Bionic limbs with nerve and bone connection

Also known as Osseointegrated prostheses, Bone-anchored prostheses, Targeted muscle reinnervation, Agonist-antagonist myoneural interface

Artificial arms and legs are being attached directly to bone and linked to the wearer's nerves and muscles, so they can be controlled more naturally and, in early studies, give some sense of touch. A bone-anchored system is approved in the US for some above-knee amputees, while nerve and muscle interfaces are still tested in small studies.

Pilot and early commercialDemonstratedReviewed 21 Oct 2026
AI illustration · Illustrative scene, not a photograph of the specific device or experiment.

In plain language

Most artificial limbs are held on by a socket that presses against the stump, and they are controlled by simple switches or by muscle signals read from the skin. Two newer approaches try to make the limb work more like part of the body. In osseointegration, an implant is fixed into the bone and comes out through the skin, so the artificial limb attaches directly to the skeleton instead of to a socket.

Other approaches work on the nerves and muscles. In targeted muscle reinnervation, surgeons move the nerves that used to supply the missing limb into muscles that are still there, so that when the wearer thinks about moving the missing hand, the muscle contracts and sensors on the skin can read the signal. Newer surgery reconnects pairs of muscles so that they still work against each other, and the 2024 study reports that this increased nerve signals from the remaining muscles. Most of these studies involve a handful of people, so they show what is possible rather than how common these devices will become.

Technical explanation

Osseointegration anchors a prosthesis to bone through a percutaneous implant, which is an implant that passes through the skin. The OPRA Implant System, submitted by Integrum AB, received US Food and Drug Administration premarket approval (PMA P190009) with a decision date of 18 December 2020, for skeletally mature patients with transfemoral (above-knee) amputation from trauma or cancer who have, or are anticipated to have, rehabilitation problems with, or cannot use, a conventional socket prosthesis. In the 51-patient prospective cohort of osseointegrated percutaneous prostheses reported by Brånemark and colleagues (2014), cumulative survival was 92% at two years, but superficial infection was the most common complication, with 41 episodes in 28 patients, and four implants were removed for loosening.

Neural interfaces use electrodes on or in nerves and muscles. Targeted muscle reinnervation was reported in a 2007 Lancet case study by Kuiken and colleagues, in which the nerves of a missing arm were transferred to chest muscles, and sensory nerves were rerouted to skin on the chest. Ortiz-Catalan and colleagues (2014) reported a bone-anchored interface in one subject followed for more than a year, in which implanted electrodes gave more precise control than surface electrodes and neurostimulation evoked appropriate sensory feedback. Raspopovic and colleagues (2014) stimulated median and ulnar nerve fascicles through intrafascicular electrodes, giving an amputee sensory information while the device decoded grasp in real time. Song and colleagues (2024) reported that their agonist-antagonist myoneural interface raised residual muscle afferent signalling to 18% of biologically intact values in seven leg amputees. Their maximum neuroprosthetic walking speed was 41% higher than in a matched group of seven amputees without the surgery (1.78 versus 1.26 m/s).

These results come from small studies that are mostly non-randomised and unblinded. The 2024 study's authors state that clinical benefits of long-term use still need to be demonstrated.

What the evidence says

Each statement is typed by the strength of its evidence and linked to the source that supports it.

Demonstrated Shown in a published experiment, test or shipping product.

The US Food and Drug Administration approved the OPRA Implant System (PMA P190009) with a decision date of 18 December 2020, for skeletally mature patients with transfemoral amputation from trauma or cancer who have, or are anticipated to have, rehabilitation problems with, or cannot use, a conventional socket prosthesis.

Sources: [1] · Approval record: decision date and indication for use

In a prospective, single-centre, non-randomised study of 51 patients with 55 transfemoral amputations, enrolled between 1999 and 2007, cumulative survival was 92% at two years.

Sources: [2] · Abstract · medium confidence

In the same 51-patient cohort, superficial infection was the most frequent complication, with 41 episodes in 28 patients, and four implants were removed for loosening.

Sources: [2] · Abstract

In one subject followed for more than a year, implanted electrodes in a bone-anchored interface gave more precise and reliable prosthesis control than surface electrodes, and neurostimulation evoked appropriate sensory feedback.

Sources: [4] · Abstract · medium confidence

Stimulating median and ulnar nerve fascicles through intrafascicular electrodes gave an amputee near-natural sensory information during real-time grasp control, and the participant used three grasp force levels consistently without visual or auditory feedback.

Sources: [5] · Abstract · medium confidence

In seven leg amputees, an agonist-antagonist myoneural interface raised residual muscle afferent signalling to 18% of biologically intact values, and maximum neuroprosthetic walking speed was 41% higher than in a matched group of seven amputees without the augmentation (1.78 versus 1.26 m/s).

Sources: [6] · Abstract and Results · medium confidence

Reported, unverified Stated by a company or team but not independently confirmed.

Kuiken and colleagues reported in 2007 that transferring nerves from a missing arm into chest muscles let a woman with a proximal arm amputation control a prosthetic arm using her intention to move the missing hand or elbow.

Sources: [3] · Abstract summary retrieved via Europe PMC; full text not checked · medium confidence

MIT reports that about 60 patients around the world have received the agonist-antagonist myoneural interface surgery.

Sources: [7] · Press release text: 'So far, about 60 patients around the world have received this type of surgery' · low confidence

MIT reports that the sensory feedback provided by the agonist-antagonist interface is less than 20 percent of what would normally be received by a person without an amputation.

