TENS looked useless in the pooled data until the analysis accounted for intensity

Pooled across 29 studies, transcutaneous electrical nerve stimulation had no statistically significant effect on chronic low back pain (d+ = 0.16, P > 0.050, across 64 effect estimates) [s1]. Split according to whether the current was delivered at an intensity the patient could clearly perceive, the picture changes completely: one subgroup showed a large effect (d+ = 0.97, 95% CI 0.65–1.30, across 12 estimates), the other showed nothing (d+ = 0.30, −0.11 to 0.72, across 6 estimates) [s1].

That is either the explanation for two decades of contradictory TENS reviews, or a subgroup analysis in a literature the same authors describe as carrying a high risk of bias and substantial heterogeneity [s1]. Both readings are defensible, and the honest position is that it depends.

The back pain analysis

The review screened 1,162 records and included 29 studies, assessing study quality with the PEDro scale and the RoB-2 tool [s1]. Its distinguishing feature is that it treated the stimulation parameters — intensity, frequency, session count, electrode placement — as moderator variables rather than averaging over them.

Only one moderator mattered. Appropriate intensity was associated with larger effects than inappropriate intensity, a difference of roughly 0.67 in standardised effect size (F(1,16) = 7.27, P = 0.016) [s1]. Frequency, number of sessions and electrode placement showed no significant impact [s1].

The authors define the effective condition as sensory-level intensity that is clearly perceptible and titrated throughout the session, and state that fixed intensity, and stimulation strong enough to elicit muscle contractions, are generally not recommended [s1]. Their own summary of certainty is low, because of high risk of bias and substantial heterogeneity across the included trials [s1].

Titration is the mechanistically interesting part. Sensory nerves accommodate to a constant stimulus, so a device left at one setting for thirty minutes delivers a diminishing perceived stimulus. A trial that fixed the intensity at the start is testing a different intervention from one that raised it as the patient adapted — and both have historically been reported as "TENS".

The knee analysis disagrees about which parameter matters

A parallel dose-response meta-analysis in knee osteoarthritis pooled 36 randomised trials covering 2,518 participants [s2]. Across all trials, TENS produced a small but statistically significant reduction in pain, larger against control or placebo comparators and smaller against other active treatments — which the authors read as reflecting effective co-interventions in the comparison arms rather than a failure of TENS [s2].

Here the moderator that mattered was session duration, with greater pain relief when stimulation lasted 40 minutes or longer [s2]. Frequency, intensity and number of sessions showed no significant moderation [s2].

So one review finds intensity decisive and duration irrelevant; the other finds duration decisive and intensity irrelevant. In different conditions, with different trial sets, and with low-certainty evidence underneath both. The knee review's own conclusion asks for higher-quality sham-controlled trials with complete parameter reporting, and with pain assessed during or immediately after stimulation, to quantify the specific analgesic effect beyond contextual influences [s2].

"Beyond contextual influences" is the crux for any device you hold against your skin and feel working. A sensation that is clearly perceptible is also a strong cue that something is happening, which is exactly the condition under which expectation effects are largest.

Percussive devices

Massage guns are marketed for both recovery and pain. A meta-analysis of 12 randomised trials of percussive therapy after exercise-induced muscle damage separates those two claims cleanly [s3].

On performance recovery, percussive therapy improved countermovement jump recovery compared with control (13 effect estimates, Hedges' g = 0.78, 95% CI 0.26 to 1.29, P < 0.01, low certainty) and reduced creatine kinase levels (7 estimates, g = −0.87, −1.57 to −0.17, P = 0.02, very low certainty) [s3].

On pain, it did nothing measurable. Delayed-onset muscle soreness showed no clear improvement (17 estimates, g = 0.14, −0.19 to 0.48, P = 0.40, very low certainty), and neither did maximum voluntary contraction recovery (8 estimates, g = 0.12, −0.10 to 0.34, P = 0.28, moderate certainty) [s3]. Exploratory dose analyses suggested that longer treatment was associated with larger jump effects — and also with higher soreness scores [s3]. The authors state explicitly that these dose-related findings are preliminary and should not be read as prescriptive thresholds [s3].

Note the certainty grades. The one outcome rated moderate certainty is a null result; the positive findings are low and very low [s3]. That is the opposite of the pattern a reader would want before buying a device for pain.

Heat, the least glamorous option

The American College of Physicians guideline for acute and subacute low back pain recommends superficial heat, and it is the only item in that recommendation carrying moderate-quality evidence — massage, acupuncture and spinal manipulation, listed alongside it, are rated low-quality [s4].

A hot water bottle has the best evidence grade in this article. It is also the cheapest thing in it.

What a reader can take from this

TENS is not nothing, and it is not reliably something. The one condition under which it produced a large effect in the back pain data — a clearly perceptible current, titrated as the session runs — is a description of what was done in trials that worked, not a recommendation to anyone [s1]. The knee data identify a different parameter [s2]. Both bodies of evidence are graded low certainty by their own authors [s1] [s2].

Percussive devices have a defensible case for short-term performance recovery and no measurable effect on soreness [s3].

What to watch

Whether anyone runs the sham-controlled, fully parameter-reported TENS trials both reviews ask for [s1] [s2]. Until then, the device literature will keep producing reviews that reach opposite conclusions from overlapping trial sets, which is what a field looks like when the intervention has not been standardised.

This article is informational and is not medical advice. Parameters described here are trial characteristics, not instructions for use.

Sources

• Dose-Response Effects of Transcutaneous Electrical Nerve Stimulation for Chronic Low Back Pain: A Systematic Review and Meta-Analysis — European Journal of Pain, 2026-01-30
• Dose-response effects of transcutaneous electrical nerve stimulation for knee osteoarthritis: A systematic review and meta-analysis — Osteoarthritis and Cartilage Open, 2026-02-28
• Effects of percussive therapy dosages on recovery from acute exercise-induced muscle damage: a systematic review and meta-analysis — Chiropractic & Manual Therapies, 2026-06-12
• Noninvasive Treatments for Acute, Subacute, and Chronic Low Back Pain: A Clinical Practice Guideline From the American College of Physicians — Annals of Internal Medicine, 2017-02-13