Lateral elbow tendinopathy · Review · 2023
Does shockwave ease tennis elbow pain in the long run?
In this review of 16 trials, people treated with shockwave reported a little less elbow pain a year or more later than people in control groups (5 trials, 537 people). Results for grip strength were less clear-cut.
Helped pain a little more than control groups- trials in the review
- 16
- people in the 5 trials of shockwave for pain
- 537
- months of follow-up
- 12–28
Good to know: The review did not describe the shockwave settings; trials followed people for 12 to 28 months.
Chen Q, Shen P, Zhang B, et al. Journal of Plastic Surgery and Hand Surgery, 2023 · Read the paper ↗
How this study worked
What happened
- trials
- 5
- people
- 537
- Compared with
- No active treatment or a placebo, depending on the trial
Long-term elbow pain
Standardised difference (no units) · a year or more after treatment
Show the numbers
| Difference (standardised difference (no units)) | likely range | |
|---|---|---|
| Pain, a year or more later | −0.19 | −0.36 to −0.02 |
What the trials used
- ESWT parameters
- No pooled device, focused/radial modality, pulse count, energy/pressure, frequency, target site, or session schedule is provided; included trials used study-specific protocols.
- Exercise and injection protocols
- No common exercise program, corticosteroid formulation/dose, or PRP preparation/injection technique is synthesized.
Who took part
People with tennis elbow in 16 trials, each following people for 12 to 28 months.
Side effects
The review did not report side effects.
All the numbers and the full appraisal
The results, in numbers.
People with lateral epicondylitis in 16 English-language RCTs. Individual study follow-up ranged from 12 to 28 months in Table 2. The synthesis does not report a single pooled age, setting, diagnostic definition, or participant profile.
Study-specific pain/function instruments standardized as SMD8 reported outcomes
ESWT vs control — long-term pain
Study-specific pain/function instruments standardized as SMD standard deviation units (SMD)
ESWT vs Control; negative/positive direction depends on source instrument coding.
5 studies · 537 participants in this estimate · I²=37%; Q-test p=0.18.
Common-effect model; no covariate-adjusted pooled estimate reported. Lower SMD indicates less pain with ESWT in the source analysis.
Statistically significant pooled reduction in pain; modest standardized estimate and heterogeneous source trials.
Analysis details and source
Lower SMD indicates less pain with ESWT in the source analysis. No common native-unit pooled mean was reported; the review reports SMD. Pooled analyzed counts were 273 treatment and 264 comparator participants (total 537); arm Ns refer to the forest-plot comparison and not study-level arms.
ESWT vs control — long-term function
Study-specific pain/function instruments standardized as SMD standard deviation units (SMD)
ESWT vs Control; negative/positive direction depends on source instrument coding.
3 studies · 240 participants in this estimate · I²=4%; Q-test p=0.35.
Common-effect model; no covariate-adjusted pooled estimate reported. The Abstract and the Results text print the CI as -0.02 to -0.49, an invalid reversed interval; Figure 3B reports -0.02 to 0.49.
The confidence interval includes zero; the pooled result does not establish a difference or equivalence.
Analysis details and source
The Abstract and the Results text print the CI as -0.02 to -0.49, an invalid reversed interval; Figure 3B reports -0.02 to 0.49. No common native-unit pooled mean was reported; the review reports SMD. Pooled analyzed counts were 119 treatment and 121 comparator participants (total 240); arm Ns refer to the forest-plot comparison and not study-level arms.
Exercise vs control — long-term pain
Study-specific pain/function instruments standardized as SMD standard deviation units (SMD)
Exercise vs Control; negative/positive direction depends on source instrument coding.
6 studies · 565 participants in this estimate · I²=80%; Q-test p=<0.001.
Random-effects model; no covariate-adjusted pooled estimate reported. Common-effect model differs: SMD -0.22 (95% CI -0.38 to -0.05), p=0.01; authors selected random effects because I²=80%.
The confidence interval includes zero; the pooled result does not establish a difference or equivalence.
Analysis details and source
Common-effect model differs: SMD -0.22 (95% CI -0.38 to -0.05), p=0.01; authors selected random effects because I²=80%. No common native-unit pooled mean was reported; the review reports SMD. Pooled analyzed counts were 270 treatment and 295 comparator participants (total 565); arm Ns refer to the forest-plot comparison and not study-level arms.
Exercise vs control — long-term function
Study-specific pain/function instruments standardized as SMD standard deviation units (SMD)
Exercise vs Control; negative/positive direction depends on source instrument coding.
5 studies · 545 participants in this estimate · I²=0%; Q-test p=0.52.
Common-effect model; no covariate-adjusted pooled estimate reported. Function pooled from PFFQ (four RCTs) and DASH (one RCT); the authors excluded the one RCT reporting pain-free grip strength because higher values on that measure indicate better function.
