RightRep EN · PT Open the app
BLOG

Can You Train With Shoulder Pain? The Umbrella-Door Audit: Calibrate the Load, Keep the Lifting (What the Research Says)

By RightRep AI Assistant · · 21 min read

The shoulder wing of this library is already built: the impingement door, the frozen shoulder door, the rotator cuff tear door, the dislocated and separated shoulder doors, the labral tear door, even the causation twin (does the bench press damage your shoulders). But every one of those doors assumes the reader holds a diagnosis. "Can I train with shoulder pain?" is asked by the reader who holds nothing but pain — an ache at the top of a press, a pinch lying on the bench, a shoulder that has stopped sleeping on one side — no MRI sentence attached, no specialist's letter, just two opposite internet scripts screaming from either side: "stop all pressing until it's pain-free" and "shoulders are supposed to hurt, train through it." The studies below were read directly; every number in this guide comes from an abstract checked against its claim. The short version: the rest script loses to a training program in this file almost everywhere it is measured, the push-through script dies on the fences below, and the honest answer is a triage step followed by a load-calibration protocol — not a yes, and definitely not a no.

Here is the shape of the file before the receipts. The operation this door's readers fear — subacromial decompression — met a training program head-to-head in a randomized trial of 90 patients and did not beat it: at twelve months the Constant score stood at 57.0 with training versus 52.7 with surgery, a non-significant difference [2]. A stricter, more specific program then cut the choice of surgery from 63% to 20% in refractory patients (OR 7.7) while adding 24 Constant-Murley points to the control program's 9 [1]. The operation itself later faced the hardest control in medicine and tied it: in the three-group CSAW trial of 313 patients, decompression did not beat arthroscopy-only placebo surgery (Oxford Shoulder Score difference −1.3, 95% CI −3.9 to 1.3) [3], and five-year MRI follow-up showed acromioplasty does not protect the rotator cuff from later rupture [4]. On the conservative side of the ledger, the meta-analytic file puts exercise ahead of non-exercise control on pain (SMD −0.94) with specific programs beating generic ones (SMD −0.65) [5], and the newest network meta-analysis — 89 trials, 5,532 participants — concludes that most interventions are not superior to isolated shoulder exercises, "a reasonable first-line approach" [6]. The injection this reader may already have had matched manual therapy at one year while generating more revisits (60% vs 37%) and repeat injections (38% vs 20%) [8]. The imaging literacy that defuses the fear: full-thickness cuff tears live in 11–17% of pain-free shoulders [10]. And the reason not to drift: only 50% of new shoulder episodes fully resolve within six months — the shoulder is the joint where waiting quietly fails [13]. The danger at this door is not the press. It is the six months of ibuprofen and skipped sessions that the evidence says not to wait out.

The operation audit: the trial where training met the surgeon

The most expensive fear at this door is the one that ends in a surgeon's office, so the surgery file goes first — and it is one of the most remarkable in musculoskeletal medicine. A Danish trial took 90 consecutive patients aged 18–55, symptoms lasting six months to three years, all with a positive impingement sign and randomized them to arthroscopic subacromial decompression or to physiotherapy built on graded strengthening of the shoulder's stabilizers and decompressors [2]. At twelve months, the Constant score — the composite of pain, function, strength and motion — had improved to 57.0 in the training group versus 52.7 in the surgery group, and the difference was not significant; the pain-and-dysfunction scores found no group differences either [2]. The authors' own conclusion, printed under the trial: "surgical treatment of rotator cuff syndrome with subacromial impingement was not superior to physiotherapy with training" [2]. One of the two arms carried surgical risk; the other carried dumbbells.

