The kinematic joint programme · September 2026

Eleven joints,
one error

Measure how the joint actually behaves under load across its arc, then let that measurement drive the plan, the geometry and the device — instead of referencing static anatomy or a fixed alignment target.

Written by an MD with deep clinical training. Not a practising surgeon and not a licensed physician — which is a real limitation on everything below, and also the reason the observation was available at all. Nobody who spends a career inside one subspecialty is standing where three of them can be seen failing the same way.

11 instruments built 1 hypothesis refuted 0 used to plan a real operation
§ 1

Three subspecialties found the same error and none of them cited the others

That is the whole claim. The instruments exist to test it.

Each of these fields had a universal anatomical target. Each discovered the target was wrong — not badly measured, but wrong in kind, because it imposed a population average on an individual. Each replaced it with a patient-specific reference. And they did it separately, in their own journals, without a shared vocabulary.

FieldUniversal targetHow it failedReplaced by
KneeNeutral mechanical axisImposed an average on constitutionally non-neutral kneesConstitutional alignment
HipLewinnek safe zone58% of dislocated hips were inside itSpinopelvic planning
SpineNeutral sagittal alignmentAchieving it overcorrects the elderly → PJKAge-adjusted targets
ShoulderGeneric version targetIgnores distance already travelledPremorbid glenoid
AnkleTibiotalar angle aloneIgnores whether the joint below shares itHindfoot behaviour

The contribution here is not any individual fix — every one of those was made by the people who own that field. It is naming the shared error, and then asking what transfers. Three things do, immediately: the hip's two-position protocol is the ankle's Coleman block asking the identical question in a different vocabulary; the spine's age-indexing exists nowhere else and probably should; and the hip's finding that a good target lets you decline the bailout — dual mobility falling from 100% to 37% of cases — is a general result currently stated only in the hip.

The method, in one line

Find the envelope. Fix the reference. Correct by position first, and escalate to geometry only on the residual.

§ 2

The eleven instruments

Same six questions asked of every joint, so that they read against each other rather than as eleven separate essays. Each is a working tool, not a diagram.

The colour on the left edge is not decoration — it is the family the joint turned out to belong to, which is a result that emerged from building them and is explained in § 3.

01KneeOwn reference

Compartment opening under load, measured across the flexion arc, driving component geometry rather than only component position. The origin instance: functional-alignment-first, with the soft-tissue envelope read per structure and a pre-op clinic rating that predicts the tier before anyone opens anything.

Valgus and varus fail differently and the geometry should answer differently — laxity appearing after the bone cuts is not the same problem as posterior osteophytes producing a fixed flexion contracture.

Honest limit. Conformity slopes, step gain, patellar-track gains and the rating-to-envelope lookup are hypotheses with no clinical validation. Stiffness is read from the tighter compartment, never an average of the two — averaging hid a tight medial beside a lax lateral until that was corrected.
Signature
The envelope
Opening under load across the arc — the measurement the implant is cut to, not a target it is aimed at.
02HipCoupled

Spinopelvic mobility: functional pelvic tilt read supine, standing and seated. The instrument returns not a target angle but the width of the window in which a cup is safe for this particular pelvis.

The finding that matters clinically: immobility narrows the window, deformity does not. A stiff spine — not a deformed one — is what removes your margin. A 1A pelvis leaves 14.5° to work in; a 2B leaves 2.5°.

Design constraint. Built deliberately on the spinopelvic envelope rather than on combined anteversion, to stay clear of active third-party art. Cup window widths are the instrument's own arithmetic, not a published table.
Signature
14.5° → 2.5°
Width of the feasible cup window, 1A pelvis versus 2B. The number is the room you have, not the angle you want.
03Shoulder · glenoidAttrition

How far the glenoid has travelled from its premorbid position, set against how far a reamer may safely go. What will not fit inside that budget is the residual, and the residual is what the construct has to supply — augment, graft, or a different operation.

The soft-tissue gate sits above the geometry, not beside it: if the envelope will not tolerate the correction, the arithmetic never runs.

No mandate reaches here. Neither TEAM nor ASM includes the shoulder. The page says so rather than implying a payment clock it does not have.
Signature
Correction
budget
Degrees needed against degrees safely reamable. The gap is what the implant must carry.
04AnkleCoupled

The hip one segment down. The subtalar joint is the adjacent segment and the Coleman block is the seated film — the same two-position question, asked in a vocabulary that has no idea the hip is asking it too.

The result worth the build: the same deformity burden and the same residual can demand a different operation, depending entirely on what the hindfoot is willing to give back.

A modelling error, corrected mid-build and left on the record. Flexibility was first credited as free self-correction. That is wrong. A flexible hindfoot is correctable, and correcting it is the osteotomy. Flexibility decides osteotomy-versus-fusion; it does not shrink the burden.
Signature
What the hindfoot gives back
Compensation available below the ankle, tested past a 5° physiologic allowance.
05Foot · halluxOwn reference

First MTP dorsiflexion available under load against roughly the 55° a gait cycle demands. Two nested two-position tests: Jack's test below for the windlass, Silfverskiöld above — because the calf loads the forefoot and the toe is downstream of it. Mid-range pain overrides the arithmetic entirely.

