

Upper limb
Hand rehabilitation and splinting: refining and deepening your practice
Jean Lichtle
Updated · published
Scaphoid fracture: why the initial radiograph does not rule it out, which clinical signs count, and what too short an immobilisation ends up costing.

Faced with a painful wrist after a fall, with anatomical snuffbox pain and a normal radiograph, we treat it as a fracture until proven otherwise.
2-5%risk of scaphoid nonunion in adults
What follows covers each of these points in detail, with sources. It is there if you need it.
This topic is taught in a course: Hand rehabilitation and splinting: refining and deepening your practice, on-site with Jean Lichtle.
1 / 12
11 article chapters · 32 min in total
Understanding◔ 3 min
The scaphoid is supplied with blood against the current: a fracture line at the middle third starves the proximal pole.
Population◔ 2 min
Young men, between 20 and 29 years, account for most scaphoid fractures.
Radiograph◔ 2 min
A normal radiograph is not enough: the probability of fracture remains 12 to 57%.
Clinical signs◔ 3 min
The anatomical snuffbox is very sensitive, but not very specific: it rules out nothing on its own.
Imaging◔ 4 min
In the emergency department, MRI is the most accurate investigation for confirming or ruling out an occult fracture.
SWIFFT◔ 3 min
The SWIFFT trial finds no difference between early surgery and cast at one year.
Immobilisation◔ 3 min
The choice of cast remains debated: an old trial favoured the long cast, recent work finds no clear advantage.
Late diagnosis◔ 4 min
A missed fracture can still unite in a cast: 82% at six months' delay.
Case report◔ 1 min
A 19-year-old basketball player was treated for a sprain before the fracture was seen.
In practice◔ 4 min
A ‘sprain’ that does not heal within the expected time should trigger a return to the doctor.
Questions◔ 3 min
A normal radiograph is not enough to rule out a scaphoid fracture.
Train on this


Upper limb
Jean Lichtle
Next comes practice: the course that teaches this topic, with Jean Lichtle.
Our courses on this topic may be covered by DPC or FIFPL, the French continuing-education funding schemes, depending on your profession and situation.
It is the fracture that hides behind a wrist sprain. A young man falls on his hand, a radiograph is taken, nothing is seen, the word sprain is used and he is sent home with nothing scheduled. In a considerable proportion of these patients the fracture is there all the same: it simply cannot be seen yet. The scaphoid is supplied with blood against the current, from the bottom upwards, which makes its proximal pole dependent on a fracture line that has just cut off its supply. It is the only common fracture in which the main risk factor for nonunion is not the violence of the impact, but the time taken to recognise it.
Summary written from primary sources checked one by one on PubMed through the E-utilities interface: identifier resolved, journal, year and author list verified, abstract read before citation. Full bibliography at the end of the article.
Scaphoid fracture in three figures
Three results that explain why this fracture is missed, and why missing it is costly
Sources: Carpenter CR et al., Acad Emerg Med 2014, meta-analysis of 75 studies, prevalence of 12 to 57% and point estimate of 25% (PMID 24673666); Jørgsholm P et al., Handchir Mikrochir Plast Chir 2020, systematic review of 42 studies (PMID 32992390); Dias JJ et al., Lancet 2020, SWIFFT randomised trial, 439 patients, adjusted difference of minus 2.1 points, 95% CI minus 5.8 to 1.6 (PMID 32771106).
What to keep in mind
The reflex that prevents nonunions
Faced with a painful wrist after a fall on the hand, with anatomical snuffbox pain and a normal radiograph, we do not conclude that it is a sprain. We treat it as a fracture until proven otherwise, and we arrange the proof: cross-sectional imaging or a scheduled reassessment. A 2025 randomised trial (180 adults with normal radiographs, of whom 16 had real fractures) found no functional inferiority at three months for a three-day bandage followed by reassessment at two weeks, compared with a two-week cast; it includes too few fractures to conclude on nonunion, and the follow-up appointment is part of the strategy19. A physiotherapist who receives a ‘wrist sprain’ that does not heal, with persistent pain in the snuffbox or at the base of the thumb, should ask this question and refer the patient back to the doctor rather than rehabilitate a stiffness.
