Shoulder Impingement: The Diagnosis That Is Being Retired, and What the Surgery Trials Found

In two separate randomised trials, surgeons operated on patients with shoulder impingement. Some received the real decompression. Others were taken to theatre, anaesthetised, given the same incisions and a look inside the joint, but the bone and soft tissue were deliberately left alone. The patients did not know which they had received.

At six months, two years, five years and ten years, the two groups did the same.

That is the central fact about shoulder impingement, and it is why the diagnosis itself is being quietly retired in much of the world.

The original idea

In the 1970s, Charles Neer proposed that shoulder pain and rotator cuff tears happen because the tendon gets compressed against the underside of the front of the acromion, the bony roof of the shoulder, as you lift your arm. He introduced an operation to shave that bone away and create more room.

The theory is intuitive, it produces a clear picture, and it is still how the condition is explained on many patient sites. The AAOS describes pain arising when the space between the acromion and the rotator cuff narrows, causing bone to rub on soft tissue.

Acromion shape was classified into three types: flat, curved and hooked, with the hooked type blamed for cuff damage.

Why the theory is in trouble

Four separate problems.

The tears are on the wrong side. If bone above the tendon were grinding it down, damage should start on the outer surface. In practice, between 76 and 91 percent of tears occur on the inner, joint-facing side of the tendon, away from any acromial contact.

The acromion classification is not reliable. When three experienced clinicians read 102 standardised X-rays, agreement between them on acromion shape was poor, with kappa values of 0.32 to 0.41. The researchers concluded the classification cannot be recommended and urged humility when interpreting acromial shape.

Removing the bone changes nothing. Covered below in detail. If the mechanical lesion were causing the pain, taking it out should help more than pretending to.

Tears happen after decompression anyway. Rotator cuff tears occur following acromioplasty, which contradicts the idea that the surgery prevents them.

The practical result is a terminology shift. Many clinicians now say subacromial pain syndrome or rotator cuff related shoulder pain, both of which describe where it hurts without claiming to know why. One NHS trust states plainly that rotator cuff related shoulder pain replaces the terms bursitis, tendinitis and impingement.

Worth knowing: US and UK patient sources genuinely differ here. If your American surgeon explains mechanical impingement and your British physiotherapist says the term is outdated, they are not contradicting each other out of ignorance. They are on different sides of an unfinished argument.

The trials

These are unusual. Placebo-controlled surgical trials are rare, difficult and ethically complex. Two were done on this operation, and both are worth knowing in detail.

CSAW, published in The Lancet in 2018

Design. 313 patients across 32 UK hospitals with 51 surgeons, randomised into three groups.

Who was included. Subacromial pain lasting more than three months, with intact rotator cuff tendons, who had already completed a non-operative program including exercise therapy and at least one steroid injection. Full thickness tears were excluded.

The three arms:

Arthroscopic subacromial decompression, the real operation, 106 patients
Investigational arthroscopy only, in which the essential surgical element of bone and soft tissue removal was omitted, 103 patients
No treatment, one reassessment appointment at three months, 104 patients

Patients in both surgical arms were blinded to which operation they had received.

Results at six months, on the Oxford Shoulder Score where 48 is best:

Decompression: 32.7
Placebo arthroscopy: 34.2
No treatment: 29.4

Decompression versus placebo surgery: a difference of minus 1.3 points, not statistically significant. The real operation scored numerically lower than the sham.

Both surgical groups did beat no treatment at all, by 2.8 and 4.2 points. But the trial had been designed around a target difference of 4.5 points as the threshold for clinical importance, and the authors stated these differences were not clinically important. They suggested the gap might reflect a placebo effect or the post-operative physiotherapy rather than the decompression.

Harms: six study-related complications, all frozen shoulders, two in each group.

FIMPACT, published in the BMJ, with five and ten year follow-ups

Design. A double-blind, sham-controlled trial across three Finnish public hospitals. Adults aged 35 to 65 with impingement symptoms for more than three months. Three arms: decompression, diagnostic arthroscopy as placebo surgery, and exercise therapy.

The primary outcomes were shoulder pain at rest and on arm activity, on a 0 to 100 scale, with a minimum important difference set in advance at 15 points.

At two years: decompression minus placebo was 4.6 points at rest and 9.0 points on activity. Neither reached statistical significance and neither approached the 15 point threshold.

