Massage

Massage

Massage is sold as a muscle treatment. The literature is a family of manual loads, not a single physiology. This article asks what those loads do to skeletal muscle and related tissues, and whether any important muscle-health effect is difficult to obtain by other means. Uniqueness is not assumed. It is established only if the evidence requires it. The evidence does not.

Compiled by South Beach Longevity · 20 August 2026 Copyright 2026 Series SBL-41 / SP-MASSAGE · Register A scientific article Sources peer-reviewed human trials, meta-analyses, Doppler and biopsy experiments, labelled animal mechanotherapy, and clinical pain reviews · verified NCBI records Constraint This document describes published research. It is not medical advice. No human use, dose, route, schedule, or treatment prescription is recommended anywhere in this document.

How to read this document Every finding is labelled, in the sentence that reports it, by the kind of study that produced it. A rabbit cyclic-compression protocol is not a Swedish massage hour. A vastus lateralis biopsy at 2.5 hours is not a training adaptation. A drop in soreness is not a return of force. A fascia model is not a palpated “release.” Where two results conflict, both are given. Findings are graded in place as established, strongly supported, emerging, plausible, or speculative. Two further labels mark careful absences rather than verdicts: not established, where the evidence is too thin to place a claim on the ladder at all — untested or insufficient, an absence of proof rather than disproof; and not supported, where the weight of evidence leans against a claim but stops short of a formal refutation. Nothing here is a recommendation.

01 A family of loads, not a single intervention

Massage, in this document, is manual or device-delivered soft-tissue loading intended to deform skin, subcutaneous tissue, muscle, and the connective layers between them. Weerapong, Hume, and Kolt, in the field’s still-useful mechanisms review, listed the usual justifications: more blood flow, less tension, less neural excitability, and a sense of well-being (Weerapong, Hume, and Kolt, 2005). Those justifications are hypotheses. They are not findings.

Four things follow and will be enforced.

First, massage is not one technique. Effleurage, petrissage, tapotement, friction, trigger-point pressure, Thai stretching-massage, acupuncture-massage, myofascial release, and sports-recovery strokes are different force–time histories. Collapsing them into one verb is how a low-back trial is used to sell a post-race rub.

Second, it is not a muscle-building stimulus. Resistance training loads the fibre through active tension and repeated high force. Massage loads the fibre and its matrix through passive compression and shear. The two share the word “mechanotransduction.” They do not share a hypertrophy literature.

Third, it is not a circulatory pump that the Doppler literature forgot to notice. The blood-flow claim is tested in Part Two. It fails.

Fourth, it is not medical advice and it is not a recovery prescription. This document describes published research. It recommends no session, pressure, oil, or schedule for any person.

The controlling question is narrower than the spa sentence. What does massage do to skeletal muscle? Does any of that matter for health, recovery, pain, or performance? And is any important muscle effect hard to get another way?

02 Modalities

The table is a map, not a ranking. Each row is a force recipe. The evidence attached to one row does not transfer to the next.

That table is the first required artifact. The rest of the document is the attempt to keep each row from stealing the next row’s citation.

03 Mechanical pressure and tissue deformation

Hands apply a surface pressure. The tissue under the hand is a layered, anisotropic, fluid-filled composite. Skin, adipose, deep fascia, epimysium, perimysium, and muscle fibres do not share a modulus. A force that dimples skin need not shorten a sarcomere, and a force that the therapist feels as “release” need not be plastic deformation of fascia lata.

Chaudhry, Schleip, Ji, Bukiet, Maney, and Findley built a three-dimensional finite-deformation model of human fasciae and asked what stresses would produce even 1% compression and 1% shear (Chaudhry et al., 2008). For fascia lata and plantar fascia the required forces sat outside the normal physiologic range of manual therapy. Superficial nasal fascia was different: substantial compression and shear were available at ordinary hands. Their conclusion is the one this article keeps: the palpable “release” reported over firm limb fasciae cannot be plastic deformation of those fasciae. Softer superficial layers remain candidates. The finding is a model plus comparison to earlier dense-connective-tissue experiments. It is strongly supported as a constraint on the fascia-release story. It is not a claim that nothing under the hand moves.

What does move is easier. Skin slides. Subcutaneous fat displaces. Muscle can change shape under compression. Intramuscular pressure rises locally and then falls. Those events are enough to stimulate mechanosensitive afferents. They are not enough to license “breaking adhesions.”

04 Mechanotransduction — what a passive load can say to a fibre

Mechanotransduction is the conversion of a physical load into a biochemical signal. Ingber reviewed the cell-scale version: integrins, the cytoskeleton, and nuclear connections can change transcription when the cell is stretched or compressed (Ingber, 2006). In skeletal muscle the training-relevant version is usually active tension — costameres, titin, FAK, mTORC1 — as reviewed by Hornberger and by Wackerhage and colleagues (Hornberger, 2011; Wackerhage, Schoenfeld, Hamilton, Lehti, and Hulmi, 2019). Massage is a different input: intermittent external compression and shear, often on a muscle that is not producing high force.