A scaphoid fracture is not missed because it is rare or deceptive. It is missed because a normal radiograph gives false reassurance, and because the word ‘sprain’ closes the file.
A bone that bridges the two rows of the carpus, and that receives its blood against the current: two anatomical particularities, two clinical consequences.
The scaphoid is the most lateral bone of the proximal row of the carpus. Its elongated, oblique shape makes it cross the midcarpal joint line: it is the only bone to articulate with both rows, which makes it the mechanical link of the carpus. Three portions are described: the distal pole (or tubercle), the waist or middle third, and the proximal pole.
This hinge position explains its exposure. In a fall on the hand in hyperextension, the stress is concentrated at the junction of the two rows, precisely where the scaphoid crosses the joint line. That is why the middle third accounts for most of the fracture lines: Jørgsholm's systematic review finds 60 to 69% of fractures at the middle third1.
The decisive feature is vascular. The main arterial supply of the scaphoid enters through its distal and dorsal portion, then runs inside the bone towards the proximal pole. This blood supply is described as retrograde: it flows against what the anatomy would lead one to expect. The proximal pole has no significant arterial entry of its own; it depends entirely on this flow coming up from below.
The consequence is mechanically inevitable. A fracture line at the middle third, and a fortiori at the proximal pole, interrupts the supply of the fragment lying upstream. The more proximal the line, the larger the devascularised fragment, and the slower and more uncertain the union becomes. This is what distinguishes this fracture from almost all others: it is not only the bone that is broken, it is also its supply.
Buijze's review of scaphoid nonunion lists the risk factors in this order: displacement and instability in first place, then delayed or missed diagnosis, an inadequate treatment, the location of the fracture line and the blood supply12. Three of these five factors depend on what the care system does, or does not do, in the first weeks.
Retrograde blood supply, and what a fracture line does to it
The more proximal the line, the larger the fragment deprived of blood supply
Schematic illustrating the classical vascular reasoning. The distribution of fracture lines (60 to 69% at the middle third) comes from Jørgsholm P et al., Handchir Mikrochir Plast Chir 2020 (PMID 32992390). The hierarchy of risk factors for nonunion follows Buijze GA, Ochtman L, Ring D, J Hand Surg Am 2012 (PMID 22541157).
Key points
The identikit is useful because it guides the pre-test probability, and therefore the imaging decision.
Duckworth's prospective Scottish study, which recorded over one year all radiologically confirmed scaphoid fractures in a defined adult population, finds 151 fractures and an annual incidence of 29 per 100,000 (95% CI 25 to 34). Male predominance is clear (105 men out of 151, p < 0.001), and the median age of men is significantly younger than that of women (p = 0.002)2.
Jørgsholm's systematic review, covering 42 studies, specifies the age distribution: in men, the peak lies between 20 and 29 years, with reported incidences of 107 to 151 per 100,000 men; in women, the peak is earlier, between 10 and 19 years, with incidences of 14 to 46 per 100,0001. In children, the fracture mainly affects boys aged 12 and over, and is rare before 9.
The gap between Duckworth's overall incidence (29 per 100,000) and Jørgsholm's figures comes down to the reference population: the first concerns all adults, the others age and sex bands in which the risk is concentrated. Both are consistent, and Jørgsholm moreover concludes that there is substantial heterogeneity between studies and a globally low level of epidemiological evidence1.
Duckworth provides a distinction that is useful for clinical reasoning. The low-energy falls, from standing height, accounted for 40.4% of mechanisms, but men were significantly more likely to fracture in high-energy trauma (p < 0.001). The most frequent Herbert classification was type B2 (55 cases, 36.4%), and above all, unstable fractures were more frequent in young patients (p = 0.025) and after high-energy trauma (p = 0.042)2.