This is the weakest instrument in the programme and its page says so first, above the tool. Weight-bearing passive hallux dorsiflexion in standing has been reported not related to hallux dorsiflexion during walking. The second position this instrument rests on is contested in a way the hip's seated film and the ankle's block are not.
Signature
Arc available
vs 55° demanded
45–60° walking, ~65° running. Supply against demand, with pain as an overriding gate.
06Hand · thumbCoupled

The thumb CMC joint, and what the MCP is doing about it. As the base collapses into adduction the MCP hyperextends to keep the web open. Rebuild the base alone and the reconstruction inherits that compensation. The gate is clinical — failed non-operative management — not radiographic.

Structural, not kinematic, and labelled as such: there is no second position here, the hyperextension is measured once. The 30° decision point is published. The 20° watch level is this tool's own convention and is marked as invented. What to do past 30° is genuinely unsettled.
Signature
MCP
compensation
Hyperextension the adjacent joint has taken on. 30° published; 20° is ours.
07Lumbar spineOwn reference

The same spinopelvic films as the hip, read in the opposite direction. The hip asks will this pelvis move? The spine asks how much of that motion am I about to remove? A stiff spine is the hip's hazard and this operation's product.

Two patients, identical radiographs — PI−LL 43°, SVA 115 mm, PT 34°, all modifiers ++. At 78 the age-adjusted target demands 26.3° of correction. At 34 it demands 53.5°. A universal neutral target demands 43° of both: it undershoots the young man by 10.5° and overshoots the elderly woman by 16.7°.

A 27° swing on date of birth alone — and the asymmetry is the point. One ideal number is wrong in both directions but only dangerous in one, where junctional failure already runs near 50% and where the patients who failed had been overcorrected.

The loudest safety framing in the programme, deliberately. The age-adjusted values used are approximate brackets showing the shape, not the primary regression — take that from the source before it informs an operation. Reported correlations with junctional angle were small (r ≈ 0.32): direction supported, precision not. It models a principle, not a plan.
Signature
26.3° vs 53.5°
Correction demanded of identical films at 78 and at 34. A 27° swing on age alone.
08Elbow · forearmCoupled — confirms hardest

The forearm is a ring: proximal joint, interosseous membrane, distal joint. The radial head is a secondary valgus stabiliser — expendable only while the coronoid and MCL are intact and the membrane holds.

Excise it with an incompetent membrane and the radius migrates proximally. The failure presents as ulnar-sided wrist pain, months later, at a joint nobody operated on, often to a different surgeon. Adjacent-segment coupling at maximum displacement — a whole limb segment away, and displaced in time as well as space.

The instrument reports the disagreement rather than resolving it. The radius pull test threshold is >3 mm conventionally and >2 mm by the RAIL criterion. It gives you the gap between them instead of quietly picking one.
Signature
>3 mm / >2 mm
The two published pull-test thresholds, reported as a gap. Expendability is conditional, never a property of the part.
09Wrist · carpusAttrition — refutes

Built expecting an envelope. Found a gate. No neighbour compensates, there is no second position, and the midcarpal joint is resected rather than recruited.

What actually decides the operation is which articular surfaces have survived: a proximal row carpectomy makes the capitate articulate with the lunate fossa, so both must be intact, and stage III degeneration of the proximal capitate excludes the operation rather than making it harder.

The instrument is a classifier with a gate, not a calculator — and that is the finding. This joint refuted the programme's own rule, which is why it is here at full length rather than quietly dropped.
Signature
What survives
Capitate head and lunate fossa, intact or not. A gate, not a budget.
10Shoulder girdleCoupled

The scapulothoracic share of elevation — judged against the share expected for that segment of the arc, not against one taught ratio.

Scapulohumeral rhythm is taught as 2:1. No increment is 2:1. The scapula supplies about 2.5% of the first 30° and about 52.7% of the arc from 90° to 120°. In the cited inclinometer series the ratio between 30° and 90° ran from 1.64:1 to 3.76:1 and never once equalled 2:1.

That is § 1's error in miniature, arrived at independently: a whole-arc population average taught as a per-patient, per-segment target.

Weak signals stay weak. Visible dyskinesis reaches tier B and never tier C on its own — it is reported in 68 to 100% of injured shoulders, so by itself it discriminates almost nothing. Beyond 120° the expected share is extrapolated and flagged as such.
Signature
2.5% → 52.7%
Scapular share of the first 30° versus the 90–120° segment. There is no such thing as 2:1.
11Cervical spineCoupled — displaced failure

T1 slope behaves as the cervical pelvic incidence — the fixed reference the lordosis must be built to match, correlating at about r = 0.89 with C0–C7 lordosis. The instrument reads the T1S−CL mismatch and the C2–C7 sagittal axis against a 40 mm threshold.

