Start Here Explore Stack Verdict Skip List Supplements Subscribe
The VerdictMODERATE CONVICTION

Your collarbone will heal. Watch the arm for new numbness, and keep the hand and elbow moving.

If you have any new numbness, pins and needles, weakness or coldness in the arm on the broken side, go back today. It can start a week or two after the injury even if the hospital checked you and everything was normal. Otherwise, right now: with the sling off and your arm resting in your lap, open and close your hand, bend and straighten your wrist, then bend and straighten your elbow. Ten of each. None of that should hurt at the collarbone.

  1. Here's what's really happening: the bone almost always knits, and how much the collarbone shortens has not been shown to predict how well the shoulder ends up working.
  2. What most people get wrong: assuming that a normal check at the hospital means the nerve risk has passed. In documented cases the symptoms started 8 to 30 days later in people who were completely normal on the day.
  3. What to watch for: any new numbness, tingling, weakness or coldness in that arm, at any point over the first several weeks. That is a same-day return, not a next-appointment mention.

The collarbone is a strut, like the single diagonal brace holding a shelf out from a wall. Snap the brace and the shelf drops and swings inward, which is why your shoulder sits lower and further forward on that side. The sling is not a clamp putting the brace back together. It is a prop taking the weight of the shelf off the broken ends so they can knit. And because the arm's nerve bundle runs immediately behind that brace, swelling and healing tissue in the gap can start pressing on it days after the injury, long after the original impact.

SH
Dr. Seth Holbrook, DPT — Doctor of Physical Therapy • Coach to 300+ clients
I built The Verdict to cut through recycled health advice and show what the evidence actually supports.

Shoulder

Midshaft Clavicle Fracture

A break through the middle of the collarbone. The bone almost always heals. The thing to watch is the arm, and it can start well after the hospital has sent you home.

CONVICTION: MODERATE
Read this first
If you have any new numbness, pins and needles, weakness or coldness in the arm on the broken side, go back and be seen today. This can start a week or two after the injury even if you were checked at the hospital and everything was normal. Do not wait for your next appointment.

Otherwise, right now: sling off, arm resting in your lap. Open and close your hand, bend and straighten your wrist, then bend and straighten your elbow. Ten of each.

Do that four or five times a day from day one. None of it should hurt at the collarbone, and if it does, keep your elbow tucked closer to your body. Stiffness in the hand and elbow is the one part of this injury that is entirely avoidable.

The Verdict

Your collarbone will heal. Watch the arm for new numbness, and keep the hand and elbow moving.

The collarbone is a strut, like the single diagonal brace holding a shelf out from a wall. Snap the brace and the shelf drops and swings inward, which is exactly why your shoulder sits lower and further forward on that side. The sling is not a clamp putting the brace back together. It is a prop taking the weight of the shelf off the broken ends so they can knit.

The part that catches people out is what runs behind the brace. The nerve bundle for the whole arm passes directly behind the collarbone, so swelling and healing tissue filling the gap can start pressing on it days after the impact, long after the injury itself is old news.

  1. Here's what's really happening: the bone almost always knits, and how much the collarbone shortens has not been shown to predict how well the shoulder ends up working.
  2. What most people get wrong: assuming a normal check at the hospital means the nerve risk has passed. In documented cases the symptoms started 8 to 30 days later, in people who were completely normal on the day.
  3. What to watch for: any new numbness, tingling, weakness or coldness in that arm at any point over the first several weeks. That is a same-day return, not a next-appointment mention.

Best for

Anyone with a confirmed middle-third collarbone break who has been told to manage it in a sling, and who wants to know what to expect and what to watch for.

Skip if

You have not had an X-ray, or you have tented or broken skin over the break, breathlessness, or any numbness or weakness right now. Those need assessing today, not reading about.

Want the full evidence? Keep scrolling. ⌄

What Works

Dark cinematic study of a shoulder girdle supported in a sling

Tier 1 — Route the surgical decision, and frame it honestly STRONG

Plate fixation cuts the risk of the bone failing to knit by roughly 85% (RR 0.14, 95% CI 0.06 to 0.32, Woltz 2017; independently replicated at RR 0.15, 95% CI 0.08 to 0.31, Ahmed 2018) and buys a genuine functional advantage for about six weeks. By twelve months the two paths are indistinguishable (DASH p=0.277 and Constant p=0.184, Ban 2021; DASH p=0.877, Tamaoki 2017).

The number needed to treat to prevent one symptomatic nonunion by operating on everybody is 6.2. Five of those six operations are on people who would have healed anyway.

This is the only Tier 1 item, and it is a referral decision rather than a therapy. That is a finding, not an omission. No physical therapy intervention for this fracture has strong evidence, because none has been compared against anything.