Four figures to keep in mind
What the series establish about the population, location and outcome
Sources: Duckworth AD et al., J Trauma Acute Care Surg 2012, 151 fractures (PMID 22439232) for incidence and sex; Jørgsholm P et al., Handchir Mikrochir Plast Chir 2020 (PMID 32992390) for location; Dias JJ et al., Bone Joint J 2026, 267 patients imaged at five years (PMID 41475366) for osteoarthritis.
This is the heart of the matter, and the reason this fracture is nicknamed the wrist sprain trap.
The widespread formula that ‘one scaphoid fracture in four cannot be seen on the initial radiograph’ circulates a great deal, and it needs rephrasing to be accurate.
What Carpenter's meta-analysis establishes, from 75 included studies out of 957 citations examined, is a pre-test probability: in adult emergency department patients with clinical concern for a scaphoid injury and a non-diagnostic initial radiograph, the prevalence of fracture found on later imaging ranged from 12% to 57%, with a point estimate of 25%3. In other words, the figure describes the proportion of true fractures among the suspected patients with a normal radiograph, and not the proportion of scaphoid fractures that escape radiography.
Two other studies measure the same quantity and give lower values, which should be stated rather than retaining the most striking figure:
These gaps are not contradictions: they reflect different populations and different thresholds of suspicion. A department that radiographs widely will have a low prevalence; a department that suspects only on obvious signs will have a high prevalence. The clinical conclusion is the same in all three cases, and it is robust: a normal radiograph does not rule out the fracture, and Carpenter says so plainly in the objective of his paper.
A methodological caveat that the authors themselves make
Carpenter stresses that the majority of the included studies were conducted outside the emergency department, often in orthopaedic clinics, that none followed the reporting guidelines for diagnostic studies, and that the overall methodological quality assessed by QUADAS-2 was low, which increases the risk of bias in the accuracy estimates3. These figures guide the reasoning, they do not lock it down.
Three measures of the same quantity, three different results
Proportion of true fractures among suspected patients whose initial radiograph is normal
Sources: Coventry L et al., Emerg Med J 2023 (PMID 37169546); Mallee WH et al., Cochrane Database Syst Rev 2015 (PMID 26045406); Carpenter CR et al., Acad Emerg Med 2014 (PMID 24673666). The three studies measure the same quantity in different recruitment populations.
Many tests, little specificity, and a single sign that really helps to rule out the diagnosis.
Mallee's meta-analysis identified 13 studies describing 25 clinical tests, including only studies carried out before radiographic evaluation and having an acceptable reference standard. Two tests had enough data for statistical analysis:
The heterogeneity was such that the authors declined to calculate pooled estimates. Two other tests showed high sensitivity: pain at the scaphoid tubercle (sensitivity 0.82 to 1.00; specificity 0.17 to 0.57) and pain on ulnar deviation (0.67 to 1.00; 0.17 to 0.60). Above all, three studies showed that combining tests increased specificity and post-test probability while retaining high sensitivity4.
Coventry's systematic review, focused on patients with normal radiographs, assessed which signs best predict an occult fracture. Its results deserve careful reading, because they point to tests that are rarely performed:
| Sign | Sensitivity | Specificity | Likelihood ratio |
|---|---|---|---|
| Pain on resisted supination | 100% | 97.9% | LR+ 45.0 (95% CI 6.5 to 312.5) |
| Supination strength below 10% of the uninjured side | 84.6% | 76.9% | LR+ 3.7 (95% CI 2.2 to 6.1) |
| Pain on ulnar deviation | 55.2% | 76.4% | LR+ 2.3 (95% CI 1.8 to 3.0) |
| Pronation strength below 10% of the uninjured side | 69.2% | 64.6% | LR+ 2.0 (95% CI 1.2 to 3.2) |
| Absence of anatomical snuffbox pain | 92.1% | 48.4% | LR− 0.2 (95% CI 0.0 to 0.7) |
The best performing test, pain on resisted supination, reaches remarkable values, but its very wide confidence interval (6.5 to 312.5) betrays a limited sample. The authors say so explicitly: this sign would deserve external validation before being widely adopted. And their general conclusion is unambiguous: no single sign satisfactorily rules out an occult scaphoid fracture5.
