Treatment of Acute Scaphoid Fractures

Treatment of Acute Scaphoid FracturesSystematic Review and Meta-analysis

Zhong-gang Yin, MD*; Jian-bing Zhang, MD*; Shi-lian Kan, MD*; and Pei Wang, MD†

Whether operative treatment is a better option than nonop-erative treatment for acute nondisplaced or minimally dis-placed fractures of the scaphoid is controversial. The type ofcast that should be used for nonoperative treatment is notknown. We performed a systematic review and meta-analysisof randomized and quasirandomized trials to evaluate theeffect of operative versus nonoperative treatment and theeffect of different casting methods for nonoperative treat-ment of acute scaphoid fractures on nonunion rate, return towork, grip strength, range of wrist motion, complications,patient evaluation, and incidence of osteoarthritis. Two in-vestigators assessed trial quality and extracted data. Opera-tive treatment of acute nondisplaced or minimally displacedfractures of the scaphoid waist does not provide greater ben-efits regarding nonunion rate, return to work, grip strength,range of wrist motion, or patient satisfaction than cast im-mobilization; however, it causes more complications and,perhaps, a higher risk of scaphotrapezial osteoarthritis.There is no evidence from randomized trials to determinewhether operative treatment is superior to nonoperativetreatment for an acute proximal pole fracture of scaphoidbones. There is insufficient evidence to determine which typeof cast should be used in nonoperative treatment of nondis-placed scaphoid fractures.

Level of Evidence: Level II, therapeutic study. See theGuidelines for Authors for a complete description of levels ofevidence.

The scaphoid, a bridge between the proximal and distalcarpal bones, transfers compression loads from the hand to

the forearm and maintains normal wrist motion, carpalstability, and function of the wrist flexor and extensortendon. Scaphoid fractures are the most common type ofcarpal fractures23 and occur most commonly in youngmale adults. The primary mechanism of injury is a fall onthe outstretched hand with an extended, radially deviatedwrist. Less common mechanisms of injury may involveforced palmar flexion of the wrist26 and axial loading ofthe wrist with the hand in a fisted position.14

Various classification systems based on fracture pat-tern, fracture stability, the status of healing, and the inter-val since injury often are combined to describe the featuresof a scaphoid fracture.16 Fractures may be consideredacute when less than 3 weeks old, delayed to union be-tween 4 and 6 months old, and nonunited when more than6 months old.27 Fracture location can be identified asproximal pole, waist, and distal pole. Waist fractures areclassified as transverse, vertical oblique, or horizontal ob-lique. Scaphoid fracture instability is defined as displace-ment greater than 1 mm in any direction, lateral intra-scaphoid angulation greater than 35°, substantial bone lossor comminution, and fractures associated with dorsal in-tercalated segment instability.

Treatment of acute scaphoid fractures can be classifiedas nonoperative (cast immobilization) or operative (openor percutaneous internal fixation). Nonoperative treatmentis indicated for acute nondisplaced scaphoid fractures.However, the length of the cast (below-elbow cast orabove-elbow cast), immobilization of the thumb, and po-sition of the hand in the cast are controversial.17 Surgicaltreatment usually is indicated for acute unstable scaphoidfractures. With the proliferation of fixation screws avail-able for the scaphoid bone, there is an increasing trendtoward surgical treatment for acute scaphoid fractureseven if the fracture is nondisplaced.6 However, the ben-efits and risks associated with internal fixation of nondis-placed fractures of the scaphoid have not been established.Treatment of acute proximal pole fractures has been re-ported.16 Because the fracture line moves proximally,there is more risk of displacement and nonunion.23 Some

Received: March 19, 2006Revised: August 16, 2006; December 2, 2006Accepted: January 24, 2007From the *Department of Hand and Microsurgery, Tianjin Hospital, Tianjin,China; and the †Department of Orthopaedics, General Hospital of TianjinMedical University, Tianjin, China.Each author certifies that he or she has no commercial associations (eg,consultancies, stock ownership, equity interest, patent/licensing arrange-ments, etc) that might pose a conflict of interest in connection with thesubmitted article.Correspondence to: Zhong-gang Yin, MD, Department of Hand and Micro-surgery, Tianjin Hospital, 300211, Tianjin, China. Phone: 8610-022-28336249; Fax: 8610-022-28313462; E-mail: [email protected]: 10.1097/BLO.0b013e31803d359a