The confidence interval includes zero; the pooled result does not establish a difference or equivalence.
Analysis details and source
Function pooled from PFFQ (four RCTs) and DASH (one RCT); the authors excluded the one RCT reporting pain-free grip strength because higher values on that measure indicate better function. No common native-unit pooled mean was reported; the review reports SMD. Pooled analyzed counts were 260 treatment and 285 comparator participants (total 545); arm Ns refer to the forest-plot comparison and not study-level arms.
Corticosteroid injection vs control — long-term pain
Study-specific pain/function instruments standardized as SMD standard deviation units (SMD)
Corticosteroid injection vs Control/placebo; negative/positive direction depends on source instrument coding.
4 studies · 425 participants in this estimate · I²=96%; Q-test p=<0.001.
Random-effects model; no covariate-adjusted pooled estimate reported. Pain pooled from VAS and Likert measures; high heterogeneity.
The confidence interval includes zero; the pooled result does not establish a difference or equivalence.
Analysis details and source
Pain pooled from VAS and Likert measures; high heterogeneity. No common native-unit pooled mean was reported; the review reports SMD. Pooled analyzed counts were 210 treatment and 215 comparator participants (total 425); arm Ns refer to the forest-plot comparison and not study-level arms.
Corticosteroid injection vs control — long-term function
Study-specific pain/function instruments standardized as SMD standard deviation units (SMD)
Corticosteroid injection vs Control/placebo; negative/positive direction depends on source instrument coding.
4 studies · 425 participants in this estimate · I²=67%; Q-test p=0.03.
Random-effects model; no covariate-adjusted pooled estimate reported. Function pooled from PFFQ and DASH; high heterogeneity.
The confidence interval includes zero; the pooled result does not establish a difference or equivalence.
Analysis details and source
Function pooled from PFFQ and DASH; high heterogeneity. No common native-unit pooled mean was reported; the review reports SMD. Pooled analyzed counts were 210 treatment and 215 comparator participants (total 425); arm Ns refer to the forest-plot comparison and not study-level arms.
Platelet-rich plasma vs control — long-term pain
Study-specific pain/function instruments standardized as SMD standard deviation units (SMD)
PRP vs Control; negative/positive direction depends on source instrument coding.
3 studies · 245 participants in this estimate · I²=78%; Q-test p=0.01.
Random-effects model; no covariate-adjusted pooled estimate reported. Pain pooled from VAS and Roles–Maudsley; high heterogeneity.
The confidence interval includes zero; the pooled result does not establish a difference or equivalence.
Analysis details and source
Pain pooled from VAS and Roles–Maudsley; high heterogeneity. No common native-unit pooled mean was reported; the review reports SMD. Pooled analyzed counts were 120 treatment and 125 comparator participants (total 245); arm Ns refer to the forest-plot comparison and not study-level arms. The Results text gives the PRP pain heterogeneity as I²=67% (Q-statistic p=0.03); Figure 4C prints I²=78% (p=0.01), the values stored here.
Platelet-rich plasma vs control — long-term function
Study-specific pain/function instruments standardized as SMD standard deviation units (SMD)
PRP vs Control; negative/positive direction depends on source instrument coding.
3 studies · 174 participants in this estimate · I²=79%; Q-test p=0.009.
Random-effects model; no covariate-adjusted pooled estimate reported. Function pooled from DASH and Roles–Maudsley; high heterogeneity.
The confidence interval includes zero; the pooled result does not establish a difference or equivalence.
Analysis details and source
Function pooled from DASH and Roles–Maudsley; high heterogeneity. No common native-unit pooled mean was reported; the review reports SMD. Pooled analyzed counts were 86 treatment and 88 comparator participants (total 174); arm Ns refer to the forest-plot comparison and not study-level arms. The Results text gives the PRP function heterogeneity as I²=78% (Q-statistic p=0.01); Figure 4D prints I²=79% (p=0.009), the values stored here.
How treatment was delivered
No pooled device, focused/radial modality, pulse count, energy/pressure, frequency, target site, or session schedule is provided; included trials used study-specific protocols.
No common exercise program, corticosteroid formulation/dose, or PRP preparation/injection technique is synthesized.
The meta-analysis does not provide pooled adverse-event or tolerability results. Harm incidence is unknown from this report.
Authors note few eligible long-term RCTs, original trials conducted in Western countries only, and possible lower quality for function outcomes based on participant-reported instruments.
The review restricted eligibility to English-language RCTs with downloadable data, ≥12-month follow-up, and ≥3 eligible trials in an intervention category; these criteria can limit representativeness.
The Abstract and the Results text report ESWT-function SMD 0.24 with CI -0.02 to -0.49, an invalid reversed interval; Figure 3B reports -0.02 to 0.49. The figure estimate and interval are retained with the conflict stated.
No pooled treatment parameters, exact common follow-up timepoint, or pooled harms are reported.
