Rotator Cuff Tears: Repair, Rehab or Reverse? An Evidence-Based Decision Guide

A patient in their sixties has a painful shoulder and a full-thickness supraspinatus tear on ultrasound. Repair it, refer for physiotherapy or, if the cuff is beyond saving, plan a reverse? The trials of rotator cuff tear surgery vs physiotherapy answer more of this than their reputation suggests. For small atraumatic tears in patients over 55, structured physiotherapy matches repair for several years, but unrepaired tears often enlarge and the one trial followed to 15 years favours primary repair. Massive tears, pseudoparalysis and arthritis are a separate decision. This post sets out who goes where, and on what evidence. The clinical background is on the rotator cuff tears topic page.

Four questions that sort every tear

The decision is easier when the tear is placed before the trials are quoted. Four questions do most of the work:

1. How did it start? An acute traumatic tear in an active patient is a different problem from an atraumatic, degenerative tear found in a 65-year-old.
2. How big is it, and is the muscle still muscle? Size by Cofield (small under 1 cm, medium 1–3 cm, large 3–5 cm, massive over 5 cm or two complete tendons), retraction by Patte, and fatty infiltration by Goutallier.
3. Can the patient lift the arm? Pseudoparalysis is active elevation under 90° with full passive range; it marks the shoulder whose cuff has stopped centring the head.
4. Is the joint arthritic? Superior migration, acetabularisation of the acromion and glenohumeral narrowing (Hamada grades 3–5) make this cuff tear arthropathy.

The randomised trials answer only the first branch: repairable, mostly small, mostly atraumatic supraspinatus tears. Everything beyond that rests on cohort series and consensus, and should be presented to patients as such.

Does surgery beat physiotherapy for a degenerative cuff tear?

Four randomised trials compare repair with physiotherapy head to head.

A 2019 Cochrane review included nine trials (1,007 participants). In the main comparison, repair with decompression against exercise in three trials, surgery probably provides little or no improvement in pain at one year (0.87 points better on a 0–10 scale; moderate certainty), and function was 6 points better on a 0–100 scale (low certainty). The authors were uncertain whether repair gives clinically meaningful benefit, and noted the trials mostly enrolled small degenerative supraspinatus tears. Their conclusions may not apply to traumatic tears, large tears involving the subscapularis or young people.

The Finnish design carries a second lesson. Its middle arm added acromioplasty to physiotherapy and changed nothing, which is consistent with the placebo-controlled trials discussed in our post on the CSAW trial and subacromial decompression.

What does 15 years of follow-up add?

The Norwegian trial is the reason the debate is not closed. Moosmayer randomised 103 patients with tears no larger than 3 cm to primary repair or physiotherapy, with secondary repair allowed if physiotherapy failed, and kept a blinded assessor for 15 years.

• 1 year: 9 patients (18%) had already crossed over to surgery.
• 5 years: 12 of 51 had crossed over. Repair led by 5.3 Constant points (p = 0.05) and 9.0 ASES points, differences the authors judged small and possibly below clinical importance.
• 10 years: the gap had widened, to 9.6 Constant and 15.7 ASES points. The 14 patients who crossed over scored 10.0 Constant points below the primary repair group.
• 15 years: 83 of 103 (81%) attended and 15 of 51 had crossed over. Repair led by 11.8 Constant points, 13.9 ASES points and 1.8 cm on a 10 cm pain scale. The SF-36 showed no significant difference. The 12 crossover patients who attended now matched the primary repair group, a finding the authors interpret with caution.

Two cautions keep this in proportion. It is a single-centre trial of 103 patients, and the Finnish trial, in patients over 55 with small single-tendon tears, still showed no difference at six years. The crossover comparison is also outside the randomisation: the patients who failed physiotherapy were a selected group. What the trial does show is that, in this population, "physiotherapy first, repair later if needed" gave a worse result than primary repair for at least 10 years, and only at 15 years, in the few crossover patients still attending, did the gap close.

Will an unrepaired tear get bigger?

This is the strongest argument for repair in a younger patient, and the numbers are consistent across studies:

• Moosmayer: 37% of tears treated with physiotherapy alone enlarged by more than 5 mm on ultrasound over five years, and enlargement was associated with an inferior outcome. In the 26 tears never repaired, mean anteroposterior size rose from 16.2 to 31.6 mm by 15 years.
• Ranebo: 29.2% of unrepaired small traumatic tears enlarged by more than 5 mm within 12 months.
• Kukkonen: at two years the mean sagittal tear size was 4.2 mm after repair against 11.0 mm without it. At six years, though, repair had not protected against progression of glenohumeral arthritis or cuff tear arthropathy.
• Mall (asymptomatic tears): in 195 people followed with annual ultrasound, the 44 whose shoulders became painful had larger tears at enrolment. With the onset of pain, 18% of full-thickness tears had grown by more than 5 mm, and 40% of partial tears had become full-thickness.

Enlargement matters because what follows it is hard to reverse. In Goutallier's original CT series, infraspinatus fatty degeneration never regressed after a successful supraspinatus repair. A tear that is small and repairable at 55 may not be either at 65.

Who should start with physiotherapy?

Most patients with an atraumatic full-thickness tear. The MOON Shoulder Group enrolled 452 such patients in a standardised physiotherapy programme. Fewer than 25% chose surgery by two years, and those who did generally decided between 6 and 12 weeks. At 10 years, 115 (27.0%) had had an operation, 56.5% of them within six months of enrolment, and patient-reported scores had not declined in those who never had surgery.

