What Works
1. The position audit Moderate
Find the positions provoking it and remove them. In the only randomised trial of conservative care, about 9 in 10 mild-to-moderate cases improved within 6 months doing exactly this, and nothing added on top beat it.
| Where to look | What's happening | What to change |
|---|---|---|
| How you sleep | Elbows curled for hours holds the nerve stretched. This is why it's worst at night | Keep the arm straighter. A towel loosely wrapped around the front of the elbow stops it folding |
| Your phone | Holding a phone to your ear is a long, hard bend | Speaker or headphones. Swap hands |
| Leaning on your elbows | The nerve sits right against bone with nothing to protect it. Desk, armrest, car door | Stop resting on that elbow. Pad the armrest. Move the chair |
| Driving | Elbow bent on the wheel, or propped on the door | Straighter arm. Get off the door |
| Work | Holding a tool or object in one fixed position is the one work factor actually linked to this condition | Break up the static hold. Change grip and position regularly |
2. If it comes to surgery, the smaller operation is the one High
Simple decompression and the bigger transposition operation produce equally good function, but simple decompression has significantly fewer complications. Equal benefit, less harm. This isn't your decision to make, but it's worth knowing before somebody offers you the bigger one.
Tier 2 & 3 — after surgery, and the experimental stuff
Early movement after surgery, rather than being immobilised Moderate
Supported, but the evidence is explicitly described as limited. Post-operative loading belongs to your surgeon.
Electrical stimulation after surgery, for severe cases Emerging
One double-blind trial of 31 people found better nerve recovery at 3 years. Genuinely promising in the stage where nothing else works, and genuinely tiny. Not established.
Low-level laser Emerging
Better than placebo for pain at 3 weeks, no different by 3 months. A short-term pain adjunct at most.
What Doesn't Work
- Ruling this out with a normal nerve test. Three out of four real cases come back normal. A normal result sends genuinely affected people away untreated.
- Steroid injection added to splinting. No evidence of effect.
- Blaming the extra muscle. The anconeus epitrochlearis, the anatomical variant usually blamed for causing this, turns up in 14.2% of healthy people and only 4.5% of patients. It might even be protective.
- Night splints and nerve-gliding exercises, as add-ons. Tested on top of education, they didn't add anything measurable. They're not harmful and they're not a waste of your time if they feel good. They're just not the part doing the work. The position change is.
Return to Training
- No numbness or tingling in the ring and little fingers at rest
- No night symptoms for 2 consecutive weeks
- Grip and pinch strength equal to the other side
- No muscle wasting between thumb and first finger, no curling fingers
- The provoking position has actually been identified and changed
- Training positions audited: no elbow-leaning, no long sustained bent-elbow holds
Most people never stop training for this. It was never a load problem. Keep lifting, stop resting on your elbows between sets, and watch the exercises that hold your elbow bent under tension for a long time. If your grip is getting weaker, that's a different conversation, and it's the one above in red.
Red Flags — Get Seen Now
- Your hand feels weak or clumsy, or you're dropping things
- The muscle between your thumb and first finger looks flatter than on the other side
- Your ring and little fingers are starting to curl and won't fully straighten
- Numbness has changed from coming-and-going to constant
Why this matters more here than in most conditions: once the nerve has been squeezed long enough to cause visible weakness or wasting, no treatment works reliably any more, including surgery. The window closes, and it doesn't reopen. This is the one condition on this site where "I'll give it a few more months" is genuinely the wrong call.
Refer to: hand surgery or orthopaedics. This is not an emergency room visit, but it is not a wait-and-see either.
Tonight, stop your elbow folding up while you sleep. Wrap a towel loosely around the front of your elbow so it can't bend much past a right angle. Eight hours curled up is the single biggest dose of this that most people get, and it's the one you can change tonight for free.
If your hand is already weak or clumsy, skip the towel and book an appointment instead. See the red flags above.
The Verdict
The nerve behind your elbow gets stretched every time you bend it. Stop the position, not the training.
Most nerves run down the front or the middle of a joint, where bending leaves them slack. This one runs behind the hinge of your elbow, so every bend does two things at once: it stretches the nerve longer and it tightens the tunnel it's sitting in. Sleep eight hours curled up and it's like kneeling on a garden hose all night. The flow doesn't stop, it just complains, and it complains louder the longer you kneel. It settles when you take the weight off, which is why the fix here is a position rather than a pill.
