Proximal Humerus Fracture

Touch the outer point of your shoulder, the patch a t-shirt sleeve covers. Numb? Then push your arm out sideways against your other hand. Weak too? Emergency room today.

If both are normal, your next job is movement rather than rest, and it starts this week.

Takes under 30 seconds. No equipment needed.

The Verdict

Broken shoulder? Start moving it this week, not next month. That one change is the best-tested decision here.

Bone knits back together on its own schedule, roughly six to twelve weeks, and nothing you do speeds that up. But the joint around it behaves like a door hinge left unused. Leave it still for a month and it stiffens shut, and now you are fixing two problems instead of one. Gentle early movement does not rush the bone. It stops the hinge seizing while the bone gets on with it.

Here's what's really happening: the bone almost always heals on its own, and an operation was tested against a simple sling in 250 people and made no difference at two years, or again at five.
The one thing that makes it worse: keeping the arm completely still for three or four weeks, because six trials compared that against starting gentle movement in the first week, and waiting lost.
Start here: get the hand, wrist and elbow moving from day one, and gentle pendulum swings going inside the first week.

Best for

Adults recovering from a broken upper arm bone without surgery, especially over 60 after a fall from standing height. This is by far the most common version of the injury.

Skip if

You have numbness over the outer shoulder, weakness lifting the arm, a cold or pale hand, or broken skin. Those need urgent assessment, not exercises. Also get an orthopaedic opinion first if you are under 50, or if you were told your break is in three or four pieces.

Want the full evidence? Keep scrolling

What Works

Dark cinematic rendering of a shoulder in early guided motion, emphasising the joint capsule and surrounding musculature

Tier 1 · Non-surgical management as the default HIGH

For the commonest pattern, a break through the narrow waist below the ball, a sling plus standard rehabilitation is the benchmark. PROFHER randomized 250 patients across 32 UK hospitals to surgery or a sling and found a 0.75-point difference on a 48-point shoulder score, against the 5 points that would count as meaningful. Unchanged at five years. Surgery also cost £1,758 more per patient with slightly worse quality-of-life scores.

Tier 1 · Start moving within one week HIGH for safetyMODERATE for benefit

Six randomized trials, 470 patients, comparing movement started inside one week against three to four weeks of immobilisation. Early won on combined function at 3 months and showed no difference at 6 or 12 months, with no increase in complications and no significant increase in the fracture shifting. Read the ceiling honestly: this buys a faster route to the same destination, not a better destination.

Exercise Prescription

Hand, wrist and elbow movement
10 of each · 4 to 5× daily · from day one
Arm supported in your lap. Should not hurt at the shoulder at all.

Pendulum swings
30 seconds each direction · 3× daily · starting within week 1
Lean forward, let the arm hang, let gravity move it. Mild discomfort is fine, sharp pain means stop.

Shoulder blade squeezes
2 × 10, hold 5 seconds · 2 to 3× daily
Sitting tall, draw both shoulder blades back and down. The arm itself does not move.

Assisted arm raise
2 × 10 · daily · from around week 2
Use your other hand or a stick to help lift the arm forward, only as far as comfortable.

Light band rows and outward rotation
3 × 12 · every other day · once healing is confirmed
Elbow tucked to your side. Muscle fatigue is the goal.

The timing above is backed by trial evidence. The specific sets and repetitions are not. No study has ever compared one exercise program against another for this fracture, so those numbers are standard clinical practice built on how bone heals, and your therapist will change them to suit you.

Tier 2 and Tier 3 — supervision, post-surgical timing, and delivery

A taught home program with review, rather than a standing weekly appointment MODERATE
For an older adult with a two-part fracture who can do the exercises. In the one trial to test this properly, 72 patients, both groups received a daily home exercise program. What was randomized was whether a weekly therapist visit was added on top. The difference at three months was 3.5 points on a 100-point disability score, well inside the margin of noise, with nothing at twelve months. This is a null result from a small trial rather than proof the two are equal, and it says nothing about the initial assessment, the red-flag screen, or the teaching, because every patient in the study got all three.

Early movement after a reverse shoulder replacement, rather than four to six weeks in a sling MODERATE
104 patients. Faster recovery of movement at 6 and 12 weeks, no difference at a year, and no difference in healing of the bony attachments or in complications. The same shape as the non-surgical timing finding, from a weaker study design.

Telerehabilitation where getting to a clinic is the barrier EMERGING
Studied across a mixed group of shoulder conditions at low to very low certainty, in which this fracture is one of four and only the undisplaced kind. Reasonable to offer on access grounds. Not something to claim works better.