Crane, Ogborn, Cupido, Melov, Hubbard, Bourgeois, and Tarnopolsky performed the landmark human test (Crane et al., 2012). Eleven young men completed damaging cycling. One quadriceps received 10 minutes of massage; the other did not. Vastus lateralis biopsies were taken at baseline, immediately after the 10-minute treatment, and 2.5 hours later. Massage increased phosphorylation consistent with FAK and ERK1/2 activation, increased nuclear PGC-1α, reduced nuclear accumulation of NFκB p65, and attenuated TNF-α and IL-6. Muscle glycogen and lactate did not change. HSP27 phosphorylation fell. The authors wrote that massage “appears to be clinically beneficial by reducing inflammation and promoting mitochondrial biogenesis.” That last clause is an inference from signalling, not a measurement of new mitochondria or of restored force.

The study is a genuine human biopsy experiment. It is also one session, eleven men, one control condition (no treatment on the other leg), and no independent human replication of the PGC-1α / NFκB pattern at the same depth. The signalling is emerging. The clinical-benefit sentence is speculative.

Animal cyclic compressive loading (CCL) is the other mechanistic pillar. Butterfield, Zhao, Agarwal, Haq, and Best applied 30 minutes of CCL immediately after eccentric damage in six rabbits and reported faster recovery of function, less leukocyte infiltration, less necrosis, and lower wet weight (Butterfield et al., 2008). Haas, Butterfield, Zhao, Zhang, Jarjoura, and Best later showed a dose response in twenty-four rabbits: 0.5 Hz at 10 N for 15 minutes recovered torque better than lower magnitude or frequency; duration of 15 versus 30 minutes did not matter (Haas et al., 2013). Waters-Banker, Butterfield, and Dupont-Versteegden applied four days of CCL to uninjured rat tibialis anterior at 0, 1.4, 4.5, or 11 N and found load-dependent immune-cell shifts and a contralateral systemic signature (Waters-Banker, Butterfield, and Dupont-Versteegden, 2014). Lawrence, Van Pelt, Confides, Hunt, and colleagues applied four 30-minute CCL bouts during seven days of hindlimb suspension in adult rats: protein turnover rose, ribosome degradation was attenuated, atrophy was not prevented (Lawrence et al., 2020). Van Pelt, Lawrence, Miller, Butterfield, and Dupont-Versteegden framed the programme as “massage as mechanotherapy” (Van Pelt, Lawrence, Miller, Butterfield, and Dupont-Versteegden, 2021). Those papers are internally coherent. They are not a human training study.

The warrant from cell to clinic therefore has a gap in the middle. Passive load can talk to a fibre. Whether that talk becomes a muscle-health outcome people can use is the rest of this document.

05 The blood-flow claim

The popular sentence is that massage “increases circulation,” and that more blood is why soreness falls and recovery accelerates. The Doppler literature tested the sentence.

Shoemaker, Tiidus, and Mader administered certified-therapist effleurage, petrissage, and tapotement to the forearm flexors and the quadriceps of ten healthy people (Shoemaker, Tiidus, and Mader, 1997). Mean blood velocity and calculated flow in the brachial and femoral arteries did not rise at 5, 10, or 20 seconds or at 5 minutes. Mild voluntary contractions did: brachial flow from about 39 to 126 ml·min⁻¹, femoral from about 371 to 1087 ml·min⁻¹. The authors’ conclusion is the one this article adopts: if the therapeutic goal is elevated muscle blood flow, light exercise would be beneficial and massage would not. That result is established for conduit-artery flow during and immediately after classic strokes in healthy limbs.

Tiidus and Shoemaker had already combined the blood-flow and the force-recovery questions after eccentric quadriceps work (Tiidus and Shoemaker, 1995). Daily massage for four days did not raise femoral arterial or venous velocity above rest; light contractions did. Isometric and dynamic peak torques recovered on the same slow curve in the massaged and unmassaged legs, from about 60–70% of baseline immediately after exercise. Perceived soreness tended to be lower at 48–96 hours. Massage did not restore strength. Light muscle activity was the better circulatory stimulus.

Wiltshire, Poitras, Pak, Hong, Rayner, and Tschakovsky asked whether massage at least helps clear lactate and hydrogen ion after a 2-minute isometric handgrip at 40% of maximum (Wiltshire et al., 2010). Twelve subjects recovered under passive rest, light active recovery, or effleurage and pétrissage. At 30 seconds, venoarterial lactate differences were similar, but forearm blood flow was highest in passive rest (766 ml·min⁻¹), lower in active recovery (614), and lowest under massage (540). Ten-minute flow and lactate-efflux areas under the curve were lower with massage than with passive rest. Massage impeded removal of lactate and H⁺ by mechanically impeding blood flow. The “flush the lactic acid” sentence is not merely unsupported. In this design it is backwards.