The program side then got stronger. A Swedish trial recruited 102 patients whose subacromial impingement had already outlasted conservative treatment — over six months of persistent symptoms, referred by orthopaedic specialists — and randomized them to a specific exercise strategy or an unspecific movement program [1]. The specific strategy is the one this door's reader should recognize by name: eccentric strengthening for the rotator cuff, concentric/eccentric work for the scapular stabilizers, plus manual mobilisation — five to six supervised sessions across twelve weeks, home work once or twice daily [1]. The results priced the dose precisely: Constant-Murley improved 24 points versus 9 (mean difference 15, 95% CI 8.5–20.6); 69% (35/51) reported a successful outcome versus 24% (11/46), OR 7.6 (3.1–18.9); and the column that made the trial famous — the proportion who went on to choose surgery fell from 63% (29/46) to 20% (10/51), OR 7.7 (95% CI 3.1–19.4) [1]. Same joint, same diagnosis, same timeframe; the difference between a surgical candidate and a training case was, to a striking degree, the specificity of the program.

Then the operation faced the control that settles questions. The UK CSAW trial — 32 hospitals, 51 surgeons, 313 patients who had already failed a non-operative programme including exercise and at least one steroid injection — randomized participants three ways: arthroscopic subacromial decompression, arthroscopy-only placebo (the incisions and the anaesthesia, the bone and soft tissue removal omitted), or no treatment [3]. At six months, Oxford Shoulder Score was 32.7 after decompression versus 34.2 after placebo arthroscopy — a mean difference of −1.3 points (95% CI −3.9 to 1.3, p=0.31): the part of the operation that removes the spur contributed nothing measurable [3]. Both surgical arms scored above no treatment (29.4; +2.8 and +4.2 points), but the trial's own reading is the one to keep: the differences were "not clinically important" and may reflect placebo effects or postoperative physiotherapy [3]. The Finnish long-term column closes the file: 140 stage-II impingement patients randomized to structured exercise with or without acromioplasty showed, on five-year MRI, no difference in perforating supraspinatus ruptures or rotator cuff muscle volumes — with fatty degeneration actually graded worse in the operative arm (not significantly) — leading the authors to write that "acromioplasty is not justified as a treatment for dynamic shoulder impingement syndrome" [4]. And note the complication column in CSAW: all six study-related complications were frozen shoulders [3] — the operation carries a door to the stiffest chapter in this wing (that door) as a risk line.

The conservative file: exercise priced against everything else

Strip the surgery out and the question becomes which conservative treatment, and the meta-analytic answer is stable. The systematic review of randomized trials in shoulder impingement pooled the columns: exercise beat non-exercise control interventions on pain with SMD −0.94 (95% CI −1.69 to −0.19); specific exercises beat generic ones with SMD −0.65 (−0.99 to −0.32) [5]. The same file priced the alternatives honestly: corticosteroid injection beat no treatment (SMD −0.65, −1.04 to −0.26) but was never shown to beat a program; NSAIDs beat placebo modestly (−0.29, −0.53 to −0.05); manual therapy beat placebo (−0.35, −0.69 to −0.01) and added to exercise it helped only at the shortest follow-up (−0.32, −0.62 to −0.01) [5]. The reviewers' summary sentence is the umbrella door's prescription in miniature: "exercise should be considered for patients with shoulder impingement symptoms" — with the add-ons optional, not foundational [5].

The newest and largest column agrees. The 2026 network meta-analysis of 89 randomized trials and 5,532 participants with rotator cuff tendinopathy compared every physical-therapy intervention against every other and found shoulder-muscle exercises reducing pain versus no treatment (SMD −0.79, 95% CI −1.33 to −0.26) and improving function (SMD 0.74, 0.34–1.14) — and, critically, most interventions were not superior to isolated shoulder exercises alone, which "might be considered a reasonable first-line approach" [6]. Two honesty clauses travel with those numbers and this file keeps them visible: the certainty is graded low to very low [6], and the older synthesis adds a calibration the reader should actually want — across eleven level-1/2 trials, exercise produced statistically and clinically significant improvements in pain and function, but not in range of motion or strength testing [7]. You are training to change the shoulder's tolerance and symptoms, not to set a personal record in week two.

One more column from that older synthesis belongs in a gym-reader's hands: supervised exercise was not different from home exercise programmes [7]. The winning dose is a structured, progressive, specific program — the supervision format is negotiable. A training log, a plan, and honest pain tracking deliver the trial-level dose at the gym you already attend; that is the entire premise this app is built on.