CLINICAL ORTHOPAEDICS AND RELATED RESEARCHNumber 460, pp. 142–151© 2007 Lippincott Williams & Wilkins

142

Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

investigators recommend open reduction and internal fixa-tion for proximal pole fractures involving less than 25% ofthe bone,16 whereas others suggest nonoperative treatmentwith a long arm cast incorporating the thumb.23

Given lingering questions about the choice of optimaltreatment for acute scaphoid fractures, we performed asystematic review and meta-analysis of randomized andquasirandomized, controlled trials to assess the effect ofoperative versus nonoperative treatment of acute scaphoidfractures on rates of nonunion, return to work, gripstrength, range of motion (ROM), rates of complications,patient satisfaction, and incidence of osteoarthritis. Wealso assessed the effect of different casting methods ofacute nondisplaced scaphoid fractures on rates of non-union, grip strength, and ROM of the wrist.

MATERIALS AND METHODS

We searched PubMed and the Cochrane Controlled Trials Reg-ister4 without language restriction and hand-searched the refer-ences of relevant studies. In PubMed, all phases of the Cochranehighly sensitive search strategy for randomized, controlled trialsfor PubMed13 were combined with the following subject-specificsearch strategy: scaphoid bone[mh] OR carpal bones[mh] ANDfracture[mh]. To be included, the study had to meet the follow-ing criteria regarding participants, treatment, outcomes, and typeof study. The study had to report on adult patients with clinicaland radiographic evidence of acute scaphoid fractures regardlessof their gender, age, and type of fracture. The study had to reporton operative versus nonoperative treatments or comparison be-tween different casting methods. The main surgical treatmentshad to be closed reduction and percutaneous fixation or open orarthroscopic reduction and internal fixation by Kirschner wiresor screws. The primary nonoperative treatments were castingimmobilization. Studies had to report the outcomes of nonunionof acute scaphoid fracture. We included any randomized or qua-sirandomized clinical trials with results published as a full reportin the English literature. Two reviewers independently selectedthe studies (ZGY, JBZ), and disagreements were discussed toreach a consensus.

The search strategies yielded 554 citations: 523 from PubMedand 31 from the Cochrane database. Twelve studies were poten-tially eligible. Of these, five were excluded because one wasreported in German,19 one compared different cast techniquesthat were not specific for treating scaphoid fracture,5 one was astudy performed on human cadaveric scaphoids,30 one was not aprospective randomized trial,22 and one was a study about clini-cal scaphoid fracture without radiographic evidence.28 We in-cluded seven randomized, controlled trials.1–3,6,9,11,24

Two reviewers (ZGY, JBZ) independently used a checklist of11 items to assess the methodologic quality of the studies (Table1). Disagreements were resolved by discussion followed by ar-bitration by a third reviewer if necessary. The checklist wasbased on the PEDro scale,29 which has sufficient reliability forrating quality of randomized, controlled trials.18 Because it wasnot possible to blind the participants or the treatment providers,

we made some modifications: Item 5 was substituted for “therewas blinding of all subjects” and Item 6 was substituted for“there was blinding of all therapists who administered thetherapy.” We also modified the definition of intention-to-treatanalysis as stated in Item 9. There was no exclusion according tothe score, and we did not define the quality of the studies fromtheir scores because such use is controversial.15 However, lowscores usually indicate poor methodology. A study was consid-ered high quality if it fulfilled at least three of the followingcriteria: randomization, allocation concealment, blind assess-ment, and intention-to-treat analysis.

All included studies had some methodologic limitations(Table 2). Five studies reported on randomized allocation,1–3,6,24

but three failed to describe the exact randomized procedure,which need not be specified according to the PEDro scale anddoes not affect the quality score, and failed to state whether thetreatment allocation was concealed.1,3,24 Two studies were ofquasirandomized allocation.9,11 No study adopted blinding of theoutcome assessor. Dropout or lost to followup occurred in fivestudies,1,3,6,11,24 but only one performed intention-to-treat analy-sis.6 Two studies fulfilled the criteria of high methodologic qual-ity.2,6

Two reviewers (ZGY, SLK) independently extracted data onnonunion, return to work, grip strength, ROM of the wrist, com-plications, patient evaluation, and incidence of osteoarthritis.