This conclusion matches Carpenter's, worded differently: with the exception of the absence of anatomical snuffbox pain, which significantly lowers the probability of fracture with a negative likelihood ratio of 0.15, history and physical examination alone are insufficient to confirm or rule out the diagnosis, and no validated clinical decision rule exists3.
What each sign really contributes
Sensitivity and specificity of the main tests, in patients with normal radiographs
Single source: Coventry L, Oldrini I, Dean B, Novak A, Duckworth A, Metcalfe D, Which clinical features best predict occult scaphoid fractures? A systematic review of diagnostic test accuracy studies, Emerg Med J 2023;40(8):576-582 (PMID 37169546). Eight studies, 1,685 wrists. The resisted supination test rests on a limited sample and needs external validation, as the authors point out.

CT scanSagittal CT scan of the wrist: fracture line clearly visible at the carpal scaphoid.
Source : Rhemrev et al., International journal of emergency medicine, 2011, figure 3 · CC BY
Three possible investigations, three very different error profiles, and a trade-off that depends on what one fears most.
Mallee's Cochrane review compared CT, MRI and bone scintigraphy in suspected patients with normal radiographs. Eleven studies were included: 4 for CT (277 suspicions), 5 for MRI (221) and 6 for scintigraphy (543). The pooled estimates are as follows6:
| Investigation | Sensitivity | Specificity | Per 1,000 patients |
|---|---|---|---|
| Bone scintigraphy | 0.99 (95% CI 0.69 to 1.00) | 0.86 (95% CI 0.73 to 0.94) | 112 patients overtreated, 2 fractures missed |
| MRI | 0.88 (95% CI 0.64 to 0.97) | 1.00 (95% CI 0.38 to 1.00) | Not calculated in the review |
| CT | 0.72 (95% CI 0.36 to 0.92) | 0.99 (95% CI 0.71 to 1.00) | 8 patients overtreated, 56 fractures missed |
Reading this table is more subtle than it looks, and the authors guide the reading explicitly. Statistically, scintigraphy is the best investigation for establishing a definitive diagnosis: its sensitivity is the highest, and it misses the fewest fractures. But its lower specificity becomes problematic when prevalence is low: in 1,000 patients, it would lead to overtreatment of 112 people, versus 8 with CT. Conversely, CT overtreats almost no one but would let through 56 fractures where scintigraphy would miss only 26.
The authors add three important caveats: scintigraphy is more invasive, involves radiation and imposes a diagnostic delay of at least 72 hours; the confidence intervals are wide for all three investigations; and the comparisons are mostly indirect, which exposes them to confounding6.
Out of 1,000 suspected patients, the errors each investigation produces
Two opposite profiles: scintigraphy misses almost nothing but overtreats a great deal, CT does the opposite
Source of the figures: Mallee WH, Wang J, Poolman RW et al., Cochrane Database Syst Rev 2015;2015(6):CD010023 (PMID 26045406), authors' calculation on a median prevalence of 20%. The confidence intervals of the three investigations are wide, as the review stresses. The preference for MRI in the emergency setting comes from Carpenter CR et al., Acad Emerg Med 2014 (PMID 24673666).
Carpenter reaches a more clear-cut recommendation, in the specific context of the emergency department: in patients whose initial radiograph shows no fracture but in whom concern persists, MRI is the most accurate investigation, superior to scintigraphy, CT and ultrasound, both for confirming and for ruling out the diagnosis. If MRI is not available, CT is enough to confirm a fracture but remains insufficient to rule it out. The author also notes that MRI and CT share the advantage of identifying other causes of post-traumatic wrist pain3.