Hinds and colleagues compared two 6-minute bouts of deep effleurage and pétrissage with rest after concentric quadriceps exercise in thirteen men (Hinds et al., 2004). End-of-period femoral-artery flow was 760 ± 256 ml·min⁻¹ after massage and 733 ± 161 after rest — not different. Skin blood flow (150 ± 49 versus 6 ± 4 arbitrary units) and skin temperature (32.2 ± 0.9 versus 31.1 ± 1.3 °C) were higher after massage. Muscle temperature, lactate, heart rate, and blood pressure were not. The authors proposed that a cutaneous flush without a rise in arterial inflow could divert flow away from recovering muscle. The skin can blush. The muscle conduit need not.

The blood-flow story therefore splits. Cutaneous warming is plausible and often observed. Conduit-artery muscle flow is established as not increased by classic massage in the studies that measured it, and can be reduced while the hand is on the limb. Increased muscle blood flow does not explain massage effects. Light contractions explain blood flow better than massage does.

06 Lymphatic claims

Sports language borrows a clinical technique. Manual lymphatic drainage (MLD) is a light, directional method used after cancer treatment for lymphoedema. Ezzo, Manheimer, McNeely, Howell, Weiss, Johansson, Bao, Bily, Tuppo, Williams, and Karadibak reviewed MLD for lymphoedema after breast-cancer treatment and treated it as a defined clinical question, not as a recovery drink for athletes (Ezzo et al., 2015). That literature does not license the claim that a sports petrissage “drains metabolites” or “clears inflammation” via lymph.

Lymphatic vessels are not veins. They fill from interstitial fluid and propel it by intrinsic contraction and external compression. A very light, directional stroke can be a rational input to a damaged lymphatic bed. A deep sports stroke that occludes a vein, as in Wiltshire’s forearm, is a different input. Using MLD citations to sell post-match recovery is a category error.

For ordinary post-exercise swelling, Zainuddin and colleagues did see a smaller rise in upper-arm circumference after massage (Zainuddin, Newton, Sacco, and Nosaka, 2005). That is a girth measurement, not a lymphoscintigram. The mechanism is unspecified. It is emerging as a short-term volume effect and speculative as a lymphatic-pump effect.

07 Autonomic effects, stress, and sleep

Moyer, Rounds, and Hannum meta-analysed 37 randomly assigned massage-therapy studies (Moyer, Rounds, and Hannum, 2004). A single application reduced state anxiety, blood pressure, and heart rate, but not negative mood, immediate pain, or cortisol. Multiple applications reduced delayed pain. The largest effects of a course were on trait anxiety and depression, with magnitudes the authors compared to psychotherapy. They argued that a medical model of massage is the wrong theory, and that a psychotherapy-like account fits the data better. That is a psychological-outcomes meta-analysis, not a muscle paper. It is strongly supported for short-term autonomic and affective change after massage. It is not a cortisol-flush finding — the single-session cortisol effect was not there.

Sherman, Cherkin, and colleagues later randomised 68 people with generalised anxiety disorder to therapeutic massage, thermotherapy, or a relaxing room for ten sessions (Sherman et al., 2010). All groups improved on the Hamilton Anxiety Rating Scale. Massage was not superior. A cheap relaxing room, packaged as treatment, did as well. The distinctive ingredient was not the stroke.

Rapaport, Schettler, and colleagues have a series of immune and neuroendocrine massage studies in healthy adults and in people with generalised anxiety. Those papers report session-dependent shifts in circulating lymphocytes, cytokines, and oxytocin-related measures (Rapaport, Schettler, and Bresee, 2010; Rapaport, Schettler, and Breese, 2012). They are small, specialised, and not a muscle-recovery literature. They are emerging as evidence that a massage hour is a systemic interoceptive event. They do not show that muscle mitochondria changed.

Sleep findings are thinner. Massage is often reported to feel soporific. Controlled sleep-architecture trials in healthy athletes are scarce in the reviewed record. Sleep improvement as a reliable, specific muscle-recovery mechanism is plausible and not established.

The autonomic package is real enough to matter for how a person feels. It is also easy to confuse with a muscle treatment. A drop in heart rate and a drop in NFκB are different sentences.

08 Does increased blood flow explain massage effects?

No. The question is answered here so it is not smuggled back later.

Shoemaker, Tiidus, and Mader showed that three classic strokes do not raise brachial or femoral flow, while light contractions do (Shoemaker, Tiidus, and Mader, 1997). Wiltshire and colleagues showed that massage can lower post-exercise forearm flow and impair lactate and H⁺ efflux relative to passive rest (Wiltshire et al., 2010). Crane and colleagues showed signalling changes without a change in muscle glycogen or lactate (Crane et al., 2012). If the effects that survive in later parts — less soreness, small ROM gains, some clinical pain relief — had blood flow as their necessary cause, the Doppler papers would have to be wrong in a consistent direction. They are not.

A cutaneous blush can still occur. A person can still feel warmer. Those are not muscle perfusion.