What Doesn't Work

Routinely keeping the arm still for three to four weeks before starting movement. This was the comparison arm in six randomized trials and it lost, mildly, at three months, while gaining nothing at any point. It survives because it feels protective and because the fear of the fracture shifting is intuitive.
Acupuncture as an add-on. The pooled pain figure looks enormous and describes nothing: the seven trials behind it disagree with each other almost completely, function did not improve at all, and the reviewers themselves rate the study quality low.
Any set-and-repetition protocol sold as evidence-based. Fewer than 40 percent of upper-limb fracture exercise trials describe their program well enough for anyone to copy it, and across the 93 studies pulled for this page there is not one replicable prescription. Programs are still necessary. Calling them evidence-based is the part that is false.

Red Flags

Read this before anything else on the page.

$Cinematic anatomical rendering of the shoulder region highlighting the nerve and vascular structures that run beneath a proximal humerus fracture$

Numbness over the outer point of the shoulder, in the patch a t-shirt sleeve would cover, especially alongside weakness lifting the arm away from your body. This is the nerve that supplies the main shoulder muscle, and it changes the whole treatment plan.
A hand that is cold, pale, or has pins and needles that will not settle. This is a circulation problem, not a fracture symptom.
Skin over the shoulder that is stretched tight, gone white, or has broken open.
Trivial injury, no injury at all, or shoulder pain that was already there before the fall. A bone that breaks too easily needs a reason found.

Any of these: emergency room, today. Not a call-back appointment.

Expected, and not a red flag

Dramatic bruising spreading down the arm and across the chest over the first few days. That is gravity moving a bruise downward. It looks far worse than it is, and it frightens people badly precisely because nobody warns them it is coming.

Return to Training

Healing confirmed on X-ray, or the orthopaedic team confirms the fracture is stable
Full active shoulder movement forward, out to the side, and rotating outward, matching the uninjured side
No pain at rest and no night pain
Equal shoulder strength on both sides, with normal sensation over the outer point of the shoulder
A full session of below-shoulder loading with no symptom increase over the following 24 hours
For overhead lifting or contact sport: sign-off from the treating orthopaedic team

These criteria are clinical convention, not validated milestones. No study defines objective return-to-activity thresholds for this fracture, and none defines them at all for a younger or higher-demand patient. That is the largest evidence gap on this page.

In the meantime: lower-body training that does not need you to grip or brace with the injured arm can usually continue from week one. Nothing overhead, nothing carried in that hand, and nothing with a fall risk until healing is confirmed. Overhead pressing comes back last.

Conviction

Moderate Split by claim, because the evidence is genuinely uneven across them.

Non-surgical management as the default for the commonest patternHIGH Safety of starting movement within one weekHIGH Functional benefit of early movement at 3 monthsMODERATE Benefit of early movement still present at 12 monthsLOW Adding weekly supervision to a home program improves outcomesLOW Any specific exercise dose, sets, reps, or progressionNO EVIDENCE Return-to-activity milestonesNO EVIDENCE

What would change my mind on early movement

A trial of at least 400 patients treated without surgery, split by how many pieces the fracture is in, with the fracture shifting on X-ray as a co-primary outcome alongside function at twelve months. The current pooled estimate for shifting is compatible with anything from a slight reduction to a fivefold increase, and no existing trial is powered on it. A clear increase in the more shattered fractures would turn early movement from a default into a decision that depends on the fracture pattern.

What would change my mind on supervision

A properly designed non-inferiority trial of at least 300 patients comparing a single teaching session against a full supervised course, with a pre-set margin for what counts as "no worse". The existing 72-patient trial found no superiority, which is not the same as proving equivalence, and it cannot rule out a small benefit. It also enrolled only patients able to exercise independently, so it says nothing about anyone who is not.

Go Deeper

Most injury advice is convention wearing a lab coat. The Verdict pulls the actual trials on one condition a week and tells you which parts are tested and which parts are just what everyone does.

Join The Verdict — free

The Full Picture — Anatomy, Diagnosis & Evidence

What's Actually Going On

Dark cinematic anatomical study of the proximal humerus showing the ball, the two bony prominences, and the narrow neck below them

The top of the arm bone has three parts that matter. The ball that sits in the shoulder socket, and two bony bumps just below it where the rotator cuff tendons attach. Below those is a narrow waist, and that waist is where most of these fractures go.

The usual mechanism is a fall onto an outstretched hand or straight onto the shoulder, in someone whose bone is already thinned by osteoporosis. The rotator cuff muscles then pull the loose pieces in the same directions they normally pull the intact bone, which is why the pieces move in a predictable pattern rather than a random one. How many pieces there are, two, three, or four, is what drives the surgical conversation.

Healing is the expected outcome even in the more shattered patterns. In conservatively treated three-piece fractures the bone knitted in 95 percent of cases, and in four-piece fractures 91 percent. What separates a good outcome from a poor one is not whether the bone heals. It is which fracture you had. In that same series the average shoulder function score was 64.5 for three-piece and 54.9 for four-piece fractures. Both healed. One shoulder works considerably better than the other, and that is the honest version of "your fracture will heal".

