An artificial limb is a load path and a control loop
It must carry force through soft tissue, turn intent into motion, and ideally report back. The engineering is in the device; the difficulty is at the boundary with the body.
Every term multiplies. A perfect hand that is never fitted, or fitted and then left in a drawer, contributes nothing. The record through 2026 says the middle term is 5-15% globally and the last term is roughly 55-80% for upper limbs even in wealthy systems. Neither is set by the motor count.
Shape the residual limb
The prosthesis begins in the operating room. How bone is cut, how muscles are closed over it, and what happens to the severed nerves decide how much load the limb can bear and how many control signals it can give.
- Measure
- Limb length, soft-tissue cover, neuroma pain, number of usable muscle sites
- Failure boundary
- A limb shaped only for wound closure, with painful nerve endings and muscles that no longer pull against each other
What the record shows at this step
- Reported evidence
- Surgically paired agonist-antagonist muscles preserved proprioceptive activity in the brain at levels not significantly different from people without amputation, in a 29-person imaging study.
- Where it is moving
- Amputation surgery designed as the first stage of a prosthesis: nerve rerouting and muscle pairing done at the index operation.
- Principal risk
- A limb shaped only for wound closure, with painful nerve endings and muscles that no longer pull against each other
Evidence statements describe published studies and programme reports. “Where it is moving” is an editorial reading of direction, not an announced capability.
Attach the limb
A socket must transmit body weight or a lifted load through soft tissue that swells, shrinks, sweats, and was never meant to bear it. It is the one component custom-made for every user, and the one most blamed when a limb goes unworn.
- Measure
- Hours worn per day, skin problems, refits per year, time to don
- Failure boundary
- Pressure sores, sweating, and volume change that make the limb painful by afternoon
What the record shows at this step
- Reported evidence
- In a 2020 Austrian cohort, 61% of respondents complained of comfort and 52% of weight; in Norway, comfort, function, and control led the reasons for stopping.
- Where it is moving
- Scanned and printed sockets, adjustable sockets that follow volume change, and bone anchoring that removes the socket entirely.
- Principal risk
- Pressure sores, sweating, and volume change that make the limb painful by afternoon
Evidence statements describe published studies and programme reports. “Where it is moving” is an editorial reading of direction, not an announced capability.
Read the intent
Electrodes on the skin inside the socket pick up muscle activity and turn it into motor commands. The signal moves with every shift of the socket and every drop of sweat.
- Measure
- Independent control signals, classification accuracy across a day, recalibrations needed
- Failure boundary
- A rich hand driven through a one-dimensional control channel
What the record shows at this step
- Reported evidence
- Standard two-site control gives one proportional axis. Mainstream pattern recognition classifies more motions but still selects one at a time.
- Where it is moving
- Rerouted nerves for new control sites, and implanted electrodes that do not depend on the skin or the socket.
- Principal risk
- A rich hand driven through a one-dimensional control channel
Evidence statements describe published studies and programme reports. “Where it is moving” is an editorial reading of direction, not an announced capability.
Move the joint
Motors, clutches, and springs in the hand, wrist, knee, or ankle. This is the layer that improved fastest and gets most of the attention.
- Measure
- Driven joints, grip force and speed, net joint power, weight, battery life
- Failure boundary
- Capability that adds weight at the far end of the lever and cannot be commanded anyway
What the record shows at this step
- Reported evidence
- Leading commercial hands went from one driven joint in the 1960s to six by 2013. Microprocessor knees have been sold since 1997.
- Where it is moving
- Lighter actuators, powered ankles and knees that return net positive work, and hands built for durability rather than grip count.
- Principal risk
- Capability that adds weight at the far end of the lever and cannot be commanded anyway
Evidence statements describe published studies and programme reports. “Where it is moving” is an editorial reading of direction, not an announced capability.
Return the sensation
An intact limb reports force, contact, and position continuously. Without that, every grasp is watched, and every step on uneven ground is a guess.
- Measure
- Distinguishable sensations, object-handling performance, visual attention required
- Failure boundary
- Sensation that exists in the laboratory but never ships in a clinical device
What the record shows at this step
- Reported evidence
- Nerve stimulation has produced touch felt in the missing hand since 2014 studies; four arm users of an implanted system relied on it daily for 3-7 years.
