REF-1149 · Clifford et al., 2020 · BMJ Open Sport & Exercise Medicine · Paper · PubMed · Free full text

Effectiveness of isometric exercise in tendinopathy

Across 10 small RCTs (294 participants, five tendon sites) isometrics were not superior to isotonic exercise for pain, either straight after a session (3 patellar trials pooled, p = 0.19) or over up to 12 weeks; limited evidence (level 3), with responses varying within and across tendons

A-fairly strong evidenceMeasured on a mixed group

Clifford and colleagues reviewed every randomised trial that compared isometric exercise with another treatment or no treatment in tendinopathy. They found 10 (294 participants): patellar (4), rotator cuff (2), lateral elbow (2), Achilles (1) and gluteal (1), published 2015-2020. Three were of good overall quality and seven poor.

For pain straight after a single session in patellar tendinopathy (Rio 2015, Rio 2017, Holden 2020), a random-effects meta-analysis found no significant difference between isometric and isotonic exercise (p = 0.19).

Over up to 12 weeks in chronic tendinopathy (van Ark 2016, patellar; Clifford 2019, gluteal), isometrics and isotonic exercise improved pain, function, satisfaction and quality of life similarly. In acute rotator cuff tendinopathy isometrics did no better than ice.

Adding isometrics to isotonic work helped in one lateral elbow trial (with more than twice the loading time) but not in one Achilles trial. All conclusions are limited evidence (level 3).

The authors conclude that isometrics are not superior to isotonic exercise for chronic tendinopathy, that the response is variable, and that isometrics can be one part of a progressive loading programme. The large single-session effect of Rio 2015 (REF-627) "has not been replicated".

Included studies

10 randomised trials, n = 294, mean age 39.2 (range 16-86); two crossovers (Rio 2015, Holden 2020). Comparators: isotonic exercise (7 trials), ice (2), wait-and-see (1). Two trials in acute tendinopathy (≤12 weeks of symptoms), seven chronic, one unspecified. Treatment from a single session to 3 months; follow-up from 45 min to 3 months.
Patellar (4 trials, 76 participants): Rio 2015 (REF-627, n=6, crossover), Rio 2017 (REF-1145), van Ark 2016 (REF-1146), Holden 2020 (REF-1147, n=21, crossover).
Rotator cuff (2, 63): Parle 2017 (isometric vs ice vs both, 1 week); Dupuis 2018 (isometric vs ice for 2 weeks, then isotonic for both).
Lateral elbow (2, 74): Vuvan 2020 (isometric vs wait-and-see, 8 weeks); Stasinopoulos 2017 (isometric added to eccentric-concentric, tennis players).
Achilles (1, 44): Gatz 2020 (eccentric with vs without isometrics, 3 months; no external load).
Gluteal (1, 30): Clifford 2019 (isometric vs isotonic, 12 weeks, pilot) — the review's first author's own trial.
Isometric dose: not pooled or analysed by dose. The patellar single-session trials used Rio-type quadriceps holds (5 × 45 s at 70% MVIC in Rio 2015 and Holden 2020); the weeks-long trials were mostly unsupervised home programmes, and in van Ark 2016 and Clifford 2019 time under tension matched the isotonic arm.
Quality: good — Holden 2020, Vuvan 2020, Dupuis 2018; poor — the other seven (Cochrane risk of bias plus external validity and precision). Only four trials had a power calculation.

Key Results

ComparisonTrialsResultEvidence level
Isometric vs isotonic, pain immediately after one session, patellarRio 2015, Rio 2017, Holden 2020pooled: no significant difference (p = 0.19)3 (limited)
— per trial, isometric minus isotonic change on 0-10 pain (Table 6)Rio 2015 / Rio 2017 / Holden 2020−4.3 (−1.2 to −7.4) / −0.9 (−1.1 to −0.7) / +0.3 (1.3 to −0.7)—
— mean pain reductions isometric vs isotonic (Discussion)Rio 2015 / Rio 2017 / Holden 20206.8 vs 2.5 / 1.8 vs 0.9 / 0.8 vs 1.1—
Isometric vs isotonic, ≤12 weeks, chronic (patellar, gluteal)van Ark 2016, Clifford 2019no difference in pain, function, satisfaction or quality of life; pain effects −0.5 (−2.6 to 1.6) and +0.7 (−0.7 to 1.7)3
Isometric vs ice, acute rotator cuff, ≤6 weeksParle 2017, Dupuis 2018no difference on any outcome3
Isometric added to isotonic vs isotonic alone, ≤12 weeksStasinopoulos 2017 (elbow), Gatz 2020 (Achilles)pain: favoured adding isometrics in the elbow trial only, −1.3 (−0.8 to −1.9); function: no difference3 (conflicting)
Isometric vs wait-and-see, lateral elbow, 8 weeksVuvan 2020lower PRTEE score; grip strength and global rating of change no different (86% vs 63% improved, not significant)3
Immediate strength and cortical inhibition, isometric vs isotonicRio 2015 onlyhigher after isometrics3 (single poor-quality trial)
Heterogeneity rule: forest plots only when I² ≤ 75% and only within one tendon type; the I² of the one pooled analysis is not given in the text.
The elbow "benefit": the combined programme ran 56 min of loading per session against 22 min in the comparison arms, so more loading, not isometrics as such, may explain it (the authors' point).