How to Identify It

Cinematic clinical assessment scene showing examination of the shoulder and upper arm after trauma

The picture: a fall in an older adult, followed by immediate and total loss of the ability to lift the arm, with bruising appearing down the upper arm and into the chest wall over the following days. A rotator cuff tear or a frozen shoulder does not produce that combination suddenly, and neither follows a discrete fall with this much swelling.

There are no hands-on tests to report, and that is a measured finding

All 93 studies pulled for this page were scanned for test accuracy figures, the numbers that tell you how good a physical test is at catching a condition or ruling it out. Two results came back and both were false alarms: one was a sentence about a database search, the other a health-economics calculation. There is no orthopaedic special test with published accuracy for this fracture, because the diagnosis is made on imaging. An empty table here would look like something got left out. It did not.

What the assessment actually is: X-rays in three planes to characterise the fracture and rule out the ball having come out of the socket at the same time. A CT scan improves agreement between clinicians and is treated as the gold standard for deciding management. Imaging, not clinical testing

The one thing to test by hand, and to write down: the strength of the main shoulder muscle, plus sensation over the outer point of the shoulder. Weakness alone can just be pain putting the brakes on. Weakness with numbness in that patch points at the nerve, and that changes everything.

The Debate

How certain is "start moving early"?

Cochrane review, 2022 · 47 trials, 3,179 patients

Rated early versus delayed movement as very low certainty. "We are uncertain of the findings."

vs

Challoumas review, 2025 · 6 trials, 470 patients

Rated the same comparison moderate to high certainty, with early movement favoured at three months.

The gap is search date and method, not a disagreement about the trials themselves. Cochrane closed its search in September 2020 and graded each outcome separately. The 2025 review searched to January 2025, added a trial, and pooled a combined function score, which sharpens the estimate but mixes measuring instruments. Follow the newer review for the direction, keep the older one's caution about the size.

Does surgery ever win?

PROFHER, 2015 · 250 patients, mean age 66

Surgery gave 0.75 points on a 48-point scale against a sling, where 5 points is the smallest difference that would matter. Same at five years.

vs

Miquel trial, 2024 · 81 randomized, over-70s, 3- and 4-part

A shoulder replacement beat non-surgical treatment by 8.84 points on a different 100-point scale at one year.

These are adjacent populations, not a contradiction. PROFHER studied breaks through the narrow neck in a younger, mixed group. The 2024 trial studied more shattered fractures in the over-70s and used one specific operation. Note also that its 8.84-point result fell short of the 10 points the trial set out to detect, 66 of 81 patients finished, and 6.5 percent of the surgical group had a major complication against none in the other arm. A larger trial, PROFHER-2, is running to settle it and has not reported.

Honest Limitations

The supervision trial measured supervision, not physical therapy

What the study showed: adding a therapist visit once weekly for ten weeks to a daily home program produced a difference of 3.5 points on a 100-point disability score at three months, and nothing at all at twelve.

The gap: both groups received a home exercise program. Every patient in that trial was assessed, screened for red flags, given a program, and taught how to do it. What was randomized was the value of an extra appointment on top of all that. "Physical therapy doesn't help" inverts what was actually tested.

The adjustment: front-load the care. Invest in the first appointment and the teaching, then review rather than booking a standing weekly slot. Do not extend this to someone who cannot follow a program alone, because that person was not in the study.

The evidence is old, female, and low-demand. Plenty of patients are not

What the studies showed: every trial on this page ran on a fragility-fracture population. Average age 66 and 77 percent women in PROFHER; 74 years and 79 percent women in the conservative-management series; entry restricted to 60 and over in the supervision trial.

The gap: a 40-year-old who breaks this bone in a bike crash and wants to know when they can press overhead sits outside every study cited here. Reviewers name this gap directly.

The adjustment: say the extrapolation out loud rather than presenting age-mismatched evidence as if it applies. The timing finding is the part most likely to carry across. The outcome expectations and the surgery arithmetic are the parts least likely to.

The trials measured different things, so the numbers cannot be stacked

What the studies showed: across 74 studies, 22 different patient-reported questionnaires were in use, averaging 2.2 per study, with no professional consensus on a standard.

The gap: PROFHER's scale and the 2024 trial's scale cannot be placed on one axis, which is why the pooled figures on this page should be read as direction rather than exact size.

The adjustment: pick two measures and keep using the same two. The comparison that matters clinically is you against yourself at six weeks, not you against a trial average.

