An ice machine for knee surgery cools the joint less than a bag of ice. That is the point, and it is also the risk

A cold therapy machine is three parts: an insulated reservoir you fill with ice and water, a small pump, and a pad with channels in it that wraps the joint. The pump pushes the ice water through the pad and back, so the knee sits under steady, even cold for as long as the ice lasts, instead of under a bag of ice that is too cold for ten minutes and then a bag of water. Surgeons’ offices have handed them out after knee surgery for thirty years, usually as a Breg Polar Care, a DonJoy IceMan or an Aircast Cryo/Cuff, and often at a price that makes the Amazon versions look like a gift.

The evidence is more interesting than the sales pitch. It says cold after knee surgery helps a little, that the machine does not help more than ice, and that the one thing the machine is reliably better at, being less cold at the skin, is also why people leave it on long enough to burn themselves.

What cold therapy does after knee surgery: the numbers

The 2023 Cochrane review by Aggarwal and colleagues pooled 22 trials and 1,839 people after total knee replacement, any form of cold (ice, cooled-water devices, with or without compression) against any control:

No benefit on painkiller use, length of stay or quality of life. The authors’ own summary: the benefits “may be too small to justify its use.” The adverse events recorded, at a rate no different from controls, were skin reactions, cold injuries, superficial infections and clots.

The guidelines read the same literature and split. The American Academy of Orthopaedic Surgeons’ 2015 knee arthroplasty guideline said moderate evidence supports that cryotherapy devices do not improve outcomes, and listed frostbite as the harm; its 2022 update dropped the topic altogether, as did the 2022 ACL guideline. The American Physical Therapy Association’s 2020 knee replacement guideline says therapists should teach and encourage cryotherapy for early pain, and its own three-trial analysis found no pain difference between devices and standard cold packs.

Machine vs ice pack: the head-to-head trials

This is the question the listings answer with “medical grade” and the trials answer with “about the same.”

• After knee replacement. Schinsky’s 2016 trial randomised 100 people to a circulating cold-water device or an ice-and-gel wrap for six weeks and found the wrap non-inferior on every outcome, with lower pain at six weeks and a lower price. Thienpont’s 2014 trial of 116 people found no difference in pain, painkillers, swelling or blood loss between a computer-controlled cooling device and cold packs, and worse flexion at six weeks in the device group, which the author put down to the knee being held straight in the cuff. A 2025 three-arm trial of 108 people found a gel pack, an Aircast Cryo/Cuff and a cold-compression device equal on morphine use and flexion, with the Cryo/Cuff best on swelling at 72 hours. The one trial that found a device advantage, Su’s 2012 multicentre study of 280 people, measured 509 against 680 mg of morphine-equivalent opioid use over two weeks with a cryopneumatic device against ice and compression, and no difference after that or in range of motion. And Ivey’s 1994 trial, the one the AAOS rated highest, ran circulating pads at 50, 60 and 70°F for 72 hours after replacement and found the temperature made no difference to morphine use.
• After ACL reconstruction. Barber’s 1998 trial of 100 people found continuous-flow cold for three days gave lower pain scores and less hydrocodone than no cold, with flexion at one week 88° against 79°; his 2000 follow-up against crushed ice found the same flexion gap (88° vs 77°) and more painkiller use with ice on the first two days.

The honest reading: cold helps a little after knee surgery; the machine is not better than ice at reducing pain or swelling; it may be better at letting the knee move early after a ligament reconstruction; and it is a great deal more convenient. If your surgeon’s office is charging several hundred dollars for the clinical brand, the Amazon units do the same job for a fraction. If you were going to use ice packs, the trials say that is fine too.

How cold it actually gets, and why that is the whole safety story

The listings sell “medical grade cold” and never state a temperature. The measurements that exist are better than the listings.

Warren and colleagues, in the American Journal of Sports Medicine in 2004, put a probe above the kneecap in 12 normal knees and compared crushed ice with a circulating cold device: the temperature drop was 3.3 against 2.2°C at 30 minutes, 12.8 against 7.1°C at 60, and 15.2 against 9.7°C at 90. Ice cooled the knee more, and hurt more, at every point. That is the machine’s whole trade-off in one table: gentler, shallower cold that you can tolerate for longer.

Belsey and colleagues, in PLoS One in 2024, measured skin temperature 20 mm below the kneecap in 32 adults under five systems for 30 minutes, from a 31.4°C start: a gel wrap reached 19.9°C, the gravity-fed Aircast Cryo/Cuff 17.0°C, a Breg powered unit 17.2°C, and two powered chillers set to 1°C and 8°C reached 13.0 and 13.9°C. The set temperature on the dial was not the temperature on the skin, only the powered chillers reached the 10 to 15°C band the authors called therapeutic, and below 10°C the cold-injury risk rises.

What nobody has measured is the Amazon unit: a cooler of ice water at roughly 0 to 4°C, pumped through an uninsulated pad. The water in the pad is colder than any of the systems above at the start and warms as the ice melts, and no published study gives either number or the curve between them. That is why the barrier layer and the timer in the next section are not optional.

The burns: what the case reports say

• A 53-year-old man after patellar tendon repair on both knees used an ice-and-water cooler pump pad continuously for two weeks, over a film dressing and a paper-towel interface, and had full-thickness dead skin over both knees at four weeks (Orthopedics, 2007).
• Two knee replacement patients lost the entire skin over the kneecap to cold therapy and needed debridement, a dermal matrix and skin grafts (Journal of Arthroplasty, 2013).
• Four frostbite cases under the kneecap pad of a cold cuff after ACL reconstruction (Arthroscopy, 2006).
• A compartment syndrome from an electronic cooling pad (American Journal of Sports Medicine, 2004), and perniosis, a cold-induced inflammatory rash, under a cold-therapy cuff in two patients (Archives of Dermatology, 2012).

