Shoulder impingement is a name with an explanation packed inside it, and the best example in this subject of what happens when the explanation is finally tested. The word describes a mechanism: bone at the top of the shoulder catching a tendon underneath as the arm goes up, and therefore correctable by shaving the bone away. For decades that was the account, and the operation built on it became one of the commonest in orthopaedics.
The last ten years went badly for it. This page sets out what was found, then draws a line most pages on this subject will not draw.
The short answer
- The term names a mechanism — bone pinching a tendon — that has not held up under testing.
- A randomised trial compared the operation against placebo surgery. Both surgical groups improved and did not meaningfully differ from each other.
- Adding the operation to an exercise programme added nothing to the exercise programme.
- The field renamed the area rotator cuff related shoulder pain, because the mechanical account had failed.
- Exercise produces small benefits on low-quality evidence, with no winning programme.
- We write about range. We do not write about how to make a painful shoulder stop hurting.
When to see a clinician first
Some shoulder problems are not a matter for a web page, and the list belongs at the top of this one rather than the bottom.
See a clinician who can examine you if the pain followed a fall or a dislocation, if you cannot lift the arm at all, if there is sudden weakness, if night pain wakes you consistently, if there is numbness or pins and needles down the arm, or if shoulder pain arrives with fever or a general feeling of being unwell. Any one of those is a reason to be examined by a person, in a room, with hands.
Nothing here diagnoses anything and nothing here is treatment. That sentence is not a legal formula on this page. It is the finding.
What shoulder impingement meant
The idea was mechanical and satisfying, which is part of why it lasted.
The tendons of the rotator cuff pass through a narrow space under the acromion, the bony shelf at the top of the shoulder blade. Raise the arm and that space narrows. If the shelf is hooked, or a spur has grown on its underside, the tendon underneath gets caught — so the reasoning went — and every overhead reach wears it a little more. The painful arc, where a shoulder hurts through the middle of the movement and settles at the top, fitted the story exactly.
The treatment followed: shave the bone through a keyhole, widen the space, and the catching stops. Subacromial decompression became a high-volume operation on that logic, with a layer of exercise advice underneath it — correct the blade, open the chest, restore the space.
The mechanism was plausible. It was also never the thing that had been tested.
The trial with placebo surgery
Beard and colleagues published the CSAW trial in The Lancet in 2018, and it deserves a full paragraph rather than a clause. Patients with subacromial shoulder pain were randomised into three groups: arthroscopic subacromial decompression; placebo arthroscopy, which is the same operation performed to the same point, camera inside the joint, the surgeon looking around, without the decompression being done; and no treatment at all. Both surgical groups improved from where they started. They did not meaningfully differ from each other. The difference between either surgical group and no treatment was small and of doubtful clinical importance.
That is easy to mishear as "surgery does not work", which is not what it says. People who had the operation got better. So did people who had the theatre, the anaesthetic, the incisions and none of the decompression. What failed was the mechanism the operation was named for.
A placebo-controlled surgical trial is rare and hard to argue with. You cannot blind a patient to a scar without making the scar, and ethics committees rarely allow it, so the design exists for a handful of operations in all of orthopaedics. When one is run and the sham arm matches the real one, the burden shifts completely.
Grade: strong, and it changed practice.
Adding the operation to exercise
The other half of the case had already been made, more quietly, in Finland.
Ketola and colleagues randomised people with this diagnosis to a supervised exercise programme, or to the same programme plus the operation. The operation added nothing. That trial was first reported in 2009 and the groups were followed up at five years and again at ten, and the answer did not change with time.
Two designs, opposite directions, the same answer.
Grade: reasonably solid, on a single trial group with unusually long follow-up.
Why the name was changed
Lewis, writing in Manual Therapy in 2016, set out the consequence and proposed the honest replacement: rotator cuff related shoulder pain. Not because the new name explains more, but because it explains less, which is the correct amount.
Two things sit under that renaming. The structures blamed on a scan turn up constantly in shoulders that do not hurt — tendon changes, spurs, partial tears — so finding one in a painful shoulder does not establish that it is the reason for the pain. And a great deal about who gets this problem, and why, remains genuinely unknown. Lewis's paper is unusual for saying so plainly rather than proposing a replacement mechanism with the old one's confidence.
Grade: reasonably solid for the imaging findings in people without symptoms. What replaces the old mechanism is honest uncertainty, not a better story.
The blade and the hands-on work
Two more findings go against the standard advice.
Mintken and colleagues, in JOSPT in 2016, randomised people with shoulder pain to exercise, or to exercise plus cervicothoracic manual therapy. The manual therapy added no benefit.