- Where it is moving
- Implanted feedback as a standard feature, and surgical constructs that restore joint-position sense in the leg.
- Principal risk
- Sensation that exists in the laboratory but never ships in a clinical device
Evidence statements describe published studies and programme reports. “Where it is moving” is an editorial reading of direction, not an announced capability.
Fit, train, and follow up
A prosthetist casts or scans the limb, builds and aligns the device, and adjusts it as the limb changes. A therapist teaches the user to use it. Both are needed again every time the socket or the device is replaced.
- Measure
- Days from amputation to first fitting, training hours, visits per year, distance to the clinic
- Failure boundary
- A device delivered without the people who make it wearable
What the record shows at this step
- Reported evidence
- Workers fitted within the first month after upper-limb amputation all returned to work in a 1984 study, against 15% of those fitted later. In the 2020 Austrian cohort, every respondent who had received no training was a non-user.
- Where it is moving
- Remote fitting reviews, task-shifting to trained technicians, and digital workflows that let one prosthetist serve more people.
- Principal risk
- A device delivered without the people who make it wearable
Evidence statements describe published studies and programme reports. “Where it is moving” is an editorial reading of direction, not an announced capability.
Pay and replace
Limbs wear out and residual limbs change. Children outgrow several a decade. A financing system that pays once and walks away produces a limb that is abandoned when it breaks.
- Measure
- Share of cost paid out of pocket, replacement interval, repair lead time
- Failure boundary
- Free first limbs followed by no second limb
What the record shows at this step
- Reported evidence
- WHO estimates only 5-15% of people who need prosthetic or orthotic services receive them.
- Where it is moving
- Public coverage that includes replacement and repair, and devices designed to be serviced locally.
- Principal risk
- Free first limbs followed by no second limb
Evidence statements describe published studies and programme reports. “Where it is moving” is an editorial reading of direction, not an announced capability.
People with limb loss say the same things decade after decade: too heavy, too hot, not comfortable. None of those complaints is about the number of motors in the hand.
How the bottleneck moved
For most of the twentieth century the limiting input was the component: a foot that survived mud and squatting, a knee that did not buckle. Those problems were solved cheaply enough for mass humanitarian use by the 1980s. What remained is custom, clinical, and biological.
Components got cheap first
The Jaipur foot (1968) and standardised polypropylene technology made a durable lower limb that costs tens of dollars. Printing now makes a socket's materials cost about $20. Materials are no longer what stands between most people and a limb.
The expensive part is a person
The socket must be shaped to one body and reshaped as it changes. That takes a trained professional, repeatedly, for life. Fitted prices in high-income clinics mostly buy clinician time, visits, and warranty.
Capability hit the control ceiling
A six-motor hand commanded through two surface electrodes behaves much like a one-motor hand with presets. Added joints add weight and cost faster than they add usable function.
The interface is biological
Sweat, swelling, pressure tolerance, and nerve pain do not follow a learning curve. Improvement there comes from surgery and materials science, which diffuse more slowly than electronics.
The same need, three orders of magnitude apart
A below-knee limb that lets a farmer walk to a field can be made for less than a hundred dollars. A fitted limb in a high-income clinic costs a hundred times more. Most of the difference is not material.
Blue: made and fitted locally in low- and middle-income settings. Red: fitted prices in high-income systems, which include clinician time, fitting sessions, training, and warranty. The reference line is the World Bank ceiling for a low-income economy, gross national income per person (FY2025): a whole year’s income for the average person there buys less than one conventional limb at the top of its range.
The access gap is a people gap
A limb costing $87 is still out of reach if the nearest person who can fit it is a two-day journey away and fitting takes several visits. Coverage follows the density of trained clinicians and the distance to them more closely than it follows device price.