Mechanisms

Authors' interpretation: when time under tension is matched, contraction type may matter less, because the tendon is loaded and lengthens whatever the muscle does.
Authors' interpretation: exercise-induced hypoalgesia acts partly through descending pain inhibition; central sensitisation, present in some tendinopathy, may explain why some people get more pain with isometric or isotonic loading. Not measured in the included trials.
Authors' interpretation: progressive loading needs intensity as well as time under tension (the Achilles add-on trial used bodyweight only and found nothing).
Inference: the one large positive result (Rio 2015) is also the smallest and earliest trial, which is the pattern expected from small-study effects.

Practical Application

Are isometrics better than slow lifting for tendon pain? 10 small RCTs found no edge: not right after a session (3 patellar trials, p = 0.19, 2 of them positive) nor over 12 weeks. Use holds to load a tendon, not as a painkiller.
What isometrics are good for here: as a loading form in a progressive programme, equal to isotonic work when time under tension is matched (two trials). The authors suggest they may help selected individuals.
Do not promise immediate relief: the large single-session effect came from 6 male volleyball players and was not reproduced; responses vary within and across tendons. Not from this paper; the base's own practice: try holds, keep them if that athlete's pain settles, drop them if it flares.
Not from this paper; the base's own practice: for Achilles and patellar tendinopathy, heavy slow resistance or eccentric loading carries the long-term evidence (REF-626, REF-625); training continues under the pain-monitoring rule (REF-628, SYNTH-73).

Limitations

Few, small, mostly poor-quality trials; every conclusion is limited evidence (level 3). Only one meta-analysis was possible; the rest is vote-counting by statistical significance.
Results are generalised across tendons on the assumption that they respond alike; the short-term isometric-vs-isotonic conclusion mixes one patellar and one gluteal trial.
Cohort and extraction issues: Table 2 gives Rio 2017 van Ark 2016's cohort (n=29, 13 vs 16) and Table 1 gives both trials identical rows, while the Rio 2017 abstract reports 10 vs 10; with the text's "the same participants as their previous study", this is either a tabulation error or a sign that the two papers report one trial (REF-1145, REF-1146).The text also says Holden 2020 found no fall in pain from baseline, while Holden reported a 0.9-point fall immediately after exercise, the same for both exercises. Neither error changes the comparison between contraction types.
Most trials had no untreated control, so time and natural recovery cannot be separated from treatment.
English-language trials only; publication bias not assessed (too few trials).
The first author led one of the included trials (Clifford 2019, gluteal).
Funding: Medical Research Council UK (MR/R020515/1). Competing interests: none declared.
How this entry was checked and changed
Why this level

systematic review, but of 10 small trials (seven rated poor quality, three good), mostly synthesised by vote-counting; only one comparison was pooled (3 patellar trials, immediate pain), and the review mis-tabulates two of its patellar trials (Limitations); read from the full text

Full text (checked 2026-10-16)
Ten eligible studies with a total of n=294 participants; mean age 39.2 years (range 16–86) (full text).
Rotator cuff (n=2 studies, 63 participants), lateral elbow (n=2 studies, 74 participants), patellar (n=4 studies, 76 participants), Achilles (n=1 study, 44 participants), gluteal (n=1 study, 30 participants) (full text).
Three studies of 'good' overall quality and seven of 'poor' quality (full text).
Forest plot, isometric vs isotonic, immediate pain: no significant difference between the two interventions (p=0.19) (full text).
Table 6 treatment effects: Rio (2015) −4.3 (−1.2 to −7.4); Rio (2017) −0.9 (−1.1 to −0.7); Holden (2020) +0.3 (1.3 to −0.7) (full text).
Discussion: Rio 2015 6.8 vs 2.5 points; Rio 2017 1.8 vs 0.9; Holden 0.8 vs 1.1; Pearson 1.7 points (full text).
Table 7: van Ark (2016) −0.5 (−2.6 to 1.6); Clifford (2019) +0.7 (−0.7 to 1.7); Stasinopoulos (2017) −1.3 (−0.8 to −1.9) (full text).
Combined elbow programme 56 min of loading per session compared with 22 min for the other two programmes (full text).
Table 2 lists Rio et al35 as n=29, isometric (n=13), isotonic (n=16) (full text).
Funding: Medical Research Council UK (MR/R020515/1); "Competing interests: None declared" (full text).