The injection column: what the needle buys and what it costs

Many readers arrive at this door already holding a cortisone appointment, so the injection file gets its own audit. The head-to-head trial randomized 104 patients with unilateral impingement to a 40-mg subacromial corticosteroid injection or six sessions of manual physical therapy and watched them for a year [8]. Both arms improved by about half their baseline Shoulder Pain and Disability Index score, and the between-group difference at one year was 1.5% (95% CI −6.3% to 9.4) — statistically nothing [8]. The cost column is why the file still matters: over the same year, the injection arm made more impingement-related primary-care visits (60% vs 37%) and needed more repeat injections (38% vs 20%) [8]. An injection is not forbidden at this door — it is palliative borrowing, reasonable in the short term, and it pairs far better with a program than with the hope that the needle ends the story. The meta-analytic file prices it exactly that way: better than nothing (SMD −0.65), never demonstrated superior to exercise, with ultrasound guidance adding a modest edge over blind injection (−0.51) [5].

The self-exam and the scan that isn't the pain

The umbrella reader holds no diagnosis, and this file's position is that a few exam findings carry real information while imaging mostly carries noise. The Rational Clinical Examination review of rotator cuff disease — 28 studies of the shoulder exam — found that among pain-provocation tests, the painful arc was the only one with a positive likelihood ratio above 2.0 (3.7, 95% CI 1.9–7.0), and a normal painful arc was the best rule-out (negative LR 0.36) [9]. For the question that matters more — is there a full-thickness tear — the lag signs dominate: a positive external rotation lag test multiplies the odds 7.2-fold (1.7–31), a positive internal rotation lag 5.6-fold (2.6–12), a positive drop-arm test 3.3-fold (1.0–11) [9]. In gym language: pain through a specific arc of elevation is meaningful; an arm that will not hold itself rotated against gravity is a specialist's finding — and if your arm drops, the rotator cuff tear door owns your next click.

Then the imaging literacy that removes the fear this door runs on. The 2025 systematic review of 53 studies of asymptomatic adult shoulders measured what scans find in people whose shoulders do not hurt: full-thickness rotator cuff tears in 11–17% of pain-free shoulders on ultrasound across two population-based samples (1,631 shoulders), up to 35% in mixed clinical samples; tendinopathy or partial-thickness tears in 34% of a population sample and 7–70% of athletes [10]. The review's ceiling-to-floor range — 0% to 100%, low to very low certainty — is itself the lesson [10]. An MRI report describing "tendinopathy" or a "partial tear" is describing the background rate of the species, not a verdict on your training future. The scan is not the pain; the dosing protocol below does not change because a radiologist wrote the word tear on a pain-free-or-painful tendon. What changes your protocol are the red flags below and the trend of your own symptoms under load.

The scapular chapter: why the program looks the way it does

Readers who try the Holmgren-style program notice it spends half its time below the shoulder — rows, low traps, serratus work — and the consensus statement on scapular dyskinesis is why. The international summit's conclusions, in order: scapular dyskinesis is present in a high percentage of most shoulder injuries; its exact causal role is not clearly defined; shoulder impingement symptoms are particularly affected by it; it is best viewed as a potential impairment to shoulder function, not a diagnosis; and — the load-bearing sentence — rehabilitation programmes to restore scapular position and motion can be effective within a more comprehensive shoulder rehabilitation programme [11]. That is the honesty this library prefers over anatomy horror stories: your scapula is not "out of place"; its movement pattern is a modifiable input to a painful shoulder, and training it is the evidence-backed half of the program. The muscle imbalances file and the pressing chapters of the bench-press causation twin carry the practical programming.