The apparent divergence between these two sources has an explanation. The Cochrane review reasons on pure statistical accuracy, Carpenter's meta-analysis on clinical utility in the emergency setting, taking into account availability, delay and invasiveness. For a physiotherapist, the practical conclusion is a single one: when doubt persists, cross-sectional imaging settles it, and passive waiting is not a diagnostic strategy.
An old practice found wanting
The traditional strategy of immobilising and then repeating a radiograph some time later has been specifically evaluated, and it does not hold up. A diagnostic performance study had 81 orthopaedic surgeons read the initial and six-week radiographs of 34 suspected patients, taking the CT and MRI performed on admission as reference standards. Interobserver agreement was poor (kappa of 0.15 and 0.14 depending on the viewing mode), sensitivity ranged from 42 to 79%, specificity from 53 to 59%, accuracy from 53 to 58% and positive predictive value from 14 to 26%. The authors conclude that six-week radiographs are not suitable for assessing a suspected scaphoid fracture15. The Cochrane review also notes that this type of radiographic follow-up served as the reference standard in several included studies, which weakens their estimates accordingly6.

MRICoronal MRI of the wrist: fracture of the waist of the carpal scaphoid.
Source : Rhemrev et al., International journal of emergency medicine, 2011, figure 5 · CC BY
A large pragmatic trial, a five-year follow-up, and a conclusion that reversed a surgical trend.
SWIFFT is a pragmatic, multicentre, open-label, randomised superiority trial conducted in 31 hospitals in England and Wales. It included adults aged 16 and over with a bicortical fracture of the scaphoid middle third displaced by 2 mm or less. Of 1,047 patients assessed, 439 were randomised (mean age 33 years, 363 men, i.e. 83%) between early surgical fixation (n = 219) and below-elbow cast immobilisation, followed by immediate fixation if nonunion was confirmed (n = 220)7.
The primary outcome was the PRWE score at 52 weeks. The results:
The authors' recommendation is explicit: adults with a middle-third fracture displaced by 2 mm or less should receive initial cast immobilisation, with any suspected nonunion to be confirmed and then fixed immediately. This strategy avoids the risks of surgery and reserves it for fractures that do not unite7.
The five-year follow-up, published in 2026, reinforces this result. Of 344 patients who provided a valid PRWE score (78.4% of the initial cohort), no significant difference appeared between the groups: mean difference of 0.6 points (95% CI −2.4 to 3.6), p = 0.709. Neither the pain and function subscores nor the overall estimate of the treatment effect differed. Grip strength and joint ranges of motion were comparable. The nonunion rate among imaged patients remained very low: 3 out of 146 (2.1%) in the fixation group versus 4 out of 121 (3.3%) in the immobilisation group8.
The radiological component of the same follow-up provides two pieces of information that a physiotherapist should know9:
This last point deserves emphasis: early surgery does not protect against progression to osteoarthritis, and that progression is not silent.
SWIFFT: what early surgery brings, and what it costs
439 randomised patients, functional outcome at one year and five years, complication profile
Sources: Dias JJ et al., Lancet 2020;396(10248):390-401 (PMID 32771106) for the one-year result and complications; Dias JJ et al., Bone Joint J 2026;108-B(1):70-78 (PMID 41475372) for the five-year clinical follow-up and nonunion rates.
An old trial was clear on cast length; more recent work qualifies the choice of cast, without changing the rule: it is imaging that judges union.
Gellman's randomised trial dates from 1989; it compares two casts that both include the thumb, and therefore answers only the question of length. Fifty-one patients with a non-displaced scaphoid fracture were allocated to two immobilisation modalities, followed at least until union, with a mean duration of twelve months:
The results are as follows10:
| Modality | Mean time to union | Nonunions | Delayed unions |
|---|---|---|---|
| Long then short cast (n = 28) | 9.5 weeks | 0 | 2 |
| Short cast alone (n = 23) | 12.7 weeks | 2 | 6 |
Three lessons follow. First, the more restrictive initial immobilisation shortens the time to union by more than three weeks. Second, it clearly reduces the number of nonunions and delayed unions. Finally, and this is the most useful point in practice, the advantage depends on the location of the fracture line: it was significant for fractures of the proximal and middle thirds, whereas fractures of the distal third did well whichever immobilisation modality was used10.