This is the only joint in the programme where the adjacent segment is an output of the operation with a published rate, rather than an input you measure beforehand. Symptomatic adjacent-segment disease runs at 2.9% per year and 25.6% at ten years. The model reproduces that ten-year figure to within a tenth of a point — the only external calibration any instrument here has.

Two patients with identical radiographs, three levels: at 45 the projection is about 76%, at 75 about 24%. The target does not move. The exposure does. Age therefore enters the spine twice for entirely unrelated reasons — changing the target in the lumbar spine, and the exposure here.

Two warnings ride on the page and in the output, and the tool will not state the number without them. Past ten years it is extrapolation. And whether fusion causes adjacent-segment disease or merely reveals natural history is genuinely contested.
Signature
76% vs 24%
Projected adjacent-segment exposure from identical films at 45 and at 75. The neighbour you are about to manufacture.
12The instrument that was never builtNot built

Every one of the eleven above depends on a joint being loaded — under body weight, under an examiner's hand, under a block. Nothing in orthopaedics standardises that force. "Stress the joint" means whatever the examiner's grip means that afternoon.

A calibrated stress applicator with a load indicator would put a number under all eleven instruments at once. It is the highest-value missing piece in the programme and it does not exist here — only a note that it should.

Listed because it is absent. A programme that only shows what it built is advertising, not a record.
Status
The measurement every other instrument silently assumes somebody made consistently.
§ 3

The programme tested itself and lost half its claim

Which is worth considerably more than a confirmation would have been.

The written thesis listed six conditions under which it should be considered false. The fourth was: examine joints with obvious neighbours — the elbow, the wrist — and see whether the neighbour dominates. Both were then built, without softening the prediction first. The result was mixed.

Condition 4 · elbow

Confirmed, harder than any joint that came before it

The ring holds. The neighbour does not merely participate — it determines whether a structure is expendable at all, and it registers the failure at a different joint, months later, in front of a different surgeon.

Condition 4 · wrist

Refuted, and the page title says so

There is no neighbour to ask. Applying the adjacent-segment question to this joint produced nothing — which is exactly what happened during the build: an envelope was expected and a gate was found.

The original rule was simply too broad. What replaced it is a distinction the programme did not have when it started:

FamilyJointsCouplingThe question to ask
Coupledhip · ankle · thumb · elbow · girdle · cervicalchain, linkage, ringWhat is the neighbour doing?
Attritionshoulder · wristnone — local surface lossWhat is left to build on?
Neither cleanlyknee · foot · lumbar spineload and postureDeviation from its own reference

Then two further corrections, forced by the last two instruments

Families are properties of the question, not of the joint. The same shoulder is an attrition problem when you are asking about the glenoid surface and a coupled problem when you are asking about the girdle. The table above assigns families to joints, which is a convenience; the honest version assigns them to questions.

There is a third pattern: displaced failure. The elbow and the cervical spine share something neither the coupled nor the attrition family captures — the harm lands at a joint nobody operated on. In the cervical spine, that pattern even comes with a price list.

What was done about it

The book was rewritten rather than left standing. The chapter that stated the hypothesis now carries the result and the two-family table; the falsification chapter marks condition four tested, and partly true, and states plainly that the claim as first published was wrong.

§ 4

What is anchored and what is invented

Kept as a standing list, because a model that does not say which numbers it made up is not a model.

Anchored in published work: the Lewinnek failure rate, the hip-spine classification, Walch glenoid types, the Coleman block, Jack's and Silfverskiöld's tests, the 30° MCP decision point, SRS-Schwab modifiers and age-adjusted alignment, the radius pull test thresholds, proximal row carpectomy staging, scapular share by arc segment, and the Hilibrand adjacent-segment rate.

Invented here, and marked as such on every page that uses them: conformity slopes, step gain, cup window widths, the ankle's 12° capacity, cheilectomy gains, the 20° MCP watch level, the spine's age brackets, and the assumptions of additivity and linearity throughout.

The honest edges

  • Nothing here has planned a real operation. Eleven instruments, zero patients. Publication is not demand — the clinic rating is live and nobody has used it.
  • The convergence may be descriptive rather than causal. Three fields arriving at similar conclusions could reflect a shared cause, or simply a shared era of better imaging. That work has not been done.
  • Every constant is heuristic. The optimiser is a linearised gap model, not a robot's bone model.
  • Inventorship is not ownership. Three patents naming the author are active and assigned elsewhere; they are not this programme's lineage and are never cited as such. Two expired filings from 2008 are the clean foundation.
  • The closest third-party art predates the foundation by a month. That distinction needs counsel, not confidence.

Eleven instruments, built to test one sentence. One of them refuted the rule the others were built on, two more corrected the correction, and the twelfth — the one that would put a number under all of it — was never built.

None of them has been used to plan an operation on a living person. That is the honest state of it, and it is stated here rather than at the bottom of a disclaimer.