.

Dashed line indicates the minimum clinically important difference. BMV, betamethasone valerate medicated plaster; ESWT, extracorporeal shockwave therapy; NSAIDs = nonsteroidal anti-inflammatory drugs; US = ultrasound therapy.

aDowngraded owing to imprecision: less than 400 participants included in the meta-analysis or 95 % CI includes no effect and crosses SMD = 0.5 or MD > 10 %. bDowngraded owing to inconsistency: I² statistic was higher than 50 %, absence of overlap between CI or pooling was not possible. cDowngraded owing to risk of bias: more than 25 % of the participants in the meta-analysis were from trials with a high risk of bias (i.e., PEDro score <6 out of 10).

(0.23MB).

Fig. 5.

Summary of Very Low-Certainty of Evidence on Quality of Life in Lateral Elbow Tendinopathy.

Dashed line indicates the minimum clinically important difference. ESWT, extracorporeal shockwave therapy; NSAIDs, nonsteroidal anti-inflammatory drugs.

aDowngraded owing to imprecision: less than 400 participants included in the meta-analysis or 95% CI includes no effect and crosses SMD = 0.5 or MD > 10%. bDowngraded owing to inconsistency: I² statistic was higher than 50%, absence of overlap between CI or pooling was not possible. cDowngraded owing to risk of bias: more than 25% of the participants in the meta-analysis were from trials with a high risk of bias (i.e., PEDro score <6 out of 10).

(0.11MB).

Detailed qualitative sensitivity analyses conducted by removing high risk of bias trials (i.e., PEDro score <6 points out of 10) to investigate potential impact of poor methodological quality trials on the estimated effectiveness are presented in Supplementary material 5. Significant effect changes were found for laser therapy to reduce pain (MD −9.1, 95% CI: −16.3, −1.9, one trials,54n = 61 participants) and disability (MD −14.6, 95% CI: −25.2, −3.9, one trial,54n = 61 participants) when compared to sham in the short-term. However, such effect estimates are based on comparisons with very low certainty evidence, making the results uncertain.

This systematic review used strong methodological rigor, following the Cochrane recommendations to investigate the effectiveness of non-invasive therapy in lateral elbow tendinopathy. We performed a careful analysis classifying all interventions according to the evidence established by the GRADE classification system, which makes our review transparent and reliable.

However, our review presents some potential limitations. Due to the variety of identified interventions, it was often not possible to group included trials. In most cases, interventions entitled “physical therapy” addressed different therapies in the experimental group, making grouping impossible. Thus, we chose to rename the interventions according to the protocol applied and group those that were similar, reducing clinical heterogeneity. Relatedly, the small number of trials included in the meta-analysis made it impossible for us to explore heterogeneity and publication bias. Although several individual trials have produced statistically and clinically relevant positive results, they have all been downgraded to very low certainty of evidence due to imprecision (i.e., small sample sizes) and inconsistency (i.e., pooling was not possible).