Six hundred ninety-two patients and four comparisons wereinvolved (Table 3). Four studies1,2,6,24 involving 228 patientscompared operative treatment with nonoperative treatment. Twoof these,1,2 involving 78 patients, compared percutaneous Acu-

TABLE 1. Checklist for Assessing theMethodologic Quality of the Randomized,Controlled Trials

1. Eligibility criteria were specified2. Subjects were randomly allocated to groups3. Allocation was concealed4. The groups were similar at baseline regarding the most

important prognostic indicators (eg, age, gender, affectedside, mechanism of injury, type of fracture)

5. Cointerventions were avoided or similar for all groups6. The outcome measures (nonunion, osteoarthritis) were clearly

defined in the text with a definition of any ambiguous termsencountered

7. There was blinding of all assessors who measured at leastone key outcome

8. Measures of at least one key outcome were obtained frommore than 85% of the subjects initially allocated to groups

9. The investigators characterized the analysis of data asintention-to-treat analysis and there was an attempt to analyzedata from all randomized participants as if each subjectreceived the treatment or control condition as planned orthere was no dropout even if the analysis was not specificallydescribed as intention-to-treat

10. The results of between-group statistical comparisons arereported for at least one key outcome

11. The study provides point measures and measures ofvariability for at least one key outcome

Number 460July 2007 Treatment of Scaphoid Fractures 143


TABLE 2. Methodologic Quality Scores

Study 1 2 3 4 5 6 7 8 9 10 11 Score

Bond et al2 + + + + + + ! + + + + 10Saeden et al24 + + ? + + + ? + ! + + 8Adolfsson et al1 + + ? + + ! ? + ! + + 7Dias et al6 + + + + + + ! + + + + 10Hambidge et al11 + ! ! + + ! ? ? ! + + 5Clay et al3 + + ? + + + ! ! ! + ! 6Gellman et al9 + ! ! + + + ? + + + ! 7

All criteria were scored yes (+), no (!), or unclear (?); Table 1 lists criteria

TABLE 3. Study Characteristics

Study and Treatments Participants Fracture Type Followup Outcomes

Bond et al2

(A) Percutaneousinternal fixation;(B) Above-elbow thumbspica cast for 6 weeksfollowed by below-elbow cast

(A) n = 11, M/F: 9/2;(B) n = 14, M/F: 13/1;age: 24 years (range,18–34 years)

Acute waist nondisplacedfracture

25 months(range, 24–27months);none lost tofollowup

Nonunion, range of motion,grip strength, time tounion, return to work,patient satisfaction,complications

Saeden et al24

(A) Open reductioninternal fixation;(B) Below-elbow thumbspica cast

(A) n = 32, M/F: 27/5,age: 29 ± 13 years;(B) n = 30, M/F: 22/8,

age: 37 ± 20 years

(A) C2.1 (1), C2.2 (25),C2.3 (3), C3 (3);(B) C2.1 (2), C2.2 (27),C2.3 (0), C3 (1) (AOclassification)

11.7 years(range,10.2–12.8years); 11lost tofollowup

Nonunion, range of motion,grip strength, return towork, osteoarthritis, pain,complications

Adolfsson et al1 (A)Percutaneous internalfixation; (B)Below-elbow thumbspica cast

(A) n = 25, M/F: 20/5,age: 30 years (range,16–75 years); (B) n =28, M/F: 19/9, age: 33years (range, 15–73years)

Acute waist nondisplacedfracture

24 weeks; 3 lostat 10 weeks,14 lost at 24weeks

Nonunion, range of motion,grip strength, time tounion, complications

Dias et al6 (A) Openreduction internalfixation;(B) Below-elbow castwith thumb free

(A) n = 44, M/F: 40/4,age: 29 ± 1.32years; (B) n = 44, M/F:39/5, age: 29.9 ± 1.6years

Acute waist nondisplaced ormild displaced fracture

52 weeks; 7 lostto followup

Nonunion, range of motion,grip strength, return towork, patient satisfaction,pain, complications

Hambidge et al11

(A) Cast with wristflexion; (B) Cast withwrist extension

(A) n = 58; (B) n = 63,M/F: 95/26, age: 30years (range, 16–76years)