Tier 2 — Moderate evidence MODERATE

A simple sling rather than a figure-of-eight harness, worn for comfort. The AAOS 2022 guideline states this preference on work-group opinion, because the evidence is insufficient to show any brace superior.

Ask about smoking at the first visit and act on it. Smoking carries a relative risk of 3.68 (95% CI 1.87 to 7.23) for nonunion with conservative treatment, from 8 studies and 2,285 observations (Dietrich 2023). It is the largest modifiable risk factor in the condition, and the case for surgery is genuinely stronger in a smoker.

Set expectations against the real recovery curve. Pooled across 18 RCTs and 1,201 patients: DASH 16.1 at 3 months, 9.9 at 6 months, 6.3 at 12 months, with pain falling from 4.4 to 2.6 over the same span (Vaajala 2026). Most of the recovery is banked by three months and the curve is nearly flat after six.

Tier 3 — emerging evidence and clinical convention

Kinesiotaping alongside the sling EMERGING. One randomised trial (N=40) found better function scores at three months and return to work at 5.37 versus 7.23 weeks. A single small single-centre trial with no replication. Worth noting that its own abstract concludes it produced "higher union rates" while reporting only union times and no union-rate comparison at all, so it is cited here for pain, early function and return to work, and not for union (Dedeoglu 2022).

Graded, pain-guided restoration of active shoulder motion once the fracture is comfortable, with the elbow, wrist and hand kept fully mobile from day one. Clinical convention, ungraded, and labelled that way deliberately. No trial in this evidence base compares any motion protocol against any other, or against none.

Exercise Prescription

Read this before the table. The timings and repetitions below are standard clinical practice, not proven numbers. No study has ever compared one collarbone rehabilitation plan against another, so this is a sensible plan built on how bone heals rather than a protocol with trial evidence behind it. It is given because withholding care is not the alternative, and it is labelled so you know which is which.

ExerciseSets × RepsFrequencyPain guide
Hand, wrist and elbow movement. Arm supported in your lap, sling off10 of each4 to 5× daily, from day oneShould not hurt at the collarbone at all
Pendulum swings. Lean forward, let the arm hang and swing gently. Let gravity move it30 seconds each direction3× daily, from around week 1 to 2Mild discomfort fine, sharp pain means stop
Shoulder blade squeezes. Sitting upright, draw both blades back and down. The arm does not move2 × 10, hold 5 seconds2 to 3× dailyEffort between the blades, nothing sharp at the break
Assisted arm raise. Use the other hand or a stick to help lift the arm forwards2 × 10Daily, typically from week 3 to 4Stop where it turns sharp
Resisted rows and external rotation with a light band, elbow tucked to your side3 × 12Every other day, once healedMuscle fatigue is the goal

What Doesn't Work

  • Low-intensity pulsed ultrasound. The AAOS 2022 guideline recommends against it for nonoperative management of acute midshaft fractures. It does not accelerate healing and does not reduce nonunion.
  • The figure-of-eight harness used as a reduction device. Uncomfortable, produces axillary pressure, and no trial evidence shows it superior to a sling on any outcome that matters to a patient.
  • Treating more than 2 cm of shortening as a known harm. The only systematic review that tested the association did not find one, for shoulder function or for nonunion. Refer for the opinion. Do not deliver the prognosis.
  • Reassuring anyone that a normal emergency department nerve check means the risk has passed. All three documented delayed palsies were normal at presentation.
  • Presenting a sling duration or a start-of-loading date as evidence-based. None of them is.

Red Flags

Dark cinematic anatomical study of the clavicle and the neurovascular structures passing behind it
Refer immediately
  • Open fracture, or skin over the break that is tented, blanched, or breaking down. Risk of skin necrosis and conversion to an open fracture. EMERGENT
  • Posteriorly displaced medial-third fracture. Sits directly in front of the trachea, oesophagus and great vessels. EMERGENT
  • Neurovascular compromise at presentation, or a floating shoulder. EMERGENT
  • Breathlessness, chest wall pain, or altered breath sounds, especially after a high-energy mechanism. Associated rib fracture, haemopneumothorax and scapular fracture are described alongside these injuries (Wang 2025). EMERGENT
  • New neurological symptoms appearing days to weeks after an examination that was normal. Documented onset at 8, 30 and 14 days in three patients aged 62 to 70, none of whom had any neurovascular symptoms when first assessed (Saito 2020). URGENT
  • More than 100% displacement, comminution, more than 2 cm shortening, or an unstable distal pattern. Referred for a surgical opinion, not because harm is proven (Serpico 2021). URGENT
  • Pain escalating rather than settling after the first two weeks, or movement at the fracture with no bridging callus at three months. Around 14% of nonoperatively managed displaced fractures reach twelve months with a nonunion (Vaajala 2026). URGENT

The evidence grade on the delayed nerve flag, stated rather than implied. It rests on three cases. Three cases is not an incidence rate and none is claimed here. What they do establish is that the window exists and that it opens after the emergency department has discharged the patient. All three were displaced middle-third fractures managed in figure-of-eight bandages, which is a plausible association and not a demonstrated cause. Surgical exploration found a ruptured subclavius muscle and granulation tissue compressing the nerve bundle in every case, and full recovery took 6 to 11 months. Screening the limb at every contact costs nothing and is the only realistic way this gets caught.

