A cervical traction device needs 10 to 25 lb of pull, lying down. A neck stretcher and an inflatable collar give you neither

Amazon calls five different things a cervical traction device. Two of them apply a measurable pull to your head: a pneumatic unit you lie on and pump, and an over-the-door pulley with a chin halter and a bag of water. One applies an unmeasured, uneven pull: an inflatable collar. One applies no pull at all: a foam wedge sold as a neck stretcher. And one is somewhere between: a strap cradle that hangs your head from a door. The evidence covers the first two, and it splits them, which is the whole point of this page. The pull, the position and the dose are what the trials tested, and most of what is sold under the name has none of the three.

What the trials found

The trial to know is Fritz and colleagues, in the Journal of Orthopaedic and Sports Physical Therapy in 2014. Eighty-six people with neck pain and signs of nerve root involvement were randomised to four weeks of exercise alone, exercise plus supine mechanical intermittent traction, or exercise plus seated over-the-door traction, and followed for a year. At six months the mechanical traction group was 13.3 points better than exercise alone on the 100-point Neck Disability Index (95% CI 5.6 to 21.0) and 8.1 points better than the over-door group (0.8 to 15.3); at twelve months it was still 9.8 points ahead of exercise alone. The over-door group was not significantly better than exercise alone at any point. The mechanical protocol, from the trial’s own description and the reviews that tabulate it: supine, 15° of flexion, 60 seconds on and 20 off, starting at about 12 lb and titrated to tolerance, 15 minutes. The over-door protocol: seated, continuous, 15 minutes, capped at 20 lb by the device.

The rest of the shelf is thinner. Young and colleagues, in 2009, randomised 81 people to manual therapy and exercise with real or sham traction and found no difference at four weeks, with no longer follow-up, which is the gap Fritz filled. Jellad and colleagues, in 39 people, found both manual and mechanical traction better than rehabilitation alone at the end of treatment, without reporting effect sizes. The 2008 Cochrane review of seven trials and 958 people, never updated, found the single low-risk trial showed no difference against placebo traction and concluded the literature neither supports nor refutes traction. Since then, Romeo’s 2018 meta-analysis of five radiculopathy trials found large but wide-interval effects for mechanical traction added to physiotherapy, Colombo’s 2020 meta-analysis of seven trials found a pain difference of 6 points on 100 and called it statistically significant but not clinically relevant, and Yang’s 2017 meta-analysis of intermittent traction for neck pain in general found a small effect immediately after treatment and none at follow-up.

The guideline reads all of that and lands where a buyer should. The American Physical Therapy Association’s 2017 neck pain guideline says clinicians should provide mechanical intermittent cervical traction for chronic neck pain with radiating pain, combined with stretching, strengthening and spinal mobilisation, at Grade B, moderate evidence. For acute radiating pain it does not recommend traction. For neck pain without arm symptoms, traction is one option in a multimodal list, and the evidence for it there is the Yang review’s “none at follow-up.”

So: if your neck pain runs into your arm and has for more than a few weeks, a supine traction device is a defensible purchase, as part of a programme a clinician has set. If your neck is stiff and sore and your arms are fine, the evidence for traction is thin to nothing, and the neck massager page is the honest place to shop.

The pull: what force does what

Traction is force, and the devices are sorted by whether they can deliver it.