09 Delayed-onset soreness is not the lesion

Delayed-onset muscle soreness (DOMS) is pain and tenderness that peak a day or two after unaccustomed eccentric work. Exercise-induced muscle damage (EIMD) is the structural and functional lesion: Z-disk streaming, lost force, raised creatine kinase, swelling. Cheung, Hume, and Maxwell reviewed the older treatment literature and treated soreness and function as separable (Cheung, Hume, and Maxwell, 2003). Howatson and van Someren did the same for EIMD as a whole (Howatson and van Someren, 2008). Clarkson and Hubal mapped the human damage phenotype (Clarkson and Hubal, 2002). Paulsen, Mikkelsen, Raastad, and Peake reviewed leukocyte traffic into damaged muscle (Paulsen, Mikkelsen, Raastad, and Peake, 2012). Peake, Neubauer, Della Gatta, and Nosaka restated that muscle damage, inflammation, and adaptation are not a single knob (Peake, Neubauer, Della Gatta, and Nosaka, 2017).

The distinction is load-bearing. A treatment can lower the rating of soreness and leave the force deficit untouched. That pattern is the modal massage result.

10 Landmark human biopsy: Crane 2012, read without the press release

Crane et al. remains the paper every mitochondrial-massage claim cites (Crane et al., 2012). The design is better than the marketing.

It is a within-person contrast: both legs damaged, one massaged for 10 minutes, biopsies from both. Metabolites that would move if the treatment were a circulatory or glycogen event did not move. The signals that moved were mechanical (FAK, ERK1/2), inflammatory-transcriptional (NFκB p65, TNF-α, IL-6), a heat-shock phosphorylation (HSP27), and a mitochondrial-biogenesis transcription coactivator (nuclear PGC-1α). The time window is hours, not days of new organelle.

What the paper does not show: more mitochondria, a higher VO₂, faster force recovery, satellite-cell activation, a protein-synthesis rate, a female sample, an older sample, a trained-athlete sample, a comparison to active recovery or foam rolling, or a replication cohort. Tiidus, reviewing alternative treatments three years later, judged that massage as then practised in humans had little effect on recovery from minor exercise-induced damage (Tiidus, 2015). That is not a refutation of the biopsy. It is a reminder that a western blot is not a performance test.

Human replication of the Crane signalling set has not become a second, independent biopsy series in the reviewed record. The animal CCL programme is the nearest mechanistic neighbour, and it is a different species and a defined piston, not a therapist’s hand (Butterfield et al., 2008; Haas et al., 2013; Waters-Banker, Butterfield, and Dupont-Versteegden, 2014; Lawrence et al., 2020). Mitochondrial and NFκB findings from Crane are therefore emerging, not established, and they are not a unique human muscle-health effect until they are repeated and tied to a function other methods lack.

11 Inflammatory signalling

Tidball reviewed muscle inflammation as a repair sequence, not a toxin to be wiped away (Tidball, 2017). Chazaud described macrophages as necessary for regeneration (Chazaud, 2016). Suppressing the early inflammatory wave is not automatically “better recovery.” It can be a change in the timing of a process that still has to finish.

Crane et al. saw less nuclear NFκB and less TNF-α and IL-6 in the massaged vastus at hours (Crane et al., 2012). Dupuy, Douzi, Theurot, Bosquet, and Dugué, in a 99-study recovery-technique meta-analysis, reported overall small-to-moderate reductions in CK, IL-6, and CRP across several methods, and named massage and cold exposure as the strongest anti-inflammatory-marker techniques in that pool (Dupuy, Douzi, Theurot, Bosquet, and Dugué, 2018). Those circulating markers are not intramuscular NFκB. They are also not proof that repair improved.

Zainuddin, Newton, Sacco, and Nosaka found a lower CK peak at day 4 and less swelling after 10 minutes of massage at 3 hours, with no strength or ROM recovery advantage (Zainuddin, Newton, Sacco, and Nosaka, 2005). Hilbert, Sforzo, and Swensen found no neutrophil-count difference after 20 minutes of massage or sham at 2 hours, and no torque or ROM difference, with lower soreness intensity only at 48 hours (Hilbert, Sforzo, and Swensen, 2003). The human inflammatory picture is therefore mixed: some circulating and some biopsy signals move; the functional recovery often does not.

Animal CCL can reduce leukocyte infiltration and necrosis when applied immediately after eccentric damage (Butterfield et al., 2008). That is a rabbit finding. It is plausible as a mechanical immunomodulation hypothesis (Waters-Banker, Dupont-Versteegden, Kitzman, and Butterfield, 2014). It is not a human prescription.

12 Mitochondria, protein synthesis, satellite cells, scar

Mitochondrial biogenesis. Crane’s nuclear PGC-1α is a signalling snapshot (Crane et al., 2012). Exercise itself is the established human mitochondrial stimulus. A 10-minute post-damage massage has not been shown, in a second human laboratory, to increase mitochondrial volume, enzyme activity, or endurance performance. The biogenesis claim is emerging as signalling and speculative as an organelle or performance effect.