(A) Waist (54), distalpole (4); (B) Waist (57),distal pole (6)

6 months; lost tofollowup - notstated

Nonunion, range of motion,grip strength, pain

Clay et al3 (A) Cast withthumb enclosed;(B) Cast leaving thumbfree

(A) n =144; (B) n = 148,/F: 222/70, age: 29.7years (range, 16–71years)

(A) Waist (116), distalpole (15), proximalpole (9), unclassified (4);(B) Waist (133), distalpole (10), proximalpole (2), unclassified (3)

6 months Nonunion, life ability

Gellman et al9 (A) Longthumb spica cast;(B) Short thumb spicacast

(A) n = 28; (B) n = 23,M/F: 46/5, age: 30years (range, 14–57years)

(A) Nondisplaced waist (22),proximal pole (2), distalpole (3), tubercle (1); (B)Nondisplaced waist (18),proximal pole (3), distalpole (2)

12 months(range, 6–24months);none lost tofollowup

Nonunion, delayed union,time to union, avascularnecrosis

M/F = male/female

Clinical Orthopaedicsand Related Research144 Yin et al


trak screw (Acumed, Beaverton, Oregon) fixation with cast im-mobilization. Two studies,6,24 involving 150 patients, comparedopen reduction Herbert screw fixation with cast immobilization.Three of the four trials used below-elbow casts for the nonop-erative group.1,6,24 One initially used an above-elbow cast for6 weeks and then a below-elbow cast.2 Although there was het-erogeneity across the trials that compared operative treatmentwith nonoperative treatment such as different surgical techniques(eg, percutaneous internal fixation or open reduction and internalfixation), different types of cast (eg, long arm cast or short armcast), and different durations of external fixation, the results arepresented for the basic comparison; namely, operative treatment(internal fixation) versus nonoperative treatment (cast immobi-lization). Three of the seven included studies (464 patients) com-pared different nonoperative treatment regimens and involvedthree different comparisons: Colles-type below-elbow cast withthe wrist in flexion versus cast with the wrist in extension,11 acast with the thumb enclosed versus a cast leaving the thumbfree,3 and a long thumb spica cast for 6 weeks followed by ashort thumb spica cast versus a short thumb spica cast.9 Three ofthe four studies comparing operative with nonoperative treat-ment, involving 166 patients, stated they only included acutenondisplaced or minimally displaced fractures of the scaphoidwaist.1,2,6 The other study, involving 62 patients, reported frac-ture type according to the AO system, and the majority of frac-tures were Type C2.2 (52 of 62 fractures).24 We did not findrandomized, controlled trials that evaluated surgical treatment ofthe proximal pole fracture of scaphoid bones.

We calculated treatment effect sizes between groups and 95%confidence intervals for each randomized comparison for eachoutcome. The results of individually randomized trials werepooled whenever possible using RevMan 4.2 software (TheCochrane Collaboration, Oxford, UK). For dichotomous out-comes, relative risk and 95% confidence interval were used tosummarize the effect. For continuous outcomes, effect sizeswere reported as weighted mean differences or standardizedmean differences if the trials measured the same outcome invarious ways along with 95% confidence intervals. When thestandard deviation at followup was not available, we used thestandard deviation at baseline. If none was reported, we assumedthe average standard deviation reported by other studies for thatoutcome. We used I2, which describes the percentage of totalvariation across studies that is from heterogeneity rather thanchance and does not inherently depend on the number of studiesin the meta-analysis12 to assess the heterogeneity. The quality

lies between 0% and 100%. A value of 0% indicates no hetero-geneity and larger values show increasing heterogeneity. If I2

was less than 50%, we used the fixed effects model to combinethe results. If I2 was greater than 50%, we used the randomeffects model.

We performed sensitivity analysis under the following con-ditions: (1) changing the cutoff point of the heterogeneity to 25%and using the random effects meta-analysis model if I2 wasgreater than 25%; (2) for missing dichotomous data, performingbest-case and worst-case scenario analysis; and (3) using onlypooled results of studies with high methodologic quality. A pvalue less than 0.05 was considered significant.