On the shoulder blade, Hickey and colleagues reviewed the prospective studies in the British Journal of Sports Medicine in 2018 and found that athletes with scapular dyskinesis and no symptoms had a modestly increased risk of later shoulder pain — a real result, small in size and wide in its uncertainty. Ratcliffe and colleagues, in the same journal in 2014, found no consistent relationship between scapular orientation and shoulder pain at all. Both can be true: a movement pattern may carry a little prospective risk while the resting position of the blade tells you nothing.
Grade: weak. A long way from there to "your shoulder hurts because your blade sits wrong", which is what the finding becomes by the time it reaches a web page. Scapular control takes that apart properly.
What exercise does, and how well
Exercise-based care is what most guidelines now recommend, and it is worth knowing how strong that recommendation is underneath.
Page and colleagues reviewed exercise for rotator cuff disease for Cochrane in 2016. The evidence was low quality, the benefits were small, and no programme was clearly better than any other. Not a hedge and not a preamble to a stronger claim — the finding is that the finding is weak. Nobody has shown which components matter, and nobody has shown that range work is why anyone improves.
Grade: low quality, small benefits, no winner.
The line this site draws
Our position, stated plainly, and narrower than the market's.
We write about range, because range is measurable and responds to work. We do not write about how to make a painful shoulder stop hurting, because the evidence does not support anyone claiming that from a web page.
Mobility work addresses two things: range, meaning how far a joint travels before it stops, which you can measure today and measure again in three weeks; and how a joint feels to move through that range, which is subjective and still real. Both respond to time spent in the position.
It does not address the symptom, the diagnosis or the outcome. No movement on this site is offered as something that addresses a condition, and we do not claim to treat anything. If a page tells you which stretch is for your impingement, it is asserting more than the trials above will carry.
The shoulder mobility test is what we do instead: the Apley scratch reach, both arms, five bands, scored 10, 30, 55, 75 and 95. It is a screen rather than a diagnosis, and it reads three joints at once without saying which was short. The smallest gap it can express between your arms is 20 points, well above the 12 at which Limber calls an imbalance — so one band of difference means measure again, not finding. The Range Score composes what you have read into one number and leaves the rest blank.
A range reading says nothing about whether a shoulder hurts. That is the limit, and we would rather print it.
What the app does with the flag
Limber's move catalogue carries contraindication flags, and shoulder impingement is one of the six. The rule is an exclusion rule: any move whose flags match one of yours is removed from the pool before selection, rather than ranked lower inside it. A test generates 500 routines with each flag active and asserts that no flagged move appears in any of them.
Exactly one move in the 56-move catalogue carries that flag — M033 Sleeper Stretch, a 30-second hold per side at the app's hardest tier, which also carries an instruction to ease off at any pinching in the front or top of the shoulder. Setting the flag removes one position, not a category.
There is no route in today's build to set it. The onboarding step that asks about sore areas offers the lower back, the neck and the knees, and writes those three flags. The shoulder flag exists in the taxonomy and in the filter; nothing in the app hands it to you yet.
An exclusion rule is not a protocol. Limber has no pain protocols, does not offer one, and removing a movement from a plan is a refusal rather than a treatment. The rest of the region is in shoulders and upper back, the joint itself in rotator cuff mobility, and the range work in shoulder mobility exercises.
Questions
Is shoulder impingement a real diagnosis?
The pain is real; the name is the problem. It describes a mechanism — bone catching a tendon — that has not survived testing, and the field has largely replaced it with the deliberately vaguer rotator cuff related shoulder pain. A label naming a cause it cannot demonstrate is worse than one admitting it does not know.
Do I need surgery for it?
That is a matter for a clinician who can examine you. What can be said is that the operation named for the mechanism was compared against placebo surgery in The Lancet in 2018 and did not meaningfully beat it, and that a separate randomised trial found it added nothing to an exercise programme.
Which stretches help shoulder impingement?
We do not publish an answer, because nobody has established one. The Cochrane review of exercise for rotator cuff disease found small benefits on low-quality evidence with no programme better than another, so the confident lists elsewhere assert more than the evidence carries. Range work is worth doing on its own terms rather than as something aimed at a symptom.
Does bad posture cause it?
There is no good evidence that it does. Resting shoulder blade position shows no consistent relationship with shoulder pain, and adding hands-on work for the neck and upper back to exercise produced no extra benefit in a randomised trial. A movement pattern at the blade carries a small prospective risk in athletes, which is far weaker than the claim usually built on it.
Should I stop moving the arm if it hurts?
Not on our say-so either way — that decision belongs to someone who has examined the shoulder. If you are moving, the rule from the app applies: work to a mild stretch, never to sharp pain, and come out of a position that pinches at the front or the top rather than pushing through it.
What does Limber do for a painful shoulder?
It measures range and plans range work, and that is the whole of it. There is no pain protocol in the app, no programme aimed at a condition, and no claim that a plan changes a symptom. The only pain-related mechanism in the engine is an exclusion rule that removes flagged movements before a session is built.