The predictors are worth quoting to patients. Low expectations of physiotherapy predicted early surgery; workers' compensation status and activity level predicted later surgery. None of the three is a feature of the tear itself.

The programme on the topic page runs 6–12 weeks: periscapular strengthening for scapular control, deltoid strengthening to compensate for cuff weakness, stretching to keep passive range, and work on posterior capsular tightness. A corticosteroid injection can make the exercises tolerable but does not heal the tear. For an irreparable massive tear the equivalent is anterior deltoid re-education, and the counselling should be honest: Levy's 17 patients (mean age 80) gained elevation from 40° to 160°, but when Yian repeated the programme in 30 patients only 40% succeeded.

When is early repair the better call?

The site's guidance offers early repair to:

• Active patients with an acute traumatic tear, ideally within about three months, before the tendon retracts and the muscle atrophies
• Younger patients (under about 60–65) with a symptomatic full-thickness tear
• Patients whose weakness progresses despite physiotherapy, or whose function at work is significantly impaired

Two caveats belong with the first indication. The only randomised trial of acute traumatic tears, Ranebo's, studied small tears (median sagittal size 9.7 mm, mean age 59.7) and found no difference at 12 months, with a re-tear rate of 6.5% after repair. The case for urgent repair of a larger traumatic tear therefore rests on cohort evidence and consensus, not on a trial. The second caveat is that "acute on chronic" is common: a fall that unmasks a degenerative tear in a 70-year-old is not the same lesion as an avulsion in a 45-year-old labourer. Established fatty infiltration, which develops over the months after a complete tear, suggests the tear predates the fall.

Will the repair heal, and does it matter?

Less often than older case series suggest. In UKUFF, a UK multicentre trial of open against arthroscopic repair in patients aged 50 or over, 217 repairs were imaged at 12 months and 56% had healed: 66% of small tears, 68% of medium, 47% of large and 27% of massive. Mean age was 61 in healed repairs and 64 in those that failed, and age was the dominant risk factor after adjusting for size. Approach made no difference to the Oxford Shoulder Score.

Healing matters, but a re-tear is not a failed operation. In UKUFF the patients with healed repairs improved most, those with re-tears slightly less, and those whose tears could not be repaired least. In Lambers Heerspink's trial, patients whose repair stayed intact scored 88.5 on the Constant score against 73.7 after conservative care. At the far end, Galatz found recurrent tears in 17 of 18 large and massive arthroscopic repairs, yet ASES scores rose from 48.3 to 84.6 at a year, before falling to 79.9 at two years. Whether a second row of anchors changes healing is covered in our post on single- versus double-row rotator cuff repair.

When is a rotator cuff tear irreparable?

Repairability is judged on the MRI before it is tested in theatre. The massive rotator cuff tear page lists the signs:

• Goutallier grade 3–4 fatty infiltration (grade 3 borderline, grade 4 irreparable)
• Patte stage 3 retraction, to or beyond the glenoid
• A positive tangent sign: supraspinatus atrophy
• An acromiohumeral interval under 7 mm: static superior migration

An irreparable tear in a shoulder without arthritis still has options short of arthroplasty:

• Partial repair with margin convergence: closing what can be closed restores part of the transverse force couple.
• Latissimus dorsi transfer for posterosuperior tears: in Gerber's 69 shoulders the Subjective Shoulder Value rose from 28% to 66% of normal, but shoulders with poor subscapularis function did not improve. Lower trapezius transfer is an emerging alternative.
• Superior capsular reconstruction: in Mihata's original 24 shoulders with fascia lata autograft, 83% had no graft tear or re-tear, but this is a single-surgeon series without a control group.
• Debridement and biceps tenotomy: palliation for older, low-demand patients whose problem is pain. A biodegradable balloon spacer gives temporary improvement.

When is a reverse shoulder replacement the answer?

When the shoulder, not just the tendon, has failed. The cuff tear arthropathy page gives the indications: cuff tear arthropathy with pseudoparalysis, a massive irreparable tear, and failed previous arthroplasty. The massive tear page puts the usual patient at older than about 65–70. Non-operative care is tried first unless the head escapes anterosuperiorly from under the coracoacromial arch, which is functionally debilitating.

The results explain its place. In Sirveaux's multicentre series of 80 Grammont prostheses, the Constant score rose from 22.6 to 65.6 and active elevation from 73° to 138°, and teres minor integrity was essential for recovering external rotation. Wall's 240 reverses (mean age 72.7) improved from 23 to 60 points, with massive cuff tears and cuff tear arthropathy doing better than post-traumatic and revision cases. Sirveaux advised reserving the prosthesis for elderly patients, and Wall, given the age of his patients and the short follow-up, that it be used judiciously. How the implant substitutes the deltoid for the cuff is set out in our post on Grammont reverse shoulder biomechanics.

Under about 60, a reverse for the cuff-deficient shoulder remains contentious because of implant longevity and demand, and joint-preserving options are weighed first. The detail is on the reverse total shoulder arthroplasty page.

Repair, rehab or reverse: the decision at a glance

Examination drives every row: lag signs map which tendons are gone, and passive range separates pseudoparalysis from stiffness. The rotator cuff examination sequence is the place to rehearse it.

In the exam

Examiners want the tear classified, the evidence quoted with its limits, and a decision that follows from the patient.