Decompression did beat exercise therapy at two years, by 7.5 and 12.0 points, both statistically significant. Both were still below the 15 point threshold, and the authors cautioned that this particular comparison was unblinded and likely biased in favour of surgery.

At five years: decompression minus placebo was 2.0 points at rest and 8.0 on activity, neither significant. The advantage over exercise therapy had disappeared entirely.

At ten years: decompression minus placebo was 1.5 points at rest and 3.2 on activity. No significant differences on any outcome. The conclusion: decompression offered no benefit over placebo surgery or exercise therapy during ten year follow-up.

The pooled picture

The Cochrane review covering eight trials and 1,062 participants found that against placebo surgery, pain scored 2.6 versus 2.9 on a 0 to 10 scale, a difference of 0.26 points. Function was 72 versus 69 on a 0 to 100 scale. Quality of life was marginally worse after surgery.

Their conclusion, at high certainty: subacromial decompression does not provide clinically important benefits over placebo in pain, function or quality of life.

What the guidelines now say

The BMJ Rapid Recommendation makes a strong recommendation against subacromial decompression surgery for adults with shoulder pain lasting more than three months. The stated reasoning: almost all informed patients would choose to avoid surgery because there is no benefit but there are harms.

Their estimate of serious adverse events is around 6 to 7 per 1,000 people operated on.

In England, NHS commissioning policies have followed. One typical policy states the operation is not normally funded and should not be referred without prior approval from a funding panel, and requires a minimum of six months of active physiotherapy and no more than two steroid injections before any exception is considered.

Listed exceptions where decompression may still be appropriate: acute cuff tears, decompression done alongside a cuff repair, calcific tendinitis, large bone spurs, and post-fracture or post-traumatic bursitis. It is the standalone operation for ordinary subacromial pain that the evidence has removed.

What it feels like

Pain and stiffness reaching overhead, and often worse on the way back down. Pain lifting the arm out to the side. Pain at night, frequently worst at night, and particularly lying on that side. In longer-standing cases, strength loss and difficulty dressing or reaching a back pocket.

The tests, and why they prove less than you would hope

From a meta-analysis of 21 tests:

Test	Sensitivity	Specificity
Neer’s sign	78%	58%
Hawkins-Kennedy	74%	57%
Empty can	69%	62%
Drop arm	21%	92%
Lift-off	42%	97%

Look at the specificity column for the top three. Between 57 and 62 percent is barely better than a coin toss, which means a positive test shifts the probability of any particular diagnosis very little.

A separate high-quality review excluded Neer, Hawkins-Kennedy and empty can from its evidence entirely, and found the painful arc the most useful single finding, with a positive likelihood ratio of 3.7. The lag signs were the most useful for detecting full thickness tears.

The practical implication: these tests are useful for reproducing your symptoms and narrowing the region. They are not useful for telling you which structure is at fault, and any clinician who uses one to give you a precise structural diagnosis is overreaching.

Scans

Most guidance now says do not scan a first presentation of non-traumatic shoulder pain. One primary care pathway states routine imaging is not recommended, and that a minimum of 12 weeks is required to see if rehabilitation helps.

The reason is the same asymptomatic findings problem that undermines the rest of this diagnosis. A meta-analysis of 17 studies covering 5,988 people found rotator cuff tears in 23.6 percent of people who had never presented with a shoulder complaint. By age: 18.8 percent of 40 to 49 year olds, 23.7 percent of 50 to 59 year olds, 29.7 percent of 60 to 69 year olds.

As the authors put it, the detection of tears in asymptomatic individuals indicates that structural abnormalities do not necessarily correspond to symptomatic disease.

What actually helps

Exercise

The best supported option.

A network meta-analysis covering 99 trials and 6,764 participants across 20 treatment options found exercise had the most robust evidence across timeframes, and was the only option with moderate certainty for functional improvement up to three months. Most other evidence was low or very low certainty.

The effect size was modest, at 0.39 for function at three to six months. But it beat everything else, and it matched surgery in the trials above.

The program that worked

The Holmgren trial is worth describing because it recruited patients who were already on a surgical waiting list, all of whom had failed at least three months of conservative care.

The specific program was six exercises:

Two eccentric rotator cuff strengthening exercises
Three exercises for the shoulder blade stabilizers
One posterior shoulder stretch

Dosage: strengthening at 15 repetitions, 3 sets, twice daily for 8 weeks, then once daily for 4 weeks. The stretch for 30 to 60 seconds, 3 repetitions, twice daily. Load progressed with weights and elastic bands.

