An evidence synthesis of the surgical treatment of trapeziometacarpal osteoarthritis — the anatomy that fails, the staging that guides less than it seems, the trapeziectomy trials that removed the case for add-ons, and three generations of joint replacement — for FESSH candidates and consultant peers.
is the commonest indication for surgery at the base of the thumb. Two operations dominate. , with or without a ligament reconstruction or suspension, is supported by randomised trials with up to 17 years of follow-up and carried a lower reoperation risk than joint replacement in a 2026 network meta-analysis12. restores length and gives a faster early recovery, and its use has grown: in France it accounted for 75.1 per cent of procedures in 20243. The randomised evidence comparing them is small and short, and the long-term evidence for modern implants is still being written4.
This article sets out the anatomy and biomechanics behind the disease, the limits of radiographic staging, the options for early disease, the trapeziectomy trials, the history and current results of arthroplasty, arthrodesis and interposition spacers, and the evidence that should inform the choice for an individual patient. It ends with points that recur in examinations.

§ 01 · WHY THE THUMB BASE FAILS

Joint loading

The trapeziometacarpal joint carries loads out of proportion to the forces at the tip. In Cooney and Chao's biomechanical model, 1 kg of pinch generated 12.0 kg of compression at the joint, and up to 120 kg may occur during strong grasp5.

Which ligament matters?

Sixteen ligaments stabilise the trapezium and the joint; in Bettinger's dissection the dorsoradial and deep anterior oblique ligaments played the main stabilising role, and the deep anterior oblique ligament was proposed as a pivot for pronation during palmar abduction6. Since then two views have competed.
The dorsal view rests on morphology, innervation and sectioning studies. In Ladd's histological study the dorsal deltoid ligament complex was the thickest, most cellular and most richly innervated, whereas the anterior oblique ligament was thin, variable in position and more consistent with capsule than with a ligament proper7. Hagert found little or no innervation of the anterior oblique ligament8. In Colman's sectioning study of 17 cadaver joints, division of the increased translation in lateral pinch more than division of the deep anterior oblique ligament, and the authors concluded that it is relatively more important for stability9. A systematic review of 52 studies concluded that increasing evidence identifies the dorsoradial ligament as the most important stabiliser10.
The palmar view rests on the pathology of the disease. Pellegrini found the essential for translational stability with flexion of the thumb ray, and degeneration of the palmar cartilage was always associated with attritional detachment of the ligament11; in Doerschuk's series every eburnated joint showed frank detachment from the metacarpal12. In a 2025 study of 71 cadaver joints, detachment of the volar beak ligament, not degeneration of the dorsal capsule, was strongly associated with osteoarthritis13.
Robotic testing suggests the two ligaments play complementary roles, each mainly restraining translation towards its own insertion14.
Controversy. Whether the dorsoradial or the anterior oblique ligament is the primary stabiliser is unresolved; morphology and sectioning favour the dorsal complex, pathology favours the beak ligament, and candidates are expected to know both arguments.

Subluxation, wear and the zigzag

Dorsal subluxation of the metacarpal marks progressive disease: in a six-year CT cohort, only the thumbs whose osteoarthritis progressed showed dorsal subluxation of the first metacarpal in the flexion pose15. The sequence that follows — resorption of the anterior beak, metacarpal flexion and dorsoradial subluxation, shortened thenar excursion and compensatory metacarpophalangeal hyperextension — produces the Z-deformity16. Flexion of the metacarpophalangeal joint unloads the most palmar surfaces of the trapeziometacarpal joint whatever the severity of the arthritis17, which is the biomechanical argument for addressing hyperextension; after implant arthroplasty, hyperextension of more than 20° corrected to a mean of 9° without any procedure on the metacarpophalangeal joint18.

§ 02 · STAGING AND ITS LIMITS

Eaton–Littler and Eaton–Glickel

The grades the radiographic changes from stage I to IV19; Eaton and Glickel's 1987 version, based on careful evaluation of the trapezial articulations, particularly in the lateral projection, adds involvement of the as stage IV2021. Dell grouped thumbs by radiographic severity and treated mild disease without surgery22, and Badia proposed an arthroscopic staging to guide treatment23.

Reliability

The classification describes better than it decides. A systematic review of four studies (163 patients) found poor to fair interobserver reliability (κ 0.11–0.56) and fair to moderate intraobserver reliability (κ 0.54–0.657)24. In other series interobserver agreement was moderate, and agreement on treatment choice only fair2526. A 2026 study of all three published versions found limited interrater reliability, worst for the intermediate stages II and III, and concluded that reliance on these classifications to guide treatment is not supported27. Shortening the scale to its first three stages improved agreement, and surgeon and patient factors biased the reading of the same radiographs21.