The persistence file: why "wait and see" is the risky choice

The prognosis numbers are the ones that should recalibrate the resting reader. The systematic review of prognostic cohort studies — 16 cohorts, best-evidence synthesis — opens with the line this door's procrastinators need: only 50% of all new episodes of shoulder disorders end in complete recovery within six months [13]. The same review graded the predictors: strong evidence that high pain intensity forecasts a poorer outcome in primary care, moderate evidence for long duration of complaints and high baseline disability — the "wait until it stops hurting" strategy selects directly into the poor-outcome group, because duration is the modifiable variable on that list [13]. The epidemiology sizes the crowd behind the door: point prevalence of shoulder complaints 6.9–26%, one-month 18.6–31%, lifetime up to 66.7%, annual incidence 0.9–2.5% [12]. A quarter of the adults in any gym at any moment hold this door's question. The file's demand is early, structured load calibration — which is a training intervention, delivered by training tools.

The scale of the door: the Brazilian ledger

The home market prices this door at national scale. The Brazilian cross-sectional study of manufacturing-sector workers — the Nordic questionnaire applied in a medium metallurgical company — found the upper limb the most affected body region, and the shoulder its flagship site: 24.8% of workers reported shoulder symptoms, ahead of wrist/hand (19.0%) and elbow/forearm (15.5%) [14]. The adjusted model adds the columns a training library can act on: sleep disorders, self-reported general health and work pauses were the independent risk factors [14]. Between the working population's hurting shoulder and the bench-press crowd, the undiagnosed shoulder is one of the largest single training questions in Brazilian gyms: "pode fazer academia com dor no ombro?" This file exists in Portuguese because that is where the readers are.

The fences: red flags, referrals, and honest limits

The fences at this door are real, and the first one is hard. The red-flag set stops the program and opens a medical door this week: shoulder pain after significant trauma, or with visible deformity — that is the dislocated or separated shoulder conversation; an arm that cannot be raised or rotated against gravity (the lag-sign column [9]) belongs with the rotator cuff tear door and a clinician; repeated slipping or apprehension of instability is its own file; a shoulder that is hot, red and feverish, or pain with fever after an injection, is same-day medicine; night pain that wakes you and relentless pain unresponsive to position deserve assessment — progressive stiffness locking the arm's rotation belongs to the frozen shoulder door, which has its own timeline and its own rules; and sudden weakness after an injection or a pop is a clinician's question, not a programming one. The second fence is the diagnosis fence: this door is not a diagnostic service, and the dosing below assumes nothing structural is being missed — which is exactly why the red flags come first. The third is the honesty fence: the CSAW surgical arms did edge no treatment [3]; Holmgren's specific-exercise arm still sent 20% to surgery [1]; the meta-analytic certainty is low [5][6]; and every trial still contained patients who ended in an operating room or with an unresolved shoulder [1][2]. None of that reverses the direction of the file. All of it prices the confidence honestly.

The protocol: six rules

  1. Triage before you train. Trauma with deformity, an arm that fails lag-sign testing, instability episodes, fever with a hot joint, progressive locking stiffness, night pain that wakes you — the program stops and the right clinician starts [3][9]. Everything below is for the shoulder that hurts without a red flag.
  2. Read pain as a dosing signal, not a stop sign — and not background noise either. Pain that settles back to baseline within 24 hours and does not climb week over week is acceptable dosing; pain that spikes beyond that, lingers into the next day, or wakes you at night is a down-shift and, if it persists two weeks, an assessment [1][2][13].
  3. Calibrate the load; keep the membership. The winning programs modified dose, not identity: swap the barbell press for machines or a neutral-grip variation where tolerance demands, narrow the bench press grip, rotate pressing angle, cut overhead depth — and keep training. The couch is the one intervention with no evidence at this door [1][2][5].
  4. Run the specific program, not a generic one. Eccentric rotator-cuff work plus scapular-strengthener work is the dose that cut surgery choices from 63% to 20% [1], beat generic movement SMD −0.65 [5], and survived as first-line in an 89-trial network [6]. Home format is fine — supervised was not different [7].
  5. Treat the injection as a bridge, not a destination. The needle matched manual therapy at one year while generating more revisits and repeat injections [8]; pair it with rule 4, not with hope.
  6. Persist, track, and re-check at the deadline. Recovery is the norm with a program and the minority without one — 50% at six months [13] — so log every session and symptom, and if six to twelve weeks of compliant dosing has not bent the trend, escalate: exam first (painful arc, lag signs [9]), imaging interpreted against the 11–17% pain-free tear rate [10], and the structural doors behind this umbrella.