This trial has to be placed for what it is: 51 patients, a 1989 publication, immobilisation practices that have moved on since. It does not set a protocol, but it clearly documents the direction of the effect, and that direction is consistent with the vascular physiology described above.
A meta-analysis of seven studies (Siotos 2023) finds no significant difference in nonunion between short and long casts (9.6 versus 10.5%; odds ratio 0.79, 95% CI 0.19 to 3.26), nor according to whether the thumb is included or not (10.3 versus 10.1%). Its intervals remain wide, compatible with a notable difference in either direction, and it says nothing about time to union, which Gellman measured16. A multicentre randomised trial of 62 patients (Buijze 2014), on non-displaced or minimally displaced fractures confirmed by CT or MRI (55 of the middle third, 7 of the distal third), finds more advanced CT bridging at ten weeks without the thumb (85 versus 70%), with no difference in mobility, strength, function or pain17. In SWIFFT, the cast was applied below the elbow for six to ten weeks, and 17 of the 220 patients in the cast arm (8%) were operated on for a confirmed nonunion18.
The choice of cast is a matter for the doctor. Gellman's trial remains a reference point for fractures of the proximal third, which are absent from Buijze's trial.
The mistake not to make
Shortening the immobilisation because the patient is bothered, or because the pain has settled, is not supported by any data cited here: the disappearance of pain does not signal union of a bone whose proximal fragment may be devascularised; a referral centre (series of 64 patients, dominated by nonunions) advocates a clinical and CT check at three months14. It is imaging, not comfort, that authorises resumption.
The good news is real, and it does not remove the need to make the diagnosis early.
Grewal's study evaluated the outcome of so-called subacute fractures, that is fractures diagnosed between six weeks and six months after the injury, and treated with cast alone. Twenty-eight patients were included (20 men, 8 women, mean age 30 years), with 20 fractures of the middle third, 7 of the proximal pole and 1 of the distal pole, with union assessed by CT.
The results are encouraging11:
The authors conclude that subacute scaphoid fractures, up to six months after the injury, can unite with cast alone, even if the initial diagnosis was delayed, and that the expected time to union is shorter than what had previously been reported. This is a study of level of evidence IV, retrospective and without a control group: it describes an outcome, it does not compare strategies.
When union does not occur, the classically described course is a degenerative osteoarthritis of the wrist following a stereotyped sequence, the scaphoid nonunion advanced collapse or SNAC wrist12. This established osteoarthritis, its stages and its salvage surgeries are the subject of a separate article on this site: carpal osteoarthritis of the wrist. This article stops where that one begins, at the fracture and its union.
But Buijze's review introduces here an intellectual caveat that deserves to be reported as it stands, because it is rarely mentioned: the correlation between symptoms and disease is poor, and the true natural history is debatable, since only symptomatic patients who seek care are evaluated. The authors go further: it is not established that surgery alters the natural history, even when union is achieved12.
This finding does not invite us to neglect the diagnosis; it invites moderation in what we say. The clinical reasoning remains unchanged: early diagnosis avoids heavier treatment, there is no benefit in delaying, and caution is called for with any painful wrist that drags on. But telling a patient that progression to osteoarthritis is inevitable goes beyond what the data allow.
A series to read with caution
A review of 64 patients published in 2024 reports a nonunion rate of 82.8%, far higher than everything else in the literature. The authors themselves give the explanation: their department is a referral centre for established nonunions, which produces a major recruitment bias. This figure therefore does not describe the risk for an all-comer patient. The practical recommendation they draw from it remains useful, however: clinical follow-up and CT at three months for conservative treatment, and early referral of fractures that have not united14.