• What the force does. Liu and colleagues put 15 healthy volunteers in an MRI under supine axial traction: the foraminal area, the opening the nerve root passes through, widened by 5.8% at 5 kg (11 lb), 16.6% at 10 kg (22 lb) and 18.9% at 15 kg, with no significant gain from 10 to 15 kg. Vaughn and colleagues measured vertebral separation on a Saunders unit at 25 lb for two minutes in 20 adults. Chung and colleagues imaged 29 patients and 7 volunteers under a portable intermittent device: the column lengthened in every volunteer and 21 of 29 patients, and disc herniations reduced fully in 3 and partly in 18, during the pull. What none of those studies shows is the separation lasting after the device comes off; the Cochrane review’s conclusion is about that.
• Position matters more than the product. Fater and Kernozek tested home units at 30 lb for five minutes in 17 people: supine traction increased posterior vertebral separation, seated traction did not. That is a mechanical reading of the Fritz result, where the supine unit worked and the seated over-door kit did not: sitting, part of the pull is spent lifting the head, and the muscles fight it.
• Angle. The classic 1960s radiographic work by Colachis and Strohm, quoted by every physiotherapy text since, put maximal posterior separation at about 24° of flexion with 25 to 30 lb; the trials used 15 to 25°. The originals are not online, so treat the millimetres as textbook, not verified.
• What an air collar does. Kang and Park inflated a collar to 40, 80 and 120 mmHg on 18 men and measured the neck muscles with electromyography: activity in the splenius capitis rose at 80 and 120 mmHg. The authors’ words are that it “did not meet the condition of muscle relaxation.” No study converts collar pressure into pounds of pull on the head, and none has imaged the neck inside one.
• What a foam wedge does. Nothing we could find. A search for any trial, cohort or imaging study of foam occipital wedges and “neck stretcher” pillows returned none.

Device capacity, from a state technology assessment that compared them: over-the-door kits under 20 lb, the Pronex pneumatic unit up to 20 lb, the Saunders home unit up to 50 lb at 15°. The trials’ useful range, 12 to 35 lb supine, sits inside the pneumatic units and at the edge of the over-door kits.

The five things sold as a cervical traction device

1. Pneumatic supine unit (Saunders HomeTrac, ComforTrac, Pronex). You lie on it, a wedge grips the base of the skull, and a hand pump applies a readable force at a set angle. This is the device class the trials used. The costliest by far, and the only one on this page with a gauge.
2. Over-the-door pulley (DMI and dozens of clones). A chin-and-occiput halter on a rope over a door-top pulley, with a water bag for weight. Cheap, and it is the arm of the Fritz trial that did not beat exercise. The halter is also where the jaw complaints come from.
3. Door-anchored strap cradle (the Neck Hammock). A sling that cradles the occiput, anchored to a door handle, so you lie down and your head hangs in it. The pull is your own head weight against the strap, a few pounds at most; no study has measured it. Supine, at least.
4. Inflatable collar. Rings that inflate between your jaw and shoulders. Comfortable for some, unmeasured in force, and the one device with an electromyography study against it.
5. Foam wedge, “neck stretcher,” “chiropractic pillow.” A curved block you lie on with the neck extended. No pull. No trial. A stretch that some people find relieving, and the thing a 9,000-a-month search for “neck stretcher” actually lands on.

Dangers of cervical traction

The FDA’s adverse-event database holds eight reports under cervical traction, six of them injuries, and the home ones are specific: a 10 lb water bag on an over-door kit producing a painful chin pull; an over-door rope that snapped so the crossbar hit the user’s head; a fainting episode on a Saunders unit; a jaw displacement and closed lock after seven minutes on a Saunders unit; “permanent” temporomandibular dysfunction on a Pronex; and an unregistered inflatable collar with an air pump that pressed on the jugular veins until the user choked and passed out. The five deaths in the traction database are halo, powered-table and surgical devices, not home units.

Two published cases involve the vertebral artery, which runs through the neck vertebrae and does not like being stretched. Aldarwish and colleagues, in 2024, describe a 57-year-old woman with degenerative disc disease who bought a neck traction device from an online advertisement, used it lying down for 10 to 15 minutes, and had a right vertebral artery dissection days later, which resolved over six months on aspirin. Dickinson and colleagues, in 1995, describe bilateral vertebral artery dissection and a posterior-circulation stroke after 8 lb of traction in a patient with a skull-base abnormality. The chin-halter jaw cases go back to the 1960s.