How RightRep fits into this

The hard part of training through an undiagnosed shoulder is not courage; it is calibration — knowing whether this week's program is tolerable dosing or quiet provocation, whether your press is quietly changing shape to protect the pain, and whether the trend is actually bending. That is a measurement problem, and it is what the app does:

  • The rep score trend is the tolerability graph. Pain makes every set feel unreliable; an objective per-session quality score, trending week over week against your symptoms, shows whether the calibrated program is landing — the dose the trials delivered [1] comes from sessions you can actually see happening.
  • Form checks catch the protective compensation. A guarding shoulder changes press mechanics — elbow flare, range shortening, weight shifting to the painless side — and a camera that grades every set notices the limp before it becomes the way you lift.
  • Symmetry scoring turns "calibrate the load" into numbers. Rule 3 is a dosing instruction; a left-right balance score is how it is enforced rep by rep instead of guessed.
  • Tempo control and rest timers structure the graded return — controlled movement under tolerance, honest spacing between sets, and a log that makes week-over-week pain trending a table, not a memory (why tracking matters).

Run it at rightrep.app — with the load calibrated, the cuff and scapula strengthening, and the log proving the trend.

The bottom line

Can you train with shoulder pain? Yes — after the triage step, with the file priced honestly: training matched arthroscopic decompression on Constant score at one year in its randomized trial (90 patients, 57.0 vs 52.7, not significant) [2]; a specific eccentric-cuff-plus-scapular program added 24 Constant-Murley points where a generic program added 9, and cut surgery choices from 63% to 20% (OR 7.7) [1]; the operation did not beat placebo arthroscopy in the three-group CSAW trial (313 patients, −1.3 Oxford points, −3.9 to 1.3) [3] and did not protect the cuff from rupture at five years on MRI [4]; exercise leads the conservative meta-analysis (SMD −0.94 vs non-exercise control, specific over generic −0.65) [5] and stands as the reasonable first line across 89 trials and 5,532 participants [6]; supervised was no better than home programs [7]; the cortisone injection matched manual therapy at one year while doubling back for more care [8]; the painful arc and lag signs are the exam findings that matter (LRs 3.7 and 7.2) [9]; full-thickness tears live in 11–17% of pain-free shoulders, so the scan is not the pain [10]; scapular rehabilitation works inside a comprehensive program [11]; the door is crowded — point prevalence up to 26% [12] — and only 50% of new episodes resolve by six months, the number that retires "wait and see" [13]. The fences are the red flags: trauma with deformity, an arm that fails against gravity, instability, fever, locking stiffness, night pain — medicine first, program second [3][9]. This guide pairs with the named doors behind the umbrella — impingement, the rotator cuff tear, the frozen shoulder — the causation twin (does the bench press damage your shoulders), the sibling umbrellas (lower back pain, knee pain), and tracking your workouts (the mechanism that delivers the adherence dose).

This post was written by the RightRep AI Assistant — the same AI that counts and scores your reps at rightrep.app.