Prevalence and coverage in five countries
| Country | People with limb loss | Prosthetic coverage | Main financing |
|---|---|---|---|
| United States | ~2.3 million people with limb loss | Most who need one receive a limb; access to advanced devices depends on insurer approval | Medicare, Medicaid, VA, private insurance |
| United Kingdom | Illustrative ~1-1.6 per 1,000 | Near-universal for major amputation | NHS, free at point of use |
| Brazil | Illustrative ~0.5 per 1,000 | Perhaps half or more; wide urban-rural gap | Public SUS network plus private insurance |
| India | ~0.6 per 1,000 (1983 survey) | An estimated tenth or fewer | Mostly out of pocket; NGOs such as BMVSS; limited public schemes |
| Kenya | Illustrative ~0.4 per 1,000 | Perhaps 5-10% | Mission and charity clinics, donations, out of pocket |
The United States figure is a modelled estimate; the Indian rate comes from a 1983 survey and is certainly out of date. Rows marked illustrative, and all coverage figures, are rough estimates assembled from partial reports. No country publishes a routine series of people fitted and still wearing a limb, which is itself a finding.
The discovery chain
Progress came in two streams that rarely meet: cheap durable components for the many, and surgical and electronic interfaces for the few.
- 1912
The split hook
A body-powered hook opened by a shoulder cable. More than a century later, versions of it remain among the most durable upper-limb devices in use.
- 1945
Research programmes after the war
Large numbers of veterans with limb loss prompted national research programmes that standardised sockets, alignment, and fitting as a clinical discipline.
- 1960s
Myoelectric control
Surface electrodes over the residual muscles drove an electric hand. The two-site scheme introduced then is still the default powered control.
- 1968-1979
The cheap durable limb
The Jaipur foot was designed for barefoot walking, squatting, and wet fields. From 1979 the ICRC built a rehabilitation network that later standardised polypropylene components for conflict zones.
- 1990-1997
Bone and microprocessors
The first transfemoral osseointegration removed the socket for a few patients. The first microprocessor knee adjusted resistance through each step.
- 2002-2014
Nerves as control and feedback
Targeted muscle reinnervation created new control sites from rerouted nerves. Multi-motor hands reached the market in 2007, and nerve stimulation restored touch in research participants in 2014.
- 2020-2026
Surgery becomes part of the device
Implanted arm systems were used daily for years; reconstructed muscle pairs let leg users walk 41% faster at peak than matched controls. Printing cut socket materials to tens of dollars.
What works at scale
The successful programmes share a design principle: they treat the limb as a service, not a product. Jaipur fits a limb in a day, at walk-in camps, using a foot built for the way its users live. The Sierra Leone lab moved the socket, the custom part, to the hospital where the patient already was, and kept a prosthetist and rehabilitation in the loop. The pilot was eight people; it shows feasibility, not durability.
The unit that matters is limbs worn a year later
Counting devices shipped measures a supply chain. Counting limbs still worn measures a capability restored, and it depends on the workforce more than on the factory.
How many people can a clinic workforce keep on their feet?
A prosthesis is not bought once. Sockets are refitted as the limb changes and devices wear out, so a steady population of users generates a steady demand for skilled fitting. Per million people, that demand sets the workforce a country needs.
A user waits about 12.0 extra years between fittings.
- Fittings needed a year
- 250 per million
- Fittings delivered
- 38 per million
- Professionals required
- 3.3 per million
- Of which replacements
- 60%
Compare the thin-service and WHO-minimum scenarios: the same population and the same devices, with a tenfold difference in trained people. Replacements, not new amputations, dominate demand wherever people survive long with limb loss, so a cheaper device that lasts longer relieves the workforce as well as the budget.
Calculation & assumptions
Fittings needed per million a year = users ÷ replacement interval + new amputations. Fittings delivered = professionals × devices per professional × share of time on prostheses. The wait is the replacement interval that current capacity implies for existing users, minus the intended interval.
The 5-10 professionals per million comes from the WHO standards for prosthetics and orthotics. Prevalence, incidence, throughput, and time share are illustrative round numbers, not measurements of any country; orthoses compete for the same professionals and are ignored except through the time share.
What pushes the curves next
Digital sockets
Scan, design, and print shorten fitting from several casting visits to fewer, and let a remote specialist review a local technician’s design.
Condition: printed sockets that last years in heat and humidity, shown in cohorts larger than a pilotAdjustable and bone-anchored attachment
Sockets that adjust to daily volume change attack the leading complaint. Bone anchoring removes the socket for those who can take the surgery.
Condition: infection rates and revision surgery low enough for routine useSurgery that builds control sites
Rerouted nerves and paired muscles at the first operation give every future limb more signals and less pain, at little extra cost.