Muscle protein synthesis. Lawrence et al. found that CCL during hindlimb suspension increased protein turnover in adult rats but did not stop atrophy, and that the same anabolic parameters did not move under normal weight bearing (Lawrence et al., 2020). Human MPS after massage is not an established finding in the reviewed record. Massage as an anabolic therapy in healthy, loaded muscle is not supported. Massage as a turnover signal during disuse is an animal result.

Satellite cells. Mauro named the cell (Mauro, 1961). Yin, Price, and Rudnicki reviewed its necessity for regeneration (Yin, Price, and Rudnicki, 2013). Snijders, Nederveen, McKay, and colleagues reviewed human satellite-cell responses to exercise (Snijders et al., 2015). None of those papers is a massage trial. The reviewed record does not contain a clean human satellite-cell massage experiment. An effect on satellite-cell activity is speculative.

Scar and remodelling. Severe crush and contusion have a fibrosis literature. Sports massage is almost never studied in that model. Tiidus noted that conclusions about massage after more severe injury cannot yet be made (Tiidus, 2015). Tissue-remodelling benefits of sports massage are speculative.

13 Recovery of force

This is the functional question that marketing treats as already answered.

Tiidus and Shoemaker: no strength difference through 96 hours (Tiidus and Shoemaker, 1995). Zainuddin et al.: no strength or ROM recovery advantage, despite ~30% less peak soreness (Zainuddin, Newton, Sacco, and Nosaka, 2005). Hilbert et al.: no peak-torque difference (Hilbert, Sforzo, and Swensen, 2003). Weerapong, Hume, and Kolt: post-exercise massage reduces soreness severity but has no effects on muscle functional loss (Weerapong, Hume, and Kolt, 2005). Davis, Alabed, and Chico, in 29 randomised studies and 1,012 participants: no evidence that massage improves strength, jump, sprint, endurance, or fatigue; small improvements in flexibility and DOMS (Davis, Alabed, and Chico, 2020). Poppendieck, Wegmann, Ferrauti, Kellmann, Pfeiffer, and Meyer, in 22 trials: performance-recovery effects “rather small and partly unclear,” larger for very short recovery after mixed high-intensity work (g = 0.45 within 10 minutes; g = 0.08 beyond 20 minutes), larger in untrained than trained, and of questionable justification for competitive athletes (Poppendieck et al., 2016).

Dupuy et al. found massage the most powerful technique in their pool for DOMS and perceived fatigue (Dupuy, Douzi, Theurot, Bosquet, and Dugué, 2018). Perceived fatigue is not force. The two meta-analyses can both be true.

Force recovery after EIMD is established as not reliably accelerated by massage in the human designs that measured it. Short-term “performance recovery” in Poppendieck is a different, smaller, time-locked claim and remains modest.

14 Muscle-biopsy and mechanistic matrix

The matrix is the second required artifact. The human rows do not add up to a unique muscle-health adaptation. The animal rows add up to a research programme that has not yet been repeated as a human biopsy series against active recovery.

15 DOMS table

Soreness reduction is strongly supported. Treating that reduction as accelerated repair is the usual error.

16 Pain perception

Melzack and Wall proposed gate control: large-fibre input can inhibit nociceptive traffic in the dorsal horn (Melzack and Wall, 1965). A firm stroke is a large-fibre event. The theory is old and incomplete, but it is a better first mechanism for a 20-minute drop in pain than “increased circulation.”

Clinical pain is a different literature from DOMS.

Furlan, Giraldo, Baskwill, Irvin, and Imamura updated the Cochrane review of massage for nonspecific low-back pain: 25 trials, 3,096 participants, evidence graded low to very low (Furlan et al., 2015). Versus inactive controls, massage reduced short-term pain in acute back pain (one small trial) and reduced short-term pain (SMD −0.75; 7 trials, n = 761) and function (SMD −0.72; 6 trials, n = 725) in subacute and chronic pain, without a long-term advantage. Versus active controls, pain favoured massage in the short term (SMD −0.37) and long term (SMD −0.40), with no functional difference. Serious adverse events were not reported; increased pain occurred in 1.5–25%. The authors’ own sentence is the one to keep: they had very little confidence that massage is an effective treatment for low-back pain. That is strongly supported as a description of the evidence quality, and emerging as a short-term analgesic effect.

Qaseem, Wilt, McLean, Forciea, and the ACP committee recommended nonpharmacologic options for acute or subacute low-back pain, listing superficial heat (moderate-quality evidence) and massage, acupuncture, or spinal manipulation (low-quality evidence) among the choices patients might select (Qaseem, Wilt, McLean, and Forciea, 2017). For chronic low-back pain the moderate-quality first list was exercise, multidisciplinary rehabilitation, acupuncture, and mindfulness-based stress reduction. This document is not a treatment guideline. It records that a major professional society was willing to name massage as a low-quality-evidence option for acute back pain, not as a muscle-remodelling therapy.