Solid evidence
Solid evidence
Weak evidence
Three-day bandage and scheduled reassessment at two weeks, for a suspicion with a normal radiograph. Non-inferior to a two-week cast on function at three months in a trial of 180 patients, but with only 16 real fractures: safety for a true fracture is not established19.
Weak evidence
Weak evidence
Union is possible for a fracture diagnosed late. Retrospective series of 28 patients, level IV, no control group11.
Very weak
Comparative accuracy of cross-sectional investigations. Eleven studies, wide confidence intervals for all three modalities, comparisons mostly indirect6.
Not recommended
Follow-up radiograph at six weeks as a means of diagnosis. Judged unsuitable by a level II diagnostic performance study: poor interobserver agreement and accuracy of 53 to 58%15.
A published case, with its identifier, that reproduces exactly the trap described in this article.
A 19-year-old basketball player, diagnosed with a ‘wrist sprain’
Braun S, Yelinek A, Cureus 2020;12(8):e9793 (PMID 32953308). Case report.
History. A 19-year-old patient, a collegiate basketball player, falls on her outstretched hand during a game. She presents to an urgent care centre where the diagnosis made is that of a wrist sprain. In view of the persistence of the pain, she then consults in sports medicine, where the diagnosis reached is that of a displaced scaphoid fracture.
Management and outcome. Treatment was conservative, with cast immobilisation including the thumb, supplemented by low-intensity pulsed ultrasound as adjuvant treatment13.
What this case teaches, and what it does not establish. It illustrates the exact sequence this article describes: a typical mechanism, a first diagnosis of sprain, and a fracture revealed by persistent symptoms. The authors also recall that the diagnosis is often missed because of the particular anatomy and position of this bone. However, an isolated case demonstrates the efficacy of no treatment, and the low-intensity pulsed ultrasound used here cannot in any case be recommended on this basis. The value of this observation is diagnostic and educational, not therapeutic.
A discrepancy in the figures worth flagging
This case reports that scaphoid fractures account for 60 to 70% of fractures of the wrist bones13, whereas Carpenter's meta-analysis uses 70% of carpal fractures3 and the SWIFFT trial 90%7. These gaps come down to the denominator used, carpal bones or wrist bones, and to the populations studied. We report them rather than choosing one: the robust finding is that the scaphoid is the most frequently fractured carpal bone1, not the exact value of the percentage.
What these data change in the assessment, in referral and in the conduct of rehabilitation.
The physiotherapist is not the one who prescribes imaging, but is very often the one who sees these patients repeatedly, and therefore the one who can catch a missed diagnosis. Three situations should trigger a return to the doctor:
Conversely, the absence of anatomical snuffbox pain is the only element that substantially lowers the probability of fracture, with a negative likelihood ratio of 0.153-0.25 depending on the sources. This does not rule it out, but it is the most useful clinical information available.
Faced with a painful wrist after a fall on the hand
A reasoning framework, and at each branch point the data that justifies it
Reasoning framework assembled from: Carpenter CR et al., Acad Emerg Med 2014, LR− of the absence of snuffbox pain and superiority of MRI (PMID 24673666); Coventry L et al., Emerg Med J 2023, LR− of 0.2 and absence of any single sign allowing exclusion (PMID 37169546); Mallee WH et al., Arch Orthop Trauma Surg 2016, unsuitability of six-week radiographs (PMID 27026536). Assembling these into a single tree is a proposal for clinical reasoning: no validated decision rule exists on this subject, as Carpenter explicitly points out.
Further reading on this site
No, and that is the central message of this article. In emergency department patients with clinical suspicion and a non-diagnostic initial radiograph, the pre-test probability of fracture is estimated at 25%, with a spread of 12 to 57% depending on the study3. Other studies measuring the same quantity find 9.0%5 or a median of 20%6. Whichever figure is used, the conclusion is the same: a normal radiograph does not rule out the diagnosis, and the strategy of a follow-up radiograph at six weeks has been judged inadequate15.