Return to Training

Stated plainly: no validated return-to-activity milestone battery exists for this fracture. The best available review establishes only that operative management improves return rates and times to sport, with no numeric criteria published (Robertson 2016). The list above is built from union status, symmetry and symptom response, which are the measurable things available. It is convention, not a validated tool.

Loading, deliberately without percentages. Roughly 6 to 12 weeks with nothing loaded through the injured shoulder girdle, including a bar across the back. Hip-loaded lower body work continues unbroken from day one, as does upper body work on the uninjured side. Rebuild through below-shoulder-height pressing and pulling before anything overhead. No percentage-of-one-rep-max landmarks are given here on purpose: this condition's evidence base carries no exercise dose at all, and importing them from general resistance-training practice would dress convention up as evidence.

Conviction

MODERATE

The direction of the surgical findings is solid: the reduction in nonunion replicates across four independent syntheses and two randomised trials, and so does the twelve-month convergence in function. What is weak is everything about the rehabilitation, and the precision of any individual pooled number.

What would change this: a randomised trial of the rehabilitation rather than the surgery. At least 250 adults with nonoperatively managed displaced midshaft fractures, randomised to sling-for-comfort with unrestricted pain-guided motion from week one versus strict immobilisation for four weeks then graded motion, primary endpoint DASH at six weeks, nonunion at twelve months as co-primary, reported so it can be replicated. Nothing resembling it exists.

What would change my mind on the shortening threshold
A prospective cohort of at least 400 nonoperatively managed displaced fractures with shortening measured on CT as a continuous exposure, DASH at twelve months as primary and nonunion as co-primary, powered to find a breakpoint by spline regression rather than by a pre-chosen 2 cm cut. If a breakpoint exists, the criterion earns its place. If none exists, the threshold in current guidelines should be retired.
What would change my mind on the surgery question
Nothing, and that is the honest answer. Six independent syntheses and two clean randomised trials agree on both the reduction in nonunion and the twelve-month convergence. Further surgery-versus-sling trials would be a poor use of the field's capacity while the rehabilitation question sits untouched. The one caveat worth carrying is that this whole literature has a fragility index of 4, so trust the direction and not the decimal places.
The Full Picture — Anatomy, Diagnosis & Evidence

What's Actually Going On

Dark cinematic anatomical rendering of the clavicle as a strut between sternum and shoulder

The clavicle is the only bony strut between the axial skeleton and the arm. It fails in the middle third because that is where the bone is thinnest in cross-section and where nothing crosses it: no muscle, no ligament, no joint capsule.

Medially, sternocleidomastoid pulls the inner fragment upwards. Laterally, the weight of the arm drags the outer fragment down while the chest and scapular muscles pull it inwards, which shortens the strut. That is the whole mechanism of the classic deformity, and it is also why a sling changes the fracture position very little. A sling supports the weight of the arm. It does not reduce anything and was never designed to.

Healing is ordinary secondary bone healing. Union runs around 8 to 12 weeks where it is reported, and plate fixation shortens time to union by about 5.1 weeks (Guerra 2019). Where union fails, the mean time to union after surgery for a nonunion is 13.6 weeks.

How to Identify It

Dark cinematic study of shoulder girdle contour and palpation landmarks
  • Focal bony tenderness with a palpable step over the middle third. Sn/Sp: DATA UNAVAILABLE
  • Radiograph, AP plus cephalic tilt. This is what makes the diagnosis. Sn/Sp: DATA UNAVAILABLE
  • Whole-limb neurological screen, C5 to T1 myotomes and dermatomes plus the radial pulse, repeated at every visit rather than only the first.
  • Piano Key sign: unreliable here. When a midshaft fracture and an acromioclavicular joint injury coexist, the sign is typically negative because the lateral clavicle is floating, and that combined injury is commonly missed (Wang 2025).

Why two cells above say DATA UNAVAILABLE rather than carrying a number. Across 73 retrieved papers and roughly twenty randomised trials, not one reports a sensitivity or specificity for any clinical test in this condition. That is not a gap in the search, it is what the literature contains. Borrowing a plausible figure from general shoulder examination and attaching it to this fracture would be inventing evidence, so the cells stay empty and say so.