Sources: [7] · Press release text · medium confidence

Limitations and open problems

  • Most evidence comes from small studies: 51 patients in the osseointegrated cohort reported by Brånemark and colleagues (2014), seven amputees against seven in the 2024 agonist-antagonist walking study, and single-subject or single-case reports for several nerve interface studies (Ortiz-Catalan and colleagues, 2014; Kuiken and colleagues, 2007).
  • Bone-anchored implants carry a risk of infection where the implant passes through the skin. In the 51-patient cohort, superficial infection was the most frequent complication, with 41 episodes in 28 patients (Brånemark and colleagues, 2014).
  • Sensory feedback is only partial. The MIT release reports that the agonist-antagonist interface provided less than 20 percent of the feedback a person without an amputation normally receives.
  • The 2024 study was prospective and non-randomised, and participants and the research team were not blinded to testing conditions. Its authors state that clinical benefits of long-term use still need to be demonstrated (Song and colleagues, 2024).
  • In the 2024 study, the battery module ran the bionic system for about two hours on a single charge, and the ankle-foot unit weighed 2.42 kg without the battery (Song and colleagues, 2024).
  • The sources reviewed here do not report verified long-term outcomes or the cost of these procedures.

Possible future uses

Speculative

Ideas, not predictions. None of these has been demonstrated.

  • Nerve and muscle interfaces for arm amputations. MIT reports that the surgery can also be done for people with arm amputations, but the 2024 study tested leg amputees only.
  • Combining bone-anchored sensory feedback with brain-computer interfaces, so that a wearer could sense grip force while using a prosthetic hand in daily life.
  • Everyday leg prostheses that adjust to slopes, stairs and obstacles from the wearer's own nerve signals, if long-term studies confirm the 2024 results.

Timeline

  1. 2007Kuiken and colleagues publish a Lancet case study of targeted reinnervation in a woman with a proximal arm amputation, using nerve transfers to control a prosthetic arm. Sources: [3]
  2. Jan 2014Brånemark and colleagues publish a prospective study of 51 patients with transfemoral osseointegrated implants, reporting 92% cumulative survival at two years. Sources: [2]
  3. Feb 2014Raspopovic and colleagues report near-natural sensory feedback through nerve stimulation in a hand prosthesis for an amputee. Sources: [5]
  4. Oct 2014Ortiz-Catalan and colleagues report a bone-anchored interface with implanted electrodes, followed for more than a year in one subject. Sources: [4]
  5. 18 Dec 2020The FDA approves the OPRA Implant System (PMA P190009) for transfemoral amputees who cannot use, or have trouble with, a conventional socket prosthesis. Sources: [1]
  6. 1 Jul 2024Song and colleagues publish in Nature Medicine a study of an agonist-antagonist myoneural interface in seven leg amputees, with a 41% higher maximum walking speed than a matched group. Sources: [6], [7]

Futures this could help bring about

AI illustration
How far away?

Artificial limbs that move and feel like real ones

Pilots · 7 of 10 steps done

Glossary

Osseointegration
The direct structural connection between living bone and the surface of an implant, which lets an implant be anchored in bone.
Targeted muscle reinnervation
A surgery that moves nerves from a missing limb into nearby muscles, so that the muscles give signals that a prosthesis can read.
Agonist-antagonist myoneural interface
A surgical method that reconnects pairs of muscles so they still work against each other inside the residual limb, which increases nerve signals from the remaining muscles.
Afferent signal
A nerve signal that travels from the body toward the brain, such as one carrying information about muscle length or touch.

Sources

  1. government report
    PMA P190009: OPRA Implant System
    US Food and Drug Administration, 18 Dec 2020. Accessed 21 Oct 2026.
  2. peer reviewed paper
    A novel osseointegrated percutaneous prosthetic system for the treatment of patients with transfemoral amputation: A prospective study of 51 patients
    R. Brånemark, O. Berlin, K. Hagberg. The Bone & Joint Journal, Jan 2014. DOI: 10.1302/0301-620X.96B1.31905. Accessed 21 Oct 2026.
  3. peer reviewed paper
    Targeted reinnervation for enhanced prosthetic arm function in a woman with a proximal amputation: a case study
    T. A. Kuiken, L. A. Miller, R. D. Lipschutz. The Lancet, 2007. DOI: 10.1016/s0140-6736(07)60193-7. Accessed 21 Oct 2026.
  4. peer reviewed paper
    An osseointegrated human-machine gateway for long-term sensory feedback and motor control of artificial limbs
    M. Ortiz-Catalan, B. Håkansson, R. Brånemark. Science Translational Medicine, Oct 2014. DOI: 10.1126/scitranslmed.3008933. Accessed 21 Oct 2026.
  5. peer reviewed paper
    Restoring natural sensory feedback in real-time bidirectional hand prostheses
    S. Raspopovic, M. Capogrosso, F. M. Petrini. Science Translational Medicine, Feb 2014. DOI: 10.1126/scitranslmed.3006820. Accessed 21 Oct 2026.
  6. peer reviewed paper
    Continuous neural control of a bionic limb restores biomimetic gait after amputation
    Hyungeun Song, Hugh M. Herr. Nature Medicine, 1 Jul 2024. DOI: 10.1038/s41591-024-02994-9. Accessed 21 Oct 2026.
  7. university press release
    A prosthesis driven by the nervous system helps people with amputation walk naturally
    MIT News, 1 Jul 2024. Accessed 21 Oct 2026.

Revision history

3 changes since this profile was created.

Changes on 21 Oct 2026
Published on owner instruction (content loop iteration 47)21eac85
Sources and claims re-checked against originalsfbb434c
Initial profile drafted from primary sources7e66249