The Nuance

Dark cinematic anatomical comparison of shoulder structures involved in the main differential diagnoses

Surgery versus conservative, with the numbers on the table. For the commonest pattern the operation does not make the shoulder better. PROFHER randomized 250 patients across 32 hospitals and found three quarters of a point of difference on a 48-point scale, holding at two years and again at five, with the same number of patients in each group needing further surgery. It cost £1,758 more per patient for slightly fewer quality-adjusted life years, giving surgery under a 10 percent chance of being cost-effective at the UK threshold. An independent review puts the complication rate 3.3 times higher with surgery.

The genuinely unresolved question is narrower than the argument usually is. It is not "surgery or not" in general. It is whether a reverse shoulder replacement helps a patient over 70 whose fracture is in three or four pieces. One small trial says probably a little, by less than it set out to detect, at the cost of a 6.5 percent major complication rate. PROFHER-2 is running to answer it properly.

The other thing that gets missed entirely. Breaking this bone in a fall from standing height means the bone was already weak. The bone-health and falls assessment that should follow is the single most-skipped part of this injury, and it is the part that determines whether there is a next fracture.

Sources

Rangan A, et al. 2015. JAMA. PMID 25756440. The PROFHER randomized trial. 250 patients across 32 UK hospitals, surgery versus sling for displaced surgical-neck fractures. Oxford Shoulder Score difference 0.75 points (95% CI −1.33 to 2.84) against a 5-point threshold. Large pragmatic RCT.
Handoll HH, et al. 2017. Bone Joint J. PMID 28249980. PROFHER at five years, 149 patients with data. No statistically or clinically significant difference at three, four or five years. Extended RCT follow-up.
Challoumas D, et al. 2025. BMC Musculoskelet Disord. PMID 40016664. Six RCTs, 470 patients. Early (within one week) versus delayed (after three to four weeks) mobilisation: combined function SMD 0.4 (95% CI 0.1 to 0.7) at three months only. GRADE moderate to high. Meta-analysis.
Handoll HH, et al. 2022. Cochrane Database Syst Rev. PMID 35727196. 47 trials, 3,179 participants, 26 comparisons. Rates early versus delayed mobilisation very low certainty. Search closed September 2020. Cochrane systematic review.
Østergaard HK, et al. 2024. J Shoulder Elbow Surg. PMID 38311103. 72 patients aged 60+, displaced 2-part fractures. Supervised plus home exercise versus home exercise alone: DASH difference 3.5 (95% CI −5.0 to 12.5), not clinically relevant. Assessor-blinded multicentre RCT.
Hodgson SA, et al. 2003. J Bone Joint Surg Br. PMID 12729121. 86 patients. Immediate physiotherapy within one week versus delayed after three weeks: less pain and better function at 16 weeks, differences no longer significant at 52. Randomized controlled trial.
Soler-Peiro M, et al. 2020. J Orthop Surg Res. PMID 32831119. 133 patients, conservatively treated 3- and 4-part fractures. Union 95% and 91%, Constant 64.5 and 54.9, malunion 21%, avascular necrosis 9%. Systematic review, Level IV.
Bruder AM, et al. 2017. J Physiother. PMID 28941967. 22 trials, 1,299 participants with upper-limb fractures. Moderate evidence favouring early exercise with reduced immobilisation. Fewer than 40% of trials described the program well enough to replicate. Systematic review.
Corbacho B, et al. 2016. Bone Joint J. PMID 26850418. Economic evaluation alongside PROFHER. Surgery cost £1,758 more per patient with a QALY difference of −0.0101, under 10% probability of cost-effectiveness at the £20,000 threshold. Trial-based cost-utility analysis.
Miquel J, et al. 2024. J Shoulder Elbow Surg. PMID 38548095. 81 randomized, 66 completed, over-70s with 3- or 4-part fractures. Reverse shoulder arthroplasty versus non-operative: Constant difference 8.84 (95% CI 1.57 to 16.11) against a 10-point powering threshold, 6.5% major complications in the surgical arm. Randomized controlled trial.
Rudran B, et al. 2022. Br J Hosp Med. PMID 35938761. Assessment per British Orthopaedic Association standards with attention to axillary nerve and vascular status, three-plane radiographs, CT as gold standard. Names the evidence gap in younger patients. Narrative review.
Hohmann E, et al. 2023. J Shoulder Elbow Surg. PMID 36736653. 22 studies. No difference in clinical outcomes or range of motion between surgical and conservative treatment; complication rate 3.3 times higher with surgery. Authors rate the body of evidence very low certainty. Meta-analysis.

Additionally noted: NICE guideline NG38 recommends non-surgical management for uncomplicated displaced low-energy proximal humerus fractures in adults, reserving surgery for open injury, skin tenting, vascular injury, fracture-dislocation or head-split patterns. NICE guidance is not indexed in the literature databases searched for this page, so it is flagged as unverified here; its direction is independently corroborated by the trials and reviews above.

Broken shoulder? Start moving it this week, not next month.