- Here's what's really happening: the nerve behind your elbow is the only one in your arm that has to stretch every single time you bend it, and eight hours curled up asleep is usually what tips it over the edge.
- The myth that won't die: that a normal nerve test means your nerve is fine. Three out of four people with obvious symptoms have a completely normal test.
- The one change that matters: find the position doing it, sleeping curled up, phone to your ear, elbow parked on a desk or car door, and stop that one thing.
Best for
Numbness or tingling in your ring and little fingers, worse at night or with your elbow bent, and your grip strength is still normal.
Skip if
Your hand is weak or clumsy, the muscle between your thumb and first finger looks flatter, or your fingers are starting to curl. That needs a hand surgeon now, not a self-help plan.
Conviction
Moderate
Stratified by claim, because the evidence is genuinely lopsided here. The surgical claims are strong (multiple randomised trials pooled, plus a network analysis of 2,894 limbs). The conservative claims rest on a single unreplicated trial of 57 followed patients, which is the best conservative evidence that exists for this condition. That's a statement about the field, not about the trial.
What would change my mind: "education and position change is the right first-line"
A multi-centre trial of 250+ patients with properly staged mild-to-moderate disease, randomising structured education alone against education plus a defined nerve-gliding dose against education plus night splinting, with blinded assessors and a validated cubital-tunnel-specific outcome score at 12 months. If nerve gliding separated from education by a meaningful margin, this recommendation would be overturned. That trial does not exist.
What would change my mind: "a normal nerve test doesn't rule it out"
A prospective study of consecutive patients with typical symptoms, all receiving both nerve conduction studies and high-resolution ultrasound against a proper reference standard, would either confirm or kill the 76%-normal figure. Right now it rests on one study.
Numb fingers are easy to ignore for a year, and this is one of the few conditions where that year actually costs you something.
Join The Verdict for free weekly protocols and stop guessing which injuries can wait.
The Full Picture — Anatomy, Diagnosis & Evidence
What's Actually Going On
The ulnar nerve runs behind the axis of the elbow joint. That single fact drives the whole condition. When the elbow bends, the nerve isn't simply compressed, it's lengthened and forced to slide, while the tunnel it runs through narrows as its roof tightens. Two provocative forces, not one, and both scale with how far the elbow is bent and how long it stays there.
The second route is direct pressure. At the bony bump on the inside of your elbow the nerve sits just under the skin, against bone, with almost no padding. That's why leaning on it reproduces symptoms so reliably.
Sustained squeezing and stretching produces, in order, numbness in the ring and little fingers, then weakness in the small muscles of the hand. The order matters. By the time the weakness is visible, the window in which anything works reliably has largely closed.
How to Identify It
- The distribution is the diagnosis: little finger plus the ulnar half of the ring finger. If your thumb and index finger are involved, it's something else.
- The key discriminator: is the back of your hand numb on the pinky side? The branch supplying it leaves the nerve above the wrist, so involvement points to the elbow, while a trapped nerve at the wrist spares it. One finding, no machine, separates the two most confusable diagnoses.
- Ultrasound, nerve cross-sectional area over 10mm² at the inner elbow Sn: 85% | Sp: 91% — the best confirmatory tool available.
- Scratch-collapse test Sn: 38% | Sp: 94% — confirms when positive, excludes nothing when negative.
- The elbow flexion test and Tinel's sign are in universal clinical use and no reliable accuracy figures were found for them. Use them, but don't give them weight they haven't earned.
The Debate
No clinical practice guideline exists for this condition as of July 2026. No NICE, APTA, BOA, EULAR or ACR guideline governs it. So the debate isn't guideline versus trial. It's the field arguing with itself, and with its own history.
The old answer (2000, 30 staged studies): conservative care has the highest recurrence; the bigger transposition operation is best for moderate disease.
The newer answer (randomised trials): no difference between the operations at all, and 89.5% of mild-to-moderate cases improved with education alone.
Why they disagree: the 2000 analysis pooled studies where patients were allocated by severity. Worse nerves got the bigger operation, and the conservatively-managed group was never a comparable population. Textbook confounding by indication. The randomised evidence wins on the operation question; the old analysis survives only for its severity-staging signal, which is still the most decision-relevant thing in this literature.
The anatomy lore: anatomical variants at the tunnel compress the nerve and cause the syndrome.
The pooled data: the anconeus epitrochlearis appears in 14.2% of healthy people versus 4.5% of patients (p<0.001), and may be protective.