In the emergency department setting, Carpenter's meta-analysis concludes that MRI is the most accurate investigation, superior to scintigraphy, CT and ultrasound, both for confirming and for ruling out the diagnosis; failing that, CT can confirm but remains insufficient to rule out3. The Cochrane review, which reasons on pure statistical accuracy, names scintigraphy as the best performing, while stressing that it is more invasive, involves radiation, requires a delay of at least 72 hours and would lead to overtreatment of 112 patients per 1,000 versus 8 with CT6. The decision rests with the prescriber.
For a bicortical middle-third fracture displaced by 2 mm or less, no. The SWIFFT trial found no difference in functional score at one year between early surgery and cast, whereas potentially serious complications affected 14% of the operated patients versus 1% of those in a cast7. The absence of difference still held at five years, with nonunion rates of 2.1% and 3.3%8. The recommended strategy is cast first, with immediate fixation if nonunion is confirmed. Fractures displaced by more than 2 mm and those of the proximal pole fall outside this framework.
Because the fracture line often cuts the blood supply to the proximal fragment. The main artery enters through the distal pole and runs up towards the proximal pole: the more proximal the fracture line, the more dependent the fragment deprived of supply. That is why time to union depends on location, as shown by Grewal's study, with 11 weeks of immobilisation on average for the middle third versus 14 weeks for the proximal pole11.
The decision rests with the doctor and is made on imaging, not on comfort: the disappearance of pain does not signal union of a bone whose proximal fragment may be devascularised, and a referral centre advocates a clinical and CT check at three months14. If the cast is poorly tolerated, the doctor should be told, and will judge whether it can be adapted.
No, and this is reassuring information to give. In Grewal's series of fractures diagnosed between six weeks and six months after the injury and treated with cast alone, the union rate reached 82%, and 96% after excluding diabetic patients, comminuted fractures and humpback deformities11. This is a retrospective series of 28 patients, level of evidence IV: it describes a favourable outcome, it does not guarantee an individual result.
The classically described course is that of the SNAC wrist, a progressive degenerative osteoarthritis. But Buijze's review adds a seldom-cited caveat: the correlation between symptoms and disease is poor, the true natural history is debatable since only symptomatic patients who seek care are evaluated, and it has not been established that surgery alters this natural history, even when union is achieved12. It should also be noted that in the SWIFFT trial, 52.4% of patients imaged at five years had osteoarthritis in at least one joint, with a similar prevalence after surgery and after cast9.
None of the sources used for this article provides a validated return-to-sport criterion specific to this fracture, and it is better to say so than to put forward a convenient timescale. The reasoning rests on union confirmed by imaging, and not on the disappearance of pain alone. Two useful reference points: observed immobilisation times range from 9.5 to 14 weeks depending on the modality and location1011, and a referral centre advocates a clinical and CT check at three months14.
Nineteen references, all individually verified on PubMed through the E-utilities interface: identifier resolved, title, journal, year, pagination and author list checked on the record itself, abstract read before citation. The links point to the PubMed record.
Note on method
The bibliographic foundation was built before writing, subject by subject, by querying PubMed through the E-utilities interface: each identifier was resolved on its XML record, each reference checked for authors, journal, year and pagination, and each abstract read before being cited. Several publication years differ from those commonly cited elsewhere, the record being authoritative: Mallee's meta-analysis on clinical tests is dated 2014, the Cochrane imaging review 2015, and the five-year SWIFFT follow-up 2026. Three points are explicitly treated as uncertain rather than settled: the exact proportion of carpal fractures attributable to the scaphoid, which varies from 60 to 90% depending on the denominator used; the prevalence of fracture in suspected patients with a normal radiograph, measured between 9 and 25% depending on the thresholds of suspicion; and the natural history of nonunion, which Buijze's review itself describes as debatable. The absence of a validated return-to-sport criterion is flagged rather than filled in. The classification codes in the structured data block were checked against the official tables: ICD-11 NC53.0 on the WHO MMS linearisation, ICD-10 S62.0 on the official ICD-10 browser. Article written on 16 August 2026.