Findings checked against the full source by a separate automated reviewer.
Limitations noted by the reviewers
- Authors note few eligible long-term RCTs, original trials conducted in Western countries only, and possible lower quality for function outcomes based on participant-reported instruments.
- The review restricted eligibility to English-language RCTs with downloadable data, ≥12-month follow-up, and ≥3 eligible trials in an intervention category; these criteria can limit representativeness.
- The Abstract and the Results text report ESWT-function SMD 0.24 with CI -0.02 to -0.49, an invalid reversed interval; Figure 3B reports -0.02 to 0.49. The figure estimate and interval are retained with the conflict stated.
- No pooled treatment parameters, exact common follow-up timepoint, or pooled harms are reported.
Source-linked critical reading
Meta-analysis of long-term comparative trials
Technical source record
Complete publisher PDF (7 pages) read; all methods/results/discussion, Tables 1–2, Figures 1–4, captions and disclosures checked. Figures 3–4 were visually inspected. The abstract and Figure 3B conflict over the ESWT function CI; Figure 3B gives the valid ordered interval -0.02 to 0.49.
The review suggests ESWT may provide a small long-term pain benefit compared with control, but does not show a clear function benefit. The other pooled treatment categories had uncertain long-term effects. Study protocols and controls differed, so these averages do not select a specific shock-wave device or regimen.
Among people with lateral epicondylitis followed for at least 12 months, what did randomized-trial syntheses show for ESWT, exercise, corticosteroid injection, and platelet-rich plasma versus control?
Systematic review and meta-analysis of English-language randomized trials identified in PubMed and Embase through March 2023. The authors retained 16 studies; analyses were conducted separately by treatment and outcome when at least three eligible trials supplied data. Follow-up eligibility was ≥12 months.
People with lateral epicondylitis in 16 English-language RCTs. Individual study follow-up ranged from 12 to 28 months in Table 2. The synthesis does not report a single pooled age, setting, diagnostic definition, or participant profile.
The review compared four treatment categories: extracorporeal shock-wave therapy (ESWT), exercise-based physiotherapy, corticosteroid injection (CI), and platelet-rich plasma (PRP). It does not pool device, focused/radial modality, pulse count, energy/pressure, frequency, session schedule, injection product, or shared co-interventions. The ESWT category should therefore not be read as a single reproducible treatment protocol.
Eligible RCTs compared conservative treatments with no active treatment or placebo/control. The actual control varied across trials and treatments; the synthesis does not provide a common comparator or treatment-contact match.
Pain measures included VAS, Likert scales, and Nirschl Phase Rating Scale; function measures included grip strength, pain-free grip strength, DASH, PFFQ, and Roles–Maudsley score. The authors standardized scores as SMDs. Follow-up was at least 12 months, but exact timepoint varied by study.
Pooled analyses found no statistically significant long-term pain or function difference for exercise, corticosteroid injection, or PRP (random-effects models, except a common-effect model for exercise function). ESWT was associated with lower pain scores than control (SMD -0.19, 95% CI -0.36 to -0.02; p=0.03; 5 trials, 537 participants), but not a significant function difference (SMD 0.24, 95% CI -0.02 to 0.49; p=0.07; 3 trials, 240 participants). The abstract prints an impossible reversed ESWT-function CI (-0.02 to -0.49); Figure 3B gives -0.02 to 0.49. Exercise pain depended on the random-effects model because I²=80%: SMD -0.21 (95% CI -0.60 to 0.18; p=0.29), while its common-effect estimate was statistically significant.
This review does not report pooled adverse-event counts or comparative tolerability for the four treatment categories. Harm incidence is unknown from this synthesis.
The authors noted few long-term trials, all original data came from Western countries, and many function outcomes were self-reports/questionnaires. Eligibility was restricted to English RCTs with downloadable data, at least 12-month follow-up, and treatment categories represented by ≥3 RCTs. The abstract/forest-plot CI discrepancy for ESWT function is unresolved; the review also gives no pooled device settings or harms.
These pooled estimates describe long-term averages across different protocols, controls, instruments, and follow-up points. They do not establish which ESWT type or dose is useful for an individual patient or specific tendon presentation. The review’s ESWT pain result is modest and based on five trials.
Future reviews should preregister methods, include broader language and database coverage, obtain protocols and study-level outcome data, distinguish ESWT modalities and regimens, use consistent outcome direction/timepoints, examine risk of bias and small-study effects, and report harms.
Read the original article.
Long-term effectiveness of conservative management for lateral epicondylitis: a meta-analysis.
Chen Q, Shen P, Zhang B, Chen Y, Zheng C. · Journal of plastic surgery and hand surgery · 2023
Open the publication record or available full-text files.
Checked against the full paper · Clinical review: Dr. Marcus Yu Bin Pai, · How we review →