No peer-reviewed contraindication list exists; the ones in textbooks and insurers’ policies agree on: rheumatoid arthritis or any atlantoaxial instability, a recent neck injury or fracture, a tumour or infection in the spine, known vertebral artery or carotid disease, a skull-base abnormality, acute neck pain after trauma, uncontrolled hypertension for weighted over-door kits, and temporomandibular joint problems for any chin halter.

How to use cervical traction at home

The protocol below is the trials’ and the manuals’, in that order of authority:

1. Lying down, on a pneumatic unit or strap cradle, with the neck flexed about 15°. Seated traction did not separate vertebrae in the study that compared them, and it is where the over-door kit lost.
2. Start light. Fritz started at about 12 lb; Young started at 20. On a pneumatic unit that means a few pumps and a readable number; on an over-door kit, start the water bag at 8 to 10 lb and never past 20.
3. Intermittent. 50 to 60 seconds of pull, 10 to 20 seconds of release, for 15 minutes. Continuous pull was the seated kit’s protocol, and the Cochrane low-risk trial’s.
4. Build over sessions, not within one. Young’s protocol added 2 to 5 lb a visit to a 35 lb ceiling; the MRI data say the gain flattens above 22 lb.
5. Three to five minutes to start, once or twice a day, if you are on an over-door kit without a therapist; that is the convention from a 58-patient retrospective series and the kits’ own manuals.
6. Stop at jaw pain, dizziness, nausea, headache, visual change, or arm symptoms that get worse during the pull rather than better. Release slowly; the fainting case was coming off the device.
7. Do the exercise. In every trial, traction was added to exercise; in none was it used alone. The traction is the ten minutes; the programme is the treatment.

The best cervical traction devices: our picks

Every spec below is listing-stated on 26 October 2026, read from a US delivery session. We have not used any of them, and all of them sit in Amazon’s health category.

The device the trials used

The over-the-door kit

The strap cradle

The inflatable collar

The foam wedge, bought for what it is

Cervical traction vs the rest of the shelf

• A neck massager is the better-evidenced purchase for a stiff neck without arm symptoms, and nobody’s jaw has been dislocated by one.
• An inversion table is lumbar traction by the ankles, with its own long safety list; it does not treat a neck.
• A heating pad on the neck before the exercises is the cheapest thing on this shelf and the only one with a guideline recommendation for a sore spine.
• A massage chair neck program is where the one home-device stroke case report came from; see that page before running one.
• A knee brace page explains why most of this shelf is the exercise programme in disguise, and so is this one.

What we have not measured

Every spec here is listing-stated as of 26 October 2026. We have not pumped, hung, inflated or lain on any of these devices.

The measurement the whole category owes is the simplest on this site and nobody has published it: force. A hanging scale in series with each device, under a 10 lb head analogue on a sled, gives the pounds of pull at each pump count on the ComforTrac, at each bag weight on the DMI (and the rope’s breaking load, which is the FDA report), at each strap setting on the Neck Hammock, and at each inflation pressure on the collar, where we expect the answer to be close to zero in the axial direction and a few pounds sideways on the neck. A protractor on the ComforTrac’s three angles. And a stopwatch on the collar’s leak rate, because a collar that deflates over fifteen minutes is a different device at the end of a session than the start. A morning, n=1, which for a safety number is the limitation worth repeating.

What we would buy

If a clinician has said traction for pain that runs into your arm: the ComforTrac, used supine at 15°, intermittently, with the exercise programme, and read the third-party warranty first.

If the budget is a tenth of that and you will keep the bag under 20 lb: the DMI over-the-door kit, knowing it is the arm of the trial that did not win.

If you want the sensation lying down without a halter on your jaw: the Neck Hammock.

If you want a comfortable collar for a desk: the ELESIK, bought as a collar.

If what you actually want is a ten-minute stretch: the RESTCLOUD wedge, which is not traction and does not pretend to be once you read its listing closely. Our how we test page says what every figure here is and is not.