Cherkin, Sherman, Kahn, Wellman, Cook, Johnson, Erro, Delaney, and Deyo randomised 401 adults with chronic nonspecific low-back pain to structural massage, relaxation massage, or usual care (Cherkin et al., 2011). At 10 weeks, Roland Disability Questionnaire scores were 2.9 points lower with relaxation massage and 2.5 lower with structural massage than with usual care (clinically meaningful threshold set at 2 points). Symptom bothersomeness fell 1.7 and 1.4 points (threshold 1.5). The two massage types did not differ. A small functional advantage for relaxation massage persisted at 52 weeks. Participants were not blinded to receiving massage. The trial is the best single clinical experiment in the reviewed record. It says that some massage hours beat usual care for chronic back pain. It does not say that “releasing” a lumbar fascia was the reason — the relaxation arm did as well.

Perlman, Sabina, Williams, Njike, and Katz found WOMAC improvements after an eight-week Swedish-massage course for knee osteoarthritis versus delayed intervention (Perlman et al., 2006). A dose-finding trial nominated 60 minutes once weekly (Perlman et al., 2012). The later multisite trial (n = 222) showed 8-week WOMAC Global advantages versus light-touch (−8.16) and usual care (−9.55); at 52 weeks the omnibus test was not significant (Perlman et al., 2019). Short-term OA symptom relief is strongly supported. Durable disease modification is not shown.

Moyer, Rounds, and Hannum found that a single massage did not reduce immediate pain, while a course reduced delayed pain (Moyer, Rounds, and Hannum, 2004). Immediate analgesia is therefore not guaranteed. Delayed analgesia after repeated sessions is the better-supported clinical pattern.

Are soreness reductions mostly perceptual? In the sports studies, yes, in this precise sense: the rating moves more reliably than the dynamometer. That does not make the rating fake. Pain is a perception. It means the muscle has not been shown to be repaired. In clinical back and knee pain, function scores sometimes move with the pain scores, as in Cherkin et al. and the 8-week Perlman results. Those are still symptom and disability instruments, not biopsies.

17 Range of motion and flexibility

Davis, Alabed, and Chico found a small flexibility benefit of sports massage and no performance benefit (Davis, Alabed, and Chico, 2020). Weerapong, Hume, and Kolt had already listed increased compliance and decreased stiffness as a biomechanical hypothesis (Weerapong, Hume, and Kolt, 2005). Zainuddin et al. found no ROM recovery advantage after damaging eccentric work (Zainuddin, Newton, Sacco, and Nosaka, 2005). Hilbert et al. found none (Hilbert, Sforzo, and Swensen, 2003). The sports-massage ROM effect, when present, is therefore an acute mobility change in undamaged or lightly disturbed muscle, not a repair of the EIMD ROM loss.

Stretching remains the obvious comparator. Behm, Blazevich, Kay, and McHugh reviewed acute stretching: all forms improved ROM, typically for less than 30 minutes, via reduced muscle–tendon stiffness or increased stretch tolerance; static and PNF stretching produced small immediate performance costs that grew when a muscle group was stretched ≥60 s (Behm, Blazevich, Kay, and McHugh, 2016). Kay and Blazevich’s static-stretch performance review is the quantitative sibling for the cost side of that trade (Kay and Blazevich, 2012). Massage is not required to obtain a short-lived ROM gain.

Foam rolling is the self-administered cousin. Wilke, Müller, Giesche, Power, Ahmedi, and Behm pooled 26 trials: foam rolling had a large acute ROM effect versus no exercise (SMD 0.74) and was not superior to stretching (SMD −0.02) (Wilke et al., 2020). Wiewelhove et al. found small pre-roll flexibility effects (g = 0.34) and small post-roll pain reductions (g = 0.47), with mostly negligible performance effects (Wiewelhove et al., 2019). MacDonald, Button, Drinkwater, and Behm reported less soreness and more ROM after rolling, and interpreted the evoked-twitch pattern as more neural and connective than contractile (MacDonald, Button, Drinkwater, and Behm, 2014). Pearcey, Bradbury-Squires, Kawamoto, Drinkwater, Behm, and Button, in eight men, reported less tenderness and smaller decrements in some dynamic tests after rolling (Pearcey et al., 2015). That last sample is small.

18 Fascia “release”

Fascia is real. Findley, Chaudhry, Stecco, and Roman reviewed it as an innervated, vascular, load-bearing network rather than an inert wrapping (Findley, Chaudhry, Stecco, and Roman, 2012). Langevin has shown that connective tissue can transduce needle twist and stretch (Langevin, 2014). None of that is a licence to treat “release” as a structural event.

Chaudhry et al. is the constraint (Chaudhry et al., 2008). Plastic deformation of fascia lata and plantar fascia at 1% compression or shear demanded forces outside physiologic manual therapy. The sensation therapists call release, over those firm layers, is therefore not a measured plastic strain of those layers. Neural change, fluid displacement, and deformation of softer superficial tissue remain available explanations. Ajimsha, Al-Mudahka, and Al-Madzhar reviewed myofascial-release trials and found a literature still short of mechanistic honesty (Ajimsha, Al-Mudahka, and Al-Madzhar, 2015).