Stage and symptoms

Disease severity does not predictably correlate with symptoms or treatment recommendations24. In a trial cohort of 100 patients, more advanced radiographic disease was associated with lower grip strength and poorer self-reported function28. An index measured on the Robert view was more reproducible than Eaton staging (intraclass correlation 0.95 against 0.73) and correlated with surgically relevant severity29.
Clinical tests carry more weight than the stage: the thumb adduction and extension tests each detected 94 per cent of cases with few false positives, whereas the grind test missed about half3031.
Examination notes. Examine before you stage, record which test was used, and let symptoms rather than the radiographic stage decide treatment.

§ 03 · EARLY DISEASE: IS JOINT PRESERVATION WORTH IT?

Ligament reconstruction and extension osteotomy

In Eaton's 1984 series of ligament reconstruction for the unstable joint, 95 per cent of thumbs with stage I or II changes had good or excellent results, and all stage I and 82 per cent of stage II joints were free of recognisable degeneration on radiographs up to 13 years later19. Wilson described basal osteotomy of the first metacarpal in 197332; its modern form is the . In Pellegrini's cadaver work the osteotomy shifted contact from the diseased palmar compartment to the dorsal compartment, but gave no biomechanical rationale in advanced arthritis33. A systematic review of 10 studies (211 thumbs) found no gain in grip or pinch but possible relief of pain and better function; disease progressed in one third of patients, most of whom needed no further surgery34.

Arthroscopy

In a systematic review, arthroscopic techniques that included a trapeziectomy lowered pain more than debridement alone, with complications in 13 per cent35. A randomised trial of arthroscopic against open hemitrapeziectomy in 90 thumbs found no arthroscopic benefit36.

Denervation

preserves anatomy and avoids immobilisation. A systematic review of nine studies (184 procedures) found improvement in pain in all, an overall complication rate of 11.4 per cent and a lack of high-quality data37. In a prospective comparison it was not superior to trapeziectomy, and the faster rehabilitation came at the expense of a 30 per cent reoperation rate38; another prospective series converted 9 per cent to trapeziectomy with ligament reconstruction at 2 years39. Denervation patients returned to function in 3.3 weeks against 4.5 months after suspension arthroplasty, but long-term outcomes remain unknown40. In a multicentre series of 76 thumbs, 28 per cent were reoperated for recurrent complaints at a median of 9 months, and involvement of the scaphotrapeziotrapezoid joint predicted less improvement41. The 2026 Cochrane review found no randomised trial of denervation42.

Injectable biologics

In a blinded four-arm randomised trial, only the combination of fat and platelet-rich plasma reduced pain significantly more than saline43; a single injection of did no better than saline at 6 months in a placebo-controlled trial44. A systematic review of fat grafting found revision rates as high as 38.7 per cent at 5 years, from a literature with a single randomised trial45. These remain investigational.

§ 04 · TRAPEZIECTOMY: DOES THE ADD-ON MATTER?

The randomised evidence

Davis's trials set the pattern. In 76 women, tendon interposition and ligament reconstruction did not improve the results of trapeziectomy in the short term46; in a larger trial of 183 thumbs, 82 per cent achieved good pain relief and 68 per cent regained enough strength for normal daily activities at 1 year, whichever variant was used47. At 5 to 18 years, good pain relief was maintained in 78 per cent across simple trapeziectomy, palmaris longus interposition and , with no long-term benefit from the additions48. A trial of 114 thumbs found no support for ligament reconstruction with temporary K-wire fixation at a mean of 6.2 years49. Field's trial of simple trapeziectomy against flexor carpi radialis suspension found more movement after trapeziectomy alone and no difference in strength50, and at 17 years there was still no significant benefit of the reconstruction — although only 28 of 65 patients attended1.
A meta-analysis of seven randomised trials (857 participants) concluded that ligament reconstruction with tendon interposition is not superior to trapeziectomy or tendon interposition alone51; another found better grip and tip pinch with the reconstruction but significantly more complications at 3 months52. The 2026 Cochrane review rated the evidence low-certainty: compared with trapeziectomy, the reconstruction may make little or no difference to pain, function or adverse events42. The British Society for Surgery of the Hand reached the same conclusion in its guideline53.

Subsidence is radiographic, not clinical

The historical case for ligament reconstruction was preservation of height: in Burton and Pellegrini's original series, proximal migration of the metacarpal averaged 11 per cent of the arthroplasty space, against nearly 50 per cent loss of height with silicone implants54. Height, however, does not predict function. Trapezial space height did not correlate with thumb strength at 1 year55; metacarpal subsidence was not associated with key pinch, tip pinch or grip56; after the scaphoid–metacarpal distance fell with time without any correlation with strength or function57; and radiographic deterioration 9 years after ligament reconstruction did not predict an unsatisfactory result58. Degenerative-like changes at the scaphometacarpal pseudarthrosis after simple excision did not influence outcome59. The exception may be the young: in patients under 56, younger age at surgery was associated with more proximal migration and outcomes deteriorated with time60.