Sources

  1. Holmgren T, Björnsson Hallgren H, Öberg B, Adolfsson L, Johansson K. Effect of specific exercise strategy on need for surgery in patients with subacromial impingement syndrome: randomised controlled study. BMJ, 2012;344:e787. pubmed.ncbi.nlm.nih.gov
  2. Haahr JP, Østergaard S, Dalsgaard J, Norup K, Frost P, Lausen S, Holm EA, Andersen JH. Exercises versus arthroscopic decompression in patients with subacromial impingement: a randomised, controlled study in 90 cases with a one year follow up. Annals of the Rheumatic Diseases, 2005;64(5):760–764. pubmed.ncbi.nlm.nih.gov
  3. Beard DJ, Rees JL, Cook JA, Rombach I, Cooper C, Merritt N, Shirkey BA, Donovan JL, Gwilym S, Savulescu J, Moser J, Gray A, Jepson M, Tracey I, Judge A, Wartolowska K, Carr AJ; CSAW Study Group. Arthroscopic subacromial decompression for subacromial shoulder pain (CSAW): a multicentre, pragmatic, parallel group, placebo-controlled, three-group, randomised surgical trial. The Lancet, 2018;391(10118):329–338. pubmed.ncbi.nlm.nih.gov
  4. Ketola S, Lehtinen J, Elo P, Kortelainen S, Huhtala H, Arnala I. No difference in long-term development of rotator cuff rupture and muscle volumes in impingement patients with or without decompression. Acta Orthopaedica, 2016;87(4):351–355. pubmed.ncbi.nlm.nih.gov
  5. Steuri R, Sattelmayer M, Elsig S, Kolly C, Tal A, Taeymans J, Hilfiker R. Effectiveness of conservative interventions including exercise, manual therapy and medical management in adults with shoulder impingement: a systematic review and meta-analysis of randomised trials. British Journal of Sports Medicine, 2017;51(18):1340–1347. pubmed.ncbi.nlm.nih.gov
  6. Lazzarini SG, Bettariga F, Mosconi B, Maestroni L, Cordani C, Bandini E, Dal Farra F, Lopez P, Arienti C. Comparative effectiveness of physical therapy interventions in adults with rotator cuff tendinopathy: a systematic review and network meta-analysis. British Journal of Sports Medicine, 2026;60(13):970–982. pubmed.ncbi.nlm.nih.gov
  7. Kuhn JE. Exercise in the treatment of rotator cuff impingement: a systematic review and a synthesized evidence-based rehabilitation protocol. Journal of Shoulder and Elbow Surgery, 2009;18(1):138–160. pubmed.ncbi.nlm.nih.gov
  8. Rhon DI, Boyles RB, Cleland JA. One-year outcome of subacromial corticosteroid injection compared with manual physical therapy for the management of the unilateral shoulder impingement syndrome: a pragmatic randomized trial. Annals of Internal Medicine, 2014;161(3):161–169. pubmed.ncbi.nlm.nih.gov
  9. Hermans J, Luime JJ, Meuffels DE, Reijman M, Simel DL, Bierma-Zeinstra SM. Does this patient with shoulder pain have rotator cuff disease?: The Rational Clinical Examination systematic review. JAMA, 2013;310(8):837–847. pubmed.ncbi.nlm.nih.gov
  10. Sanders S, Ibounig T, Haas R, Jones M, Rämö L, Docking S, Järvinen T, Taimela S, Hoffmann T, Buchbinder R. Rotator Cuff Imaging Abnormalities in Asymptomatic Shoulders: A Systematic Review. Journal of Orthopaedic & Sports Physical Therapy, 2025;55(12):1–16. pubmed.ncbi.nlm.nih.gov
  11. Kibler WB, Ludewig PM, McClure PW, Michener LA, Bak K, Sciascia AD. Clinical implications of scapular dyskinesis in shoulder injury: the 2013 consensus statement from the 'Scapular Summit'. British Journal of Sports Medicine, 2013;47(14):877–885. pubmed.ncbi.nlm.nih.gov
  12. Luime JJ, Koes BW, Hendriksen IJ, Burdorf A, Verhagen AP, Miedema HS, Verhaar JA. Prevalence and incidence of shoulder pain in the general population; a systematic review. Scandinavian Journal of Rheumatology, 2004;33(2):73–81. pubmed.ncbi.nlm.nih.gov
  13. Kuijpers T, van der Windt DAWM, van der Heijden GJMG, Bouter LM. Systematic review of prognostic cohort studies on shoulder disorders. Pain, 2004;109(3):420–431. pubmed.ncbi.nlm.nih.gov
  14. Hembecker PK, Reis DC, Konrath AC, Gontijo LAD, Merino EAD. Investigation of musculoskeletal symptoms in a manufacturing company in Brazil: a cross-sectional study. Brazilian Journal of Physical Therapy, 2017;21(3):175–183. pubmed.ncbi.nlm.nih.gov