Structural plausibility of fascia “release,” as advertised, is low for deep, dense fasciae and plausible only as a name for a sensory and superficial-tissue event. Cherkin et al. already showed that a relaxation massage, which does not pretend to remodel lumbar fascia, matched structural massage on disability (Cherkin et al., 2011).

19 Performance

Davis, Alabed, and Chico is the cleanest sports-performance synthesis: 29 randomised studies, 1,012 people, no evidence for strength, jump, sprint, endurance, or fatigue improvement (Davis, Alabed, and Chico, 2020). Poppendieck et al. allow a small, short-lived exception after mixed high-intensity work when the next test is minutes away (Poppendieck et al., 2016). Dupuy et al. support massage for perceived fatigue and DOMS, not for a dynamometer (Dupuy, Douzi, Theurot, Bosquet, and Dugué, 2018). Weerapong, Hume, and Kolt had already said that between-event massage was widely used and thinly supported (Weerapong, Hume, and Kolt, 2005).

Pre-event massage as a performance enhancer is not supported. Post-event massage as a force-restoration method is not supported. Post-event massage as a way to feel less sore is strongly supported.

20 Performance-recovery matrix

21 ROM and flexibility matrix

Massage is not a unique ROM tool. Stretching and rolling already occupy the same short-lived niche.

22 Sleep and stress, returned to as outcomes

Moyer, Rounds, and Hannum remain the quantitative core for anxiety, blood pressure, and heart rate (Moyer, Rounds, and Hannum, 2004). Sherman et al. show that a relaxing room can match massage for GAD (Sherman et al., 2010). Rapaport’s immune series shows that a massage hour is not physiologically silent (Rapaport, Schettler, and Bresee, 2010). None of these findings requires a muscle explanation, and none of them is a sleep-polysomnography programme in athletes. Stress-reduction after massage is strongly supported as a short-term state change. A unique, muscle-mediated sleep benefit is not established.

23 Active recovery, stretching, foam rolling, compression, rest

Active recovery. Shoemaker, Tiidus, and Mader already showed that light contractions raise conduit flow and massage does not (Shoemaker, Tiidus, and Mader, 1997). Wiltshire et al. showed that massage can be worse than passive rest for metabolite clearance, with active recovery in between (Wiltshire et al., 2010). Mika and colleagues, in the active-recovery lactate literature, treated light exercise as the standard way to raise muscle perfusion after hard work (Mika, Mika, Fernhall, and Unnithan, 2007). If the goal is blood flow, massage is the inferior instrument.

Stretching. Behm, Blazevich, Kay, and McHugh give ROM without a therapist (Behm, Blazevich, Kay, and McHugh, 2016). Massage’s flexibility effect in Davis et al. is the smaller cousin of that literature (Davis, Alabed, and Chico, 2020).

Foam rolling. Wiewelhove et al. and Wilke et al. show the same outcome shape as sports massage: some ROM, some less pain, little performance (Wiewelhove et al., 2019; Wilke et al., 2020). Siegel, Afonso, Thomas, and a large international survey recently documented that practitioners still believe in performance and “fascial adhesion” effects the evidence does not support (Siegel et al., 2026). The knowledge-to-action gap is the same gap this pipeline was built to refuse.

Compression garments. Hill, Howatson, van Someren, Leeder, and Pedlar meta-analysed garments for recovery and found small reductions in perceived soreness and CK with limited performance restoration — the same split massage shows (Hill, Howatson, van Someren, Leeder, and Pedlar, 2014). Dupuy et al. listed compression with massage and immersion as a fatigue-management method (Dupuy, Douzi, Theurot, Bosquet, and Dugué, 2018).

Cold-water immersion. Leeder, Gissane, van Someren, Gregson, and Howatson reviewed CWI as another soreness-and-CK method with mixed functional effects (Leeder, Gissane, van Someren, Gregson, and Howatson, 2012). Dupuy et al. grouped cold exposure with massage for inflammatory-marker reduction. CWI is not massage. It occupies the same recovery-marketing shelf.

Rest. Most EIMD force returns with time. Tiidus and Shoemaker’s unmassaged legs recovered on the same torque curve as the massaged legs (Tiidus and Shoemaker, 1995). Rest is the control that massage frequently fails to beat on function.

24 Comparative-intervention table

25 Six questions that have to be answered in the evidence’s own words

Does increased blood flow explain massage effects? No. Classic strokes do not raise femoral or brachial flow; light contractions do (Shoemaker, Tiidus, and Mader, 1997). Massage can impede post-exercise flow and metabolite efflux (Wiltshire et al., 2010). Crane’s biopsy signals moved without a metabolite change (Crane et al., 2012).

Are reductions in soreness mostly perceptual? They are perceptual in the only honest sense of the word — they are ratings — and they are more reliable than force recovery (Zainuddin, Newton, Sacco, and Nosaka, 2005; Hilbert, Sforzo, and Swensen, 2003; Davis, Alabed, and Chico, 2020). That is not a sneer. It is a split between symptom and lesion. Clinical disability scores can move with pain, as in Cherkin et al. (2011). Those still are not myofibrillar repair.