Suspensionplasty

Suture-button suspension and ligament reconstruction improved pain and strength equally in a randomised trial, with a shorter operation (63 against 91 minutes) and shorter immobilisation (2 against 6 weeks) after the suture button61. Complication rates were similar in a systematic review (12 and 13 per cent)62, but sensory disturbance was commoner after suture-button suspension in a series of 524 hands63, and index metacarpal fracture can present months after surgery without trauma64. In a pilot trial against suture-tape suspension, return to activity and complications favoured ligament reconstruction65.

Long-term reoperation and failure

In 1,369 patients followed for a median of 7.8 years after ligament reconstruction, 13 per cent had further surgery and 2.6 per cent revision66. In the Swedish national registry, pain on load and weakness persisted to some extent at 1 year67. A failed trapeziectomy is difficult to salvage: implant arthroplasty after trapeziectomy carries high complication and revision rates and is not supported68. A 2026 network meta-analysis of 71 studies found that trapeziectomy with or without suspension or ligament reconstruction had a lower reoperation risk, and that there is insufficient evidence to recommend routine joint replacement2.

Rehabilitation

Early mobilisation after trapeziectomy was as safe and effective as 6 weeks in a rigid orthosis in a randomised trial69, and after ligament reconstruction a longer immobilisation protocol improved neither function nor strength70.

§ 05 · ARTHROPLASTY: FROM SILICONE TO DUAL MOBILITY

Lessons from earlier implants

Each generation failed in a different way. Silicone implants failed in 25 per cent at 3.9 years, with silicone synovitis and bone resorption71. The de la Caffinière prosthesis reached 73.9 per cent survival for reoperation but 26.0 per cent for all failure at 26 years72. The metal-on-metal Elektra had complications in 23 of 37 thumbs and 17 revisions, 15 with metallosis, a failure the authors speculated was due to the metal-on-metal bearing73; in a direct comparison its survival was 68.6 per cent against 95.2 per cent for a dual-mobility design, at different follow-up lengths74. Every patient with the Moje implant presented with loosening, migration or tilting of a component75.
Single-mobility designs of the second generation lasted longer. The ARPE implant survived in 80 per cent at 15 years with further decline thereafter, and was difficult to master76; a small series reported 83 per cent at 5 years and 50 per cent at 30 years77. The Ivory implant survived in 95 per cent at 10 years, but 5.5 per cent were removed for dislocation or fracture of the trapezium78.

Dual mobility

A mobile polyethylene liner between head and cup reduced dislocation: 9.7 against 1.4 per cent comparing a single-mobility with a dual-mobility design79, and 4.2 against 0.5 per cent in a second comparison80. A systematic review of 1,421 dual-mobility arthroplasties found complications in 13 per cent and dislocation in 0.6 per cent, with follow-up too short for long-term survival and no proof of superiority over trapeziectomy81. Longer-term implant results need to be separated by design. The earlier MAÏA series reported survival of 93 per cent at 5 years and 88 per cent at 10 years, with cup loosening the main reason for revision82; these figures should not be presented as long-term dual-mobility results.83 A separate design had survival of 98 per cent at 9 years for cup removal for aseptic loosening (87 per cent for removal or exchange of any implant component).84 For a third design, estimated 2-year survival was 96 per cent in 130 patients85 and one revision was needed among 61 implants followed for a mean of 6.5 years86; in another series 4 of 88 were revised, two for cup dislocation after misplacement and two for impingement87. Correction — 19 October 2026: the 87% survival endpoint was clarified to include removal as well as exchange of an implant component.
Technique determines survival. Poorly centralised cups were associated with revision88; cup orientation should approximate the flexion–extension axis, and osteophytes between the first and second metacarpals must be removed to avoid impingement16. A learning curve of 33 to 40 cases has been described, although outcomes before proficiency were not worse89.

The comparative trials

Four randomised trials have compared joint replacement with a form of trapeziectomy; they are small, but they point the same way. 90In 62 women, MAÏA total joint replacement9091 was not superior to trapeziectomy on the Michigan Hand Outcomes Questionnaire at 1 year, but gave better strength and range of motion90; at 5 years the hand-function score still did not differ, while , satisfaction and willingness to undergo the same operation favoured replacement91. A double-mobility prosthesis was superior to the Epping procedure at 6 weeks but not at 6 months or 1 year92. The Elektra implant gave no better QuickDASH than ligament reconstruction at 2 years and more complications, with five revisions93. Against tendon interposition, a dual-mobility replacement improved faster, especially in strength and range of motion94. A meta-analysis of four trials (420 participants) found no clinically important improvement in pain at 1 year (mean difference 0.53 points on a 0–10 scale) and no information on long-term revision risk4; a network meta-analysis found similar pain relief, with pinch about 0.9 kg stronger after replacement95.
Non-randomised and registry data point the same way. Patients who chose an implant recovered faster, with stronger grip and better motion, than those who chose tendon interposition96. In the UK Hand Registry, function returned faster after replacement, but the difference was not sustained beyond 6 months, with heavy attrition97. In a pooled analysis, pinch strength returned to baseline at about 1.5 months after replacement against 6 months after trapeziectomy, though the difference was not statistically significant98.