RESULTS

The pooled relative risk for nonunion was 0.26 (95% con-fidence interval, 0.07–0.91; p ! 0.04; homogeneity I2 !47.8%) for the four studies that compared operative withnonoperative treatments,1,2,6,24 indicating a lower rate ofnonunion in the operative group than in the nonoperativegroup (Table 4; Fig 1). If we changed the cutoff point ofI2 to 25%, the results were pooled with a random effectsmodel and the relative risk for nonunion was 0.41 (95%confidence interval, 0.05–3.69; homogeneity I2 ! 47.8%),indicating no difference between the two groups. Thepooled relative risk for nonunion with the best-case sce-nario analysis in which the missing data were consideredas union was 0.40 (95% confidence interval, 0.04–3.87;homogeneity I2 ! 50.4%), whereas the pooled relativerisk for nonunion with the worst-case scenario analysiswas 0.63 (95% confidence interval, 0.28–1.40; homoge-neity I2 ! 16.8%), indicating no difference between thetwo groups. When the analysis was limited to the twohigh-quality studies,2,6 the pooled relative risk for non-union was 0.06 (95% confidence interval, 0–0.94; p !0.05), which approached significance. Dias et al6 reportedthe highest rate of nonunion in the nonoperative group. Intheir trial, the period of cast immobilization was 8 weeks,and four fractures that were not united at 12 weeks had noobvious mobility at the fracture site and could not be easilyidentified at the time of surgery. It was possible these fourfractures could have united with additional immobiliza-

TABLE 4. Relative Risk for Nonunion

Study

Operative Number ofNonunions/Number

of Participants

Nonoperative Numberof Nonunions/Number

of Participants

Relative Risk(fixed effect)

(95% confidence interval)

Adolfsson et al1 1/24 0/26 3.24 (0.14 to 75.91)Bond et al2 0/11 0/14 Not estimableDias et al6 0/39 9/42 0.06 (0.00 to 0.94)Saeden et al24 1/32 2/30 0.47 (0.04 to 4.91)Total 2/106 11/112 0.26 (0.07 to 0.91)



tion. If we assumed the four fractures had healed, therelative risk for nonunion was 0.39 (95% confidence in-terval, 0.10–1.44; homogeneity I2 ! 24.7%), indicatingno difference between the two groups.

The pooled weighted mean difference for return towork was !5.47 weeks (95% confidence interval, !9.97–!0.98; p ! 0.02) for the three studies2,6,24 that reportedthis outcome, suggesting patients with surgical treatmentreturned to work more rapidly (Table 5; Fig 2). There wassubstantial heterogeneity (I2 ! 92.7%). When the analysiswas limited to the two high-quality studies,2,6 the pooledweighted mean difference for return to work was !4.10weeks (95% confidence interval, !9.99–1.78), suggestingno difference between operative and nonoperative groups.However, the heterogeneity was still substantial (I2 !96.2%).

The weighted mean difference for grip strength of in-dividual studies with different timing of outcome assess-ment was 0.05 (95% confidence interval, !0.08–0.18) at16 weeks postoperatively, 0.06 (95% confidence interval,!0.01–0.13) at 26 weeks, 0.11 (95% confidence interval,!0.12–0.34) at 2 years followup, and 0.07 (95% confi-dence interval, !0.03–0.17) at 12 years followup, indicat-ing no difference between operative and nonoperativegroups; and 0.09 (95% confidence interval, 0.01–0.17;p ! 0.02) at 12 weeks postoperatively and 0.07 (95%confidence interval, 0–0.14; p ! 0.04) at 52 weeks in

favor of surgical treatment (Table 6; Fig 3). The resultscould not be pooled because the timing of the outcomeassessment was different across studies.

The weighted mean difference for ROM of the wrist inindividual studies with different timing of outcome assess-ment was 0.01 (95% confidence interval, !0.06–0.08) at12 weeks, !0.01 (95% confidence interval, !0.07–0.05) at26 weeks, 0.01 (95% confidence interval, !0.05–0.07) at52 weeks, 0.07 (95% confidence interval, !0.02–0.16) at 2years followup, and !0.02 (95% confidence interval,!0.08–0.04) at 12 years followup, indicating no differencebetween operative and nonoperative groups; and 0.07(95% confidence interval, 0.01–0.13; p ! 0.02) at 16weeks in favor of surgical treatment (Table 7; Fig 4).