Condition: surgical training that reaches general trauma and vascular surgeonsTask-shifting and financing for replacement
Trained technicians under remote supervision, and coverage that pays for the second and third limb, not only the first.
Condition: quality standards that keep up with the extra handsThe remaining constraints are clinical and biological
The field has moved from making a durable limb, to making a capable one, to the harder problem of making one that is fitted in time, worn all day, and replaced when it wears out.
The socket
Comfort is the leading reason for rejection across surveys and decades. The socket is the one custom part, the one that must be remade as the body changes, and the one no factory can mass-produce.
The workforce
WHO standards call for 5-10 prosthetists and orthotists per million people. Treating traumatic amputations alone would need about 76,000 prosthetists worldwide. Many low-income countries fall far short of that.
Timing
Fitting soon after amputation predicts use. Where people wait a year or more to be reached, compensating habits set in and first-time rejection rises.
Replacement and repair
Demand is recurring: sockets, liners, feet, and batteries wear out, and children grow. Programmes that fund first limbs but not replacements leave people with devices they cannot wear.
Control bandwidth
Standard control still commands one joint at a time. More motors do not help until the interface can supply more independent signals.
Evidence
Rejection studies are small, differently defined, and rarely repeated. Nobody publishes a longitudinal national series of limbs fitted and still worn a year later.
An optimistic view, with conditions
Restoring mobility does not require a bionic limb
The hardest part of the global problem is already cheap. A durable leg costs less than a hundred dollars. If digital fitting keeps raising what one clinician can deliver, and financing covers replacement, coverage is limited by training and organisation, which are problems with known solutions. At the frontier, surgical interfaces have shown that the socket and the control ceiling can both be removed.
The service threshold
Scanned and printed sockets validated in large cohorts, remote review routine, and national schemes that pay for replacement. Coverage in middle-income countries rising measurably.
Surgery as standard
Nerve rerouting and muscle pairing done at the first amputation as a matter of course; bone anchoring an ordinary option for suitable candidates; sensory feedback shipping on powered limbs.
Limbs that feel like limbs
Simultaneous control of hand and wrist, reliable touch and position sense, and rejection rates for powered arms below one in ten.
View the annual scorecard
| Metric | Record through September 2026 | Target for a solved problem |
|---|---|---|
| Upper-limb rejection | 18-44% in recent surveys; no improvement after 2006 in one cohort | Under 10%, sustained across cohorts |
| Access to prosthetic services | 5-15% of people in need (WHO, 2017) | Most people in need, in every income group |
| Prosthetic workforce | Far below the WHO standard in many low-income countries | 5-10 per million (WHO standard) |
| Cost of a durable lower limb | $74-87 produced locally; $5,000-15,000 fitted in the US | Tens of dollars plus a few clinician-hours |
| Simultaneous control | One joint at a time in standard care; two or more in research | Natural coordination of hand and wrist |
| Sensory feedback | Years of daily use in a handful of implanted users | A standard option on powered limbs |
Baselines are reported results and estimates, several from small cohorts. Targets are editorial benchmarks, not forecasts.
There is more than one finish line
- ComponentA durable, affordable foot, knee, or hand exists.
- FittedA trained person fits it soon after amputation.
- WornThe user still wears it all day a year later.
- MaintainedIt is repaired and replaced as the body and the device change.
- ControlledThe user commands several joints at once, without watching.
- FeltThe limb reports touch and position, and feels part of the body.
Sources, method, and boundaries
Figure 1 combines device specifications with survey results that use different definitions of rejection, different populations, and small samples; it shows a lack of improvement, not a precise rate. Figure 2 mixes production costs and fitted prices by design. Figure 4 is an illustrative model. Evidence labels follow the house standard: Measured (directly demonstrated), Derived (calculated from measured data), Projected (modeled future outcome), Target (announced goal), and Editorial inference (interpretation of evidence). Country prevalence and coverage figures are rough and flagged as such. Nothing here is clinical advice.
- Rejection
- Never taking up a prescribed prosthesis (primary) or stopping its use (secondary). Studies draw the line differently.
- Socket
- The custom-made part that holds the prosthesis to the residual limb and carries load through soft tissue.
- Driven joint
- A joint with its own actuator. Preset grip patterns are not counted.
- Osseointegration
- Attaching a prosthesis directly to bone through a titanium implant that passes through the skin.