Registries, revision and salvage

The Norwegian Arthroplasty Register reported 91 per cent survival at 5 years and 90 per cent at 10, found newer metal implants no better than older silicone and monobloc ones, and concluded that the evidence did not support widespread use — before dual mobility99. In a systematic review of 74 mostly observational studies (4,865 patients), 12.8 per cent of prostheses were revised (4.3 per cent for recent designs), against 2.0 per cent for trapeziectomy100; in the 5-year randomised trial, reoperation rates did not differ significantly91. Total joint replacement needed a salvage procedure at 2.4 per 100 procedure-years against 0.49 for trapeziectomy101.
The main reasons for revision are loosening of the trapezial component and instability102. The asymmetry matters: revision of a failed prosthesis is usually a standard trapeziectomy, whereas revision of a failed trapeziectomy is complex and uncertain100. Secondary trapeziectomy after a failed implant gave results comparable with primary trapeziectomy in a matched comparison at more than 7 years103 and in a second series104.

Pantrapezial disease

Arthritis of both the trapeziometacarpal and scaphotrapeziotrapezoid joints has long been a relative contraindication to arthroplasty. In a recent retrospective series, associated scaphotrapeziotrapezoid arthritis did not appear to impair outcomes at mid- and long-term follow-up, although it progressed radiographically in 20 per cent105; in another, thumbs with radiological pantrapezial disease that was symptomatic only at the trapeziometacarpal joint did well after arthroplasty106. The authors of one randomised trial still advise trapeziectomy with interposition when the scaphotrapeziotrapezoid joint is involved94.

§ 06 · ARTHRODESIS AND INTERPOSITION SPACERS

Arthrodesis

Arthrodesis gives a strong, stiff thumb but a high complication rate. A randomised trial against ligament reconstruction was stopped early after significantly more moderate and severe complications (71 against 29 per cent)107. In a meta-analysis, patient-rated function, tip pinch and palmar abduction were better after ligament reconstruction108, and in a series of 37 arthrodeses nine failed to unite109.

Spacers: a cautionary history

Interposition implants have repeatedly failed through foreign-body reactions. Patients with the Artelon spacer had four times the odds of being dissatisfied compared with tendon arthroplasty110, and a systematic review recommended against its use111. The PI2 interposition spacer was removed in 27 per cent at a median of 10 months112; a polyethylene-mesh spacer produced foreign-body giant-cell reactions in every case examined histologically113; and the RegJoint scaffold showed osteolysis on CT in 24 cases at 18 months114. A systematic review found the highest complication rates with polyurethane-urea and poly-L/D-lactide materials115.

§ 07 · CHOOSING FOR THE PATIENT

Return to work and cost

In Danish national registers, 91 per cent returned to work sustainably within 2 years of surgery; workers with high hand-force requirements took a median of 18 weeks, against 10 weeks for low-demand work116. In Swedish patients aged 40 to 59, sick leave was substantial after both soft-tissue and implant arthroplasty — for women, 137 days against 109 after excluding pre-operative leave117. After a dual-mobility replacement, the median return to work was 51 days118, and a cost-effectiveness study found 52 days against 84 after resection-suspension-interposition arthroplasty119.

The young and the manual worker

High occupational hand-force requirements did not have large effects on implant revision rates, in an underpowered register study120. In a claims database, patients younger than 55 and those with chronic pain were more likely to undergo revision arthroplasty121. Younger age was associated with revision after ligament reconstruction66 and with more metacarpal migration60. A 2026 review argued that indications should be set at the start, weighing stage, occupational demand, durability, revision risk and the quality of the salvage that would follow122.

The evidence still to come

No trial has compared surgery with continued non-operative care42. A three-arm randomised trial of non-surgical care, trapeziectomy and joint replacement is under way123.
Examination notes. When asked to choose, frame the answer on four axes: pain relief (equal in the trials), speed of recovery and early pinch (favour replacement), reoperation and revision (favour trapeziectomy in observational series), and salvage (secondary trapeziectomy after a failed implant performs like a primary one; the reverse is not true).

§ 08 · POINTS THAT RECUR IN EXAMINATIONS