The pooled relative risk for complications was 12.23(95% confidence interval, 2.32–64.60; p ! 0.003; homo-geneity I2 ! 0%) for the four studies1,2,6,24 comparingoperative and nonoperative treatment, suggesting surgicaltreatment caused more complications (Table 8; Fig 5). Thepooled relative risk for complications with the best-casescenario analysis was 11.82 (95% confidence interval,2.20–63.61; p ! 0.004; homogeneity I2 ! 0%), whereasthe pooled relative risk for complications with the worst-case scenario analysis was 5.03 (95% confidence interval,1.87–13.49; p ! 0.001; homogeneity I2 ! 36.3%), sug-gesting surgical treatment caused more complications.When only the two high-quality studies2,6 were pooled, the

Fig 1. When the four studies com-paring operative with nonoperativetreatment were pooled, the opera-tive treatment produced fewer non-unions because the pooled relativerisk for nonunion was 0.26 (95%confidence interval, 0.07 to 0.91).

TABLE 5. Weighted Mean Difference of Return to Work

Study

Operative Nonoperative

Weighted Mean Difference(95% confidence interval)

Number ofParticipants

Mean(standard deviation)



Bond et al2 11 8.00 (0.70) 14 15.00 (0.70) !7.00 (!7.55–!6.45)Dias et al6 39 5.00 (5.12) 42 6.00 (5.12) !1.00 (!3.23–1.23)Saeden et al24 24 6.00 (3.00) 21 15.00 (10.00) !9.00 (!13.44–!4.56)Total 74 77 !5.47 (!9.97–!0.98)



relative risk for complications was 16.90 (95% confidenceinterval, 2.22–128.36; p ! 0.006; homogeneity I2 !4.3%), suggesting surgical treatment resulted in morecomplications. The results of all four studies consideredtogether showed an incidence of complications in 15 of106 patients (14.2%) in the operative treatment group and0 of 112 patients in the nonoperative treatment group. Thecomplications included superficial infection (one patient),sensory change in the region of the palmar cutaneousbranch of the median nerve (one patient), screw problem(one patient), reflex sympathetic dystrophy (two patients),and problems related to scarring (10 patients).

The pooled standardized mean difference for patientsatisfaction for the two studies2,6 that reported this out-come with a different scale was !1.04 (95% confidenceinterval, !2.92–0.84; homogeneity I2 ! 91.6%), indicat-ing no difference between the operative and nonoperativegroups (Table 9; Fig 6).

For one study with a 12-year followup reporting theincidence of osteoarthritis,24 the relative risk for scaph-otrapezial joint osteoarthritis was 2.43 (95% confidenceinterval, 0.98–6.03), which approached significance. Theincidence of scaphotrapezial joint osteoarthritis was 60.9%(14 of 23) in the operative group and 25% (four of 16) inthe nonoperative group. However, the rate of symptomaticscaphotrapezial osteoarthritis was similar in the operativeand nonoperative groups, 13% (three of 23) versus 18.8%(three of 16), and the relative risk for symptomatic scaph-

otrapezial osteoarthritis was 0.70 (95% confidence inter-val, 0.16–3.02). The relative risk for radiocarpal osteoar-thritis was 1.25 (95% confidence interval, 0.52–3.04). Therate of radiocarpal osteoarthritis was 39.1% (nine of 23)and 31.3% (five of 16) in the operative and nonoperativegroups, respectively, and the rate of symptomatic radio-carpal osteoarthritis was also similar in the two groups,8.7% (two of 23) versus 12.5% (two of 16). When weperformed the best-case scenario analysis, the relative riskfor scaphotrapezial joint osteoarthritis was 3.28 (95% con-fidence interval, 1.22–8.86; p ! 0.02) and the relative riskfor radiocarpal osteoarthritis was 1.69 (95% confidenceinterval, 0.64–4.47), suggesting surgical treatment in-creased scaphotrapezial joint osteoarthritis risk. When theworst-case scenario analysis was performed, the relativerisk for scaphotrapezial osteoarthritis was 1.20 (95% con-fidence interval, 0.83–1.72) and the relative risk for radio-carpal osteoarthritis was 0.89 (95% confidence interval,0.59–1.34), indicating no difference between the operativeand nonoperative groups.