The Debate

The referral threshold nobody has validated

Serpico 2021 / Ropars 2017

Urgent surgical referral at more than 2 cm of shortening. Or surgery at more than 1.5 cm in young active patients.

vs

Malik 2019, systematic review, 16 studies, PROSPERO-registered

No significant association between shortening and shoulder outcome scores in 11 of 12 comparative series and 3 of 4 randomised trials, and none with nonunion.

Keep the referral, because it buys a surgical opinion and a shared decision. Drop the prognosis attached to it. The two sources quoting a threshold disagree with each other by half a centimetre, which is itself evidence that no measured breakpoint exists.

Does the surgical advantage grow, or converge?

Yan 2022

Surgery superior beyond 24 months, with a large pooled effect on disability scores.

vs

Ban 2021, Tamaoki 2017, Ahmed 2018, Woltz 2017

No clinically relevant difference at 12 months, across two randomised trials and two meta-analyses.

Follow the convergence. Yan's own paper concludes its significant effects fall short of a clinically important difference, and its beyond-24-month estimate is an order of magnitude larger than its own 6-to-12-month one, which is a heterogeneity signature rather than a benefit that grows with time.

Honest Limitations

The trial literature is statistically fragile

Across 30 randomised trials and 250 outcomes, the fragility index is 4, meaning that reversing four events flips the significance of a typical result. Worse, 63.3% of those trials lost more patients to follow-up than their own fragility index (Megafu 2025). Trust the direction of the nonunion effect, which replicates widely. Do not defend any specific pooled magnitude, and never present a risk ratio to a patient as a precise personal probability.

Seventeen different outcome measures, and the reviews pool across them

151 studies covering 15,853 fractures used 17 different patient-reported measures, averaging 1.6 per study, with significant variation by geography (Hao 2022). Cross-study comparison is partly an artefact of instrument choice. Pick one measure, use it every time, and compare the patient against their own baseline rather than a published pooled score.

Almost all of the research asks whether to operate

A review that went looking specifically for the effect of immobilisation and rehabilitation found ten studies, of which four were Level I, and reported no direct comparison between rehabilitation protocols and no study isolating either effect (Catapano 2019). Nothing across 73 retrieved papers reports a replicable exercise dose. That is a research-funding gap, not a verdict on the treatment: the question was never asked, which is different from being asked and answered badly. The honest response is the one this page takes, which is to give the plan and say which parts of it are proven and which are convention.

The Nuance

Dark cinematic anatomical comparison across the clavicle from sternal to acromial end

Surgery is not the safe option. It is a different set of risks that lands earlier and more predictably.

ConservativeSurgery
Nonunion at 12 months~14% (Vaajala 2026)Near-eliminated (RR 0.14 to 0.15)
Time to unionReference5.1 weeks shorter (Guerra 2019)
Function at 12 monthsNo clinically relevant difference
Complications incl. reintervention20.5%31.3% (Guerra 2019)
Paraesthesia near the site2.1%13.7% (Tamaoki 2017)
Secondary surgery, plate removal counted16.6%17.6% (Woltz 2017)

Roughly a quarter of patients have a second operation whichever path they take, just for different reasons: mandatory surgery for a nonunion on one side, elective plate removal for implant irritation on the other. And a nonunion is not automatically a problem to be fixed, since one third of them received no further treatment at all (Woltz 2017).

Two groups where the picture shifts. In adolescents, pooled data across 7 studies and 578 patients found no difference in function, with significantly fewer revision surgeries and fewer complications without surgery, and the AAOS guideline agrees that surgery may offer no benefit there. In smokers, conservative management carries triple the nonunion risk, and that is the one place where the surgical case genuinely strengthens.

Sources

Go Deeper

Most injury advice never tells you which parts are proven and which parts are just custom. The Verdict does, one condition at a time, free every week.

Join The Verdict

Dealing with something specific?

Every pain and rehab verdict, evidence-scored: what actually speeds recovery, what to skip, and when to get it checked.

Browse Pain & Rehab verdicts
Or find your lane in 2 questions

Get weekly evidence-based rehab verdicts

Physio conditions reviewed against clinical evidence. What works, what doesn't, and what to do — from a practising physiotherapist.

Subscribe free

Want a coach, not just research?

The Verdict is built by the same team behind Precision Metrics — a physique and health coaching practice with 300+ clients coached. Dr. Seth Holbrook, DPT and Luke Holbrook lead the coaching.

Book a free consultation

Related free research

Pain & Rehab
Accessory Deep Peroneal Nerve — The Verdict
 Pain & Rehab
Dorsal Foot and Ankle Bone Spurs as a Nerve-Compression Cause — The Verdict
 Pain & Rehab
Charcot-Marie-Tooth Disease — The Verdict

There are 500+ more inside

Conviction-scored verdicts on supplements, nutrition, training, physio, and recovery.

Explore all Get weekly verdicts