Why they disagree: a denominator failure. Surgeons operate, see the variant in the patient in front of them, and infer causation. Nobody was counting how often it sits harmlessly in people who never get the condition. Pool the healthy controls and the correlation reverses.
Honest Limitations
The evidence answers the surgeon's question, not the therapist's. Of 40 papers retrieved, the overwhelming majority compare one operation to another. Exactly one randomised study addresses conservative care. A therapist reading this literature finds a beautifully mapped surgical decision tree and almost nothing about what to do in clinic on a Tuesday.
The condition gets better on its own about 90% of the time, which makes everything look effective. Anything you layer on top of a ~90% improvement rate will appear to work. That's exactly why the control arm in that trial was so useful, and it's why treatments here should be judged against education rather than against nothing.
Outcome reporting is chaotic, so the pooled numbers are softer than they look. Across 101 studies this condition generated 45 different outcomes and 31 different measurement instruments, and there's still no validated, widely-used, cubital-tunnel-specific score. When analyses pool "improvement," they're pooling instruments that don't measure the same thing.
Every paper retrieved was abstract-only. The direction of every finding here is reliable. The precise numbers are quoted as the studies stated them and should be read as hedged.
The Nuance
Conservative success: 89.5% of mild-to-moderate patients improved at 6 months with education and avoiding provocation (57 patients followed).
Surgical success: 87% improved across 30 studies and 2,894 limbs, with a 3% complication rate.
Those two numbers look almost identical and must not be read as a head-to-head. They're different populations. The conservative figure is mild-to-moderate disease only. The surgical cohorts averaged 15 months of symptoms before their operation and include the severe cases that conservative care can't help. Nobody has randomised surgery against conservative care in a matched group, so the comparison simply doesn't exist.
What the evidence does support is a staged answer. In mild-to-moderate disease with normal strength, explaining the condition and removing the provocation is a legitimate first move with roughly nine-in-ten odds. Once weakness appears, that logic expires, because severe disease responds poorly to everything and surgery at least offers 87% at a 3% complication cost. And if the conversation does reach surgery, the evidence is unusually clear for once: the simple decompression, not the bigger transposition. Equal function, fewer complications, no reason to accept the larger operation.
Sources
- Svernlöv B, et al., 2009, J Hand Surg Eur Vol (PMID 19282413). Randomised, 70 allocated / 57 followed. Night splinting vs nerve gliding vs control, all with education. 89.5% improved, no difference between groups. 76% had normal neurophysiology. The single most important paper here.
- Zlowodzki M, et al., 2007, J Bone Joint Surg Am (PMID 18056489). Meta-analysis of 4 randomised trials. Decompression vs transposition: no difference (p=0.81).
- Chen HW, et al., 2014, Clin Neurol Neurosurg (PMID 25255159). Meta-analysis, 13 studies, 1,009 patients. Complications significantly lower with simple decompression (OR 0.32).
- Wade RG, et al., 2020, JAMA Netw Open (PMID 33231636). Network meta-analysis, 30 studies, 2,894 limbs, 8 operations. 87% improve; 3% complications; all in situ decompression beat any transposition.
- Mowlavi A, et al., 2000, Plast Reconstr Surg (PMID 10946931). 30 staged studies. No modality consistently effective at the severe stage.
- Chang KV, et al., 2018, Arch Phys Med Rehabil (PMID 28888384). Ultrasound meta-analysis, 14 trials. 10mm² cutoff: Sn 0.85, Sp 0.91.
- Jain NS, et al., 2024, Hand (N Y) (PMID 37222286). Scratch-collapse systematic review. Sn 38%, Sp 94%.
- Suwannakhan A, et al., 2021, The Surgeon (PMID 33551294). Prevalence meta-analysis, 40 studies. Anconeus epitrochlearis 14.2% healthy vs 4.5% patients (p<0.001).
- van Rijn RM, et al., 2009, Rheumatology (Oxford) (PMID 19224937). Occupational review. Cubital tunnel: holding a tool in position, OR 3.53.
- Gallo L, et al., 2020, J Hand Surg (PMID 32591175). 101 studies, 45 outcomes, 31 instruments. Why there's no MCID here.
Full 27-source evidence trail with quality ratings and conflict matrix in the research file. Every paper retrieved was abstract-only; effect directions are reliable, precise magnitudes are as-stated by the studies.