The control group did six unspecific movement exercises with no progression over 12 weeks.

Results: the specific group improved 24 points on the Constant score against 9 points for the control, a 15 point difference.

And the number that matters most: 20 percent of the specific exercise group went on to have surgery, against 63 percent of the control group.

Note what separated the two programs. Both groups exercised. One progressed the load and targeted it. The other did gentle movements that never got harder. If your program is not getting harder, it is the control arm.

Cortisone injection

A meta-analysis of 11 trials covering 726 patients found a pain benefit against placebo at four to eight weeks, with a number needed to treat of around five. Meaning roughly five people must be injected for one to get a meaningful transient reduction in pain.

There was no advantage over placebo at three months. Multiple injections were no better than a single one. The authors concluded injections cannot modify the natural course of the condition.

Useful for getting you comfortable enough to exercise. Not a treatment in itself.

Shockwave therapy

The Cochrane review of 32 trials and 2,281 participants found 42 out of 100 people achieved at least half their pain relief at three months against 38 out of 100 with placebo. Pain was 0.78 points better on a 0 to 10 scale. Side effects occurred in 26 per 100 against 7 per 100.

Their conclusion: very few clinically important benefits, and uncertainty regarding safety.

Manual therapy

The Cochrane review covering 60 trials found only one trial comparing a realistic manual therapy and exercise package against placebo. Pain improved by 6.8 points on a 100 point scale, not statistically significant. Function by 7.1 points, which was significant but small. Mild adverse events, mostly transient soreness, occurred in 31 per 100 against 8 per 100.

Electrotherapy, laser and ultrasound

The Cochrane review of 47 trials found the therapeutic ultrasound evidence rests on a single 61-participant trial in calcific tendinitis. Low-level laser showed benefit up to three weeks. TENS evidence was insufficient. Pulsed electromagnetic field therapy showed no clinically important benefit.

How long does it take?

Two well-conducted cohorts disagree, and you should see both.

One followed 349 patients with new shoulder pain through Dutch general practice and found 23 percent completely recovered at one month and 59 percent at twelve months, leaving around 41 percent with persistent symptoms at a year.

Another followed 166 patients and found only 21 percent completely recovered at six months and 49 percent at eighteen months, concluding that shoulder pain is not typically short-lived.

The honest statement: roughly a quarter of people are better within a month, and somewhere between 40 and 50 percent still have symptoms a year later.

Poor prognosis was associated with neck pain alongside the shoulder pain, and severe daytime pain at first presentation. Faster recovery was associated with a clear overuse or minor trauma trigger, and presenting early.

This is worth knowing because the standard reassurance, that it will settle in a few months, is true for around half of people and not for the rest.

When to seek urgent attention

Emergency, same day:

Feeling systemically unwell with fever and unexplained shoulder swelling, or pain unrelated to activity and not relieved by rest, which suggests joint infection
Shoulder pain with chest pain, breathlessness or sweating, which can be cardiac and needs an ambulance
Obvious deformity after trauma, with or without loss of movement
Sudden swelling, bruising or deformity of the upper arm or chest suggesting a muscle rupture

Within a week: a history of cancer, night sweats, unremitting night pain, or an unexplained lump.

A careful note on night pain. It is one of the most typical features of ordinary subacromial pain, and it is also on the red flag list. What separates them is company: night pain alone in an otherwise well person with an activity-related shoulder problem is expected. Night pain that does not let up, in someone with a cancer history or systemic symptoms, is different.

What to do with all this

If you have been told you have impingement:

The label describes where your pain is, not what is wrong. The mechanism it implies is not well supported.

Progressive, targeted, loaded exercise is the treatment with the best evidence, and it needs to actually progress over at least 12 weeks.

An injection may help you get started. It will not change the course of the condition.

Decompression surgery performs the same as a fake operation at every follow-up point out to ten years, and the strongest available recommendation is against it.

And if you are in the 40 percent still struggling at a year, that is a known outcome rather than a personal failure. It is also the point to ask whether the diagnosis is right, rather than repeating the same treatment.

Related conditions

Rotator cuff tendinopathy · Rotator cuff tears · Subacromial bursitis · Cuff tear arthropathy

This article is for general education and is not medical advice. The evidence summarised here applies to ordinary subacromial pain without a significant tear. Acute traumatic tears, calcific tendinitis and cuff repair surgery are different situations with different evidence.