A 2021 review lists the complications as skin irritation, frostbite, perniosis and peripheral nerve injury, and says they are reduced by patient education and shorter applications. Breg’s own instructions for its clinical unit say the pad must never touch bare skin, to stop if the barrier gets wet, to wait at least 30 minutes between uses, to check the skin every one to two hours as the prescriber directs, and that the unit can be cold enough to cause full skin necrosis. That is the manufacturer of the most-prescribed machine telling you what it can do.

What to look for in a cold therapy machine

1. A real timer with on and off cycles. The one feature that addresses the case reports. Several units here cycle on for a set period and off for a set period on their own; the Polar unit puts the control inline on the hose, which matters when you cannot reach the floor.
2. Reservoir size and what it takes. Nine to 20 quarts here. Bigger runs longer between refills; Polar’s unit takes six frozen water bottles instead of loose ice, which is cleaner and reusable. The listings’ “up to 18 hours” claims are for the ice lasting, not for the skin.
3. The pad. A knee-specific pad with three straps wraps a swollen knee better than a universal rectangle; a universal pad fits a shoulder, hip or ankle that a knee pad will not. Decide which joint it is for. Pads with an insulated cover are safer than bare ones.
4. Pump noise and flow control. The quieter listings claim under 18 dB with no method; the useful spec is adjustable flow, because less flow means less cold at the pad.
5. Hose length and insulation. Five feet on the units that state it; insulated tubing keeps the water cold between cooler and pad and keeps condensation off the bed.
6. Who sells it. Several units here are sold by third-party stores through Amazon rather than by a brand with a parts channel. A pad that leaks on day three is the usual failure; buy from somewhere that will replace it.

What does not matter: “medical grade”, “cryotherapy system”, “FSA eligible” (a claim about the card, not the IRS; see our HSA and FSA page), and the temperature display on units that have one, which reads the reservoir, not your skin.

The best ice machines for knee surgery: our picks

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

The one the clinics’ brand sells on Amazon

The cheaper units that do the same job

The bigger unit for a shoulder, hip, or two knees

The ICYTHRP unit (12-quart collapsible bucket, programmable on and off cycles) is the same class as the GALINAND and ONLYCARE and read “only 9 left in stock” when we checked, so we have not linked it. The Aircast Cryo/Cuff, the gravity-fed cuff in the temperature study above, is sold on Amazon as the cuff alone, without the cooler, by a third party; the full system is a surgeon’s-office item.

How to use an ice machine after knee surgery

Your surgeon’s protocol first. Where it is silent, this is what the manufacturers and the case reports agree on:

1. Fill with ice and water to the marked line, or the frozen bottles the unit specifies. Ice alone does not flow; water alone does not cool.
2. Barrier on. The pad’s cover, a dry towel, or a dry dressing between pad and skin. If it gets wet with condensation, change it. Never pad to skin.
3. Wrap snug, not tight. The pad should stay put when you shift, not leave a mark. Compression from a pad is not the measured compression of the cuff trials.
4. Set the cycle short. Twenty to thirty minutes on, then at least thirty off, through the day. The 180-minute settings exist; the burns were from continuous use.
5. Look at the skin at every break, and have someone else look if you cannot see the back of the knee: white, waxy, blistered, mottled or persistently red skin means stop.
6. Overnight only on a cycled timer your surgeon has approved, and never with a pad over a numb area you will not check for hours.
7. Elevate while you ice, which is in the physical therapy guideline and costs nothing.
8. Stop when the knee stops swelling. The Cochrane trials ran cold for days, not weeks. Cold after that is comfort, and the heating pad takes over for stiffness once the swelling is gone.

Ice machine vs the rest of the shelf

• A bag of crushed ice in a pillowcase is colder, free, and equal in the trials. The machine wins on steadiness and on not getting up.
• An ice bath is whole-body cold for a different reason; our cold plunge benefits page covers what immersion does and does not do, and none of it is post-surgical.
• Compression boots are the compression without the cold, and the cuff trials suggest the compression may be doing some of the swelling work.
• A knee brace is what the knee wears after the ice, when it is walking again.
• A heating pad is for the stiff knee weeks later, not the swollen one now.

What we have not measured

Every spec here is listing-stated as of 26 October 2026. We have not filled, run, or worn any of these machines.

The measurement the whole category owes is the simplest on this site. A thermocouple inside the pad’s water channel and a second one on 1-inch foam under the pad’s own cover, logged every 30 seconds from a full ice load until the reservoir reaches 15°C, at each flow setting, for each unit. That gives the number no listing or study provides: how cold the water reaching your knee is, and how fast it warms, on a $120 machine against the clinical systems whose skin temperatures Belsey measured. Alongside: a sound meter at one metre for the “under 18 dB” pumps, a stopwatch on each timer’s stated cycle, a scale for how much ice a fill actually takes, and a leak test of each pad under a 20 lb weight for 24 hours, because the pad is what fails. A day per unit, n=1, on foam rather than a knee, which carries heat away in ways foam does not.

What we would buy

For a knee replacement or ACL reconstruction and a surgeon who said “ice it”: the Polar Active Ice 3.0, for the inline cycling timer and a maker with a clinical line.

If the budget is tighter and you will set a stopwatch anyway: the PhysioNatural, bought for the pump and not its listing.

If you want the cycle to run itself while you are awake: the ONLYCARE, with its 18 timers and remote.

For a shoulder, a hip, or two knees: the CRYOFORCE CF-3 Pro, for the reservoir and the insulated cover.

Whichever one: a dry barrier, a short cycle, and eyes on the skin. Our how we test page says what every figure here is and is not.