Does massage accelerate functional recovery? Not in the human EIMD designs that measured peak torque through days (Tiidus and Shoemaker, 1995; Zainuddin, Newton, Sacco, and Nosaka, 2005; Hilbert, Sforzo, and Swensen, 2003). Sports-performance metas are null on strength, jump, sprint, and endurance (Davis, Alabed, and Chico, 2020). Poppendieck et al. allow a small, minutes-scale exception (Poppendieck et al., 2016). Rabbit CCL can speed torque recovery (Butterfield et al., 2008; Haas et al., 2013). That is not yet a human result.

Are mitochondrial and mechanotransduction findings replicated? FAK/ERK and PGC-1α / NFκB changes exist in one human biopsy paper (Crane et al., 2012). A second, independent human biopsy series of that set was not present in the reviewed record. Animal CCL replicates the broader idea that compression is biologically non-silent (Waters-Banker, Butterfield, and Dupont-Versteegden, 2014; Lawrence et al., 2020). Human mitochondrial biogenesis as an outcome is not replicated.

Is fascia “release” structurally plausible? Not for fascia lata and plantar fascia under ordinary hands (Chaudhry et al., 2008). Softer superficial layers can deform. Sensation can change. Cherkin et al. found structural and relaxation massage equivalent for chronic back disability (Cherkin et al., 2011). The structural story is not required by the best clinical trial.

Does massage have any scientifically established muscle benefit genuinely difficult to obtain through other means? No. That is the load-bearing answer. Soreness relief is shared with foam rolling, compression, cold, and active recovery. ROM is shared with stretching and rolling. Autonomic quieting is shared with a relaxing room (Sherman et al., 2010). Force restoration is not an established massage benefit. Mitochondrial biogenesis is not an established massage benefit. Fascia remodelling is not an established massage benefit. Satellite-cell activation is not shown. Human MPS is not shown. The remainder that is hardest to photocopy is the social and tactile fact of being treated by another person. That is a real human event. It is not a unique muscle-health adaptation.

26 The uniqueness verdict, without marketing language

The evidence supports a short list.

Massage can reduce the intensity of delayed soreness. Massage can produce a small, short-lived gain in flexibility. Massage can reduce state anxiety, heart rate, and blood pressure in a single session, and can reduce delayed pain across a course (Moyer, Rounds, and Hannum, 2004). Massage can beat usual care for chronic low-back disability at 10 weeks, whether the strokes are “structural” or relaxing (Cherkin et al., 2011). Massage can improve 8-week WOMAC scores in knee osteoarthritis, without a 52-week group difference (Perlman et al., 2019). Massage can change intramuscular signalling for a few hours after damage in eleven young men (Crane et al., 2012).

The evidence does not support the list that sells the hour. Massage does not raise muscle blood flow in the Doppler studies. It can lower it. It does not reliably restore force. It does not have a replicated human mitochondrial-biogenesis outcome. It does not plastically deform deep fascia at physiologic force. It does not uniquely occupy a recovery niche that active recovery, stretching, foam rolling, compression, or rest cannot enter.

Uniqueness was the question this pipeline was told not to assume. After the reviewed record, it is not granted.

Standing constraint This document describes published research. It is not medical advice. No human use, dose, route, schedule, or massage prescription is recommended anywhere in this document.

References

The numbered list is generated at build from verified NCBI MEDLINE records. Author–year citations in the prose are the keys.

28 Evidence handling

Study type is named in the reporting sentence. Animal cyclic-compression work is labelled as animal work and is not written as a human outcome. Meta-analytic effect sizes are the authors’ reported g or SMD, not a re-analysis. Cochrane’s own “very little confidence” sentence is preferred to a secondary cheer. Project 06 was queried read-only for discovery; it is not citation authority. Project 07 journalism, when reachable, is not evidence. Local Firecrawl was used for open pages; paywalled Cochrane HTML was not scraped around the wall. Citations are verified against NCBI records.

Conflicting results are left in conflict. Dupuy et al. can rank massage first for DOMS while Davis et al. find no performance effect. Both are kept. Crane et al. can change signalling while Tiidus and Shoemaker find no force recovery. Both are kept.

29 Limitations

The evidence and argument were tested against six critical perspectives applied to the peer-reviewed record. Some classic papers without recoverable database records are absent rather than invented; no commissioned or third-party plate is used.

30 Glossary

DOMS. Delayed-onset muscle soreness: a pain rating, not a histology.

EIMD. Exercise-induced muscle damage: structural and functional lesion after unaccustomed eccentric or high-force work.

Effleurage / petrissage / tapotement. Gliding, kneading, and striking strokes of classic massage.

Mechanotransduction. Conversion of load into biochemistry. Shared word; not a shared training adaptation.

PGC-1α. Transcriptional coactivator used as a mitochondrial-biogenesis signalling marker. A nuclear increment is not a new mitochondrion.

CCL / MLL. Cyclic or massage-like compressive loading in animal devices that quantify force, frequency, and duration.

Conduit flow. Blood flow in a named artery (brachial, femoral), as opposed to skin flush.

References