Three studies that reported different comparisons be-tween nonoperative treatments3,9,11 found no difference innonunion rate between different casting methods. Ham-bidge et al11 compared the Colles-type below-elbow castwith the wrist in 20° flexion and in 20° extension; theposition of the wrist did not influence the rate of nonunion(8.6% versus 12.7%; relative risk ! 0.68; 95% confidenceinterval, 0.24–1.96). Clay et al3 reported no difference in

TABLE 6. Weighted Mean Difference for Grip Strength

Study Timing

Operative NonoperativeWeighted MeanDifference (95%

confidence interval)Number

of ParticipantsMean

(standard deviation)Number ofParticipants


Dias et al6 12 weeks 44 0.84 (0.16) 44 0.75 (0.20) 0.09 (0.01–0.17)Adolfsson et al1 16 weeks 12 0.88 (0.17) 21 0.83 (0.22) 0.05 (!0.08–0.18)Dias et al6 26 weeks 44 0.94 (0.13) 44 0.88 (0.19) 0.06 (!0.01–0.13)Dias et al6 52 weeks 44 0.99 (0.10) 44 0.92 (0.20) 0.07 (0.00–0.14)Bond et al2 2 years 11 0.95 (0.29) 14 0.84 (0.29) 0.11 (!0.12–0.34)Saeden et al24 12 years 32 0.95 (0.17) 30 0.88 (0.22) 0.07 (!0.03–0.17)

Fig 2. When the three studies com-paring operative with nonoperativetreatment and included time of re-turn to work as an outcome werepooled, the operative treatment per-formed better because the pooledweighted mean difference for returnto work was !5.47 weeks (95% con-fidence interval, !9.97 to !0.98).



nonunion rate with a below-elbow thumb spica or Colles-type below-elbow cast (9.8% versus 10.1%; relative risk! 0.97; 95% confidence interval, 0.48–1.93). Gellman etal9 compared initial use of an above-elbow thumb spicacast followed by the use of a below-elbow thumb spicacast with the use of a below-elbow thumb spica cast only;no difference was found (0% versus 8.7%; relative risk !0.17; 95% confidence interval, 0.01–3.28).

Hambidge et al11 reported the Colles-type cast with thewrist in 20° extension resulted in greater grip strength(weighted mean difference ! 5.0; 95% confidence inter-val, 1.02–8.98; p ! 0.01) and better range of wrist exten-sion (weighted mean difference ! 12°; 95% confidence in-terval, 7.61–16.39; p < 0.0001) at 6 months postoperatively.

DISCUSSION

This systematic review and meta-analysis of randomized,controlled trials involving the treatment of acute scaphoidfracture met most of the methodologic criteria that have

been suggested for meta-analysis,20 but some study limita-tions need to be considered when interpreting our findings.

First, we included only studies published as a full reportin English, which produces publication and language bias.Unpublished studies and studies presented only in abstractform are more likely to have nonsignificant results.7,25

Trials with significant results are more likely to be pub-lished in English.8 However, we identified only one fullreport not published in English, which evaluated the roleof pulsed low-intensity ultrasound in accelerating healingof acute scaphoid fractures.19 Therefore, the language biasmight have only limited impact on our results. Second, weassumed standard deviations to calculate treatment effectsizes of grip strength and ROM. In theory, this would notbias our estimates because the same assumption was ap-plied to treatment and control groups. Calculating treat-ment effect sizes allows meaningful comparisons acrossstudies.10 A third limitation was that few studies met theinclusion criteria, and there was heterogeneity across thestudies. The small number of included studies hampered

Fig 3. The weighted mean differ-ence for grip strength of individualstudies with different timing of theoutcome assessment did not show adifference between the operativeand nonoperative groups at 16weeks, 26 weeks, 2 years, and 12years followup.

TABLE 7. Weighted Mean Difference for Range of Wrist Motion

Study Timing


Weighted Mean Difference(95% confidence interval)





Dias et al6 12 weeks 44 0.81 (0.19) 44 0.80 (0.14) 0.01 (!0.06–0.08)Adolfsson et al1 16 weeks 12 0.94 (0.08) 20 0.87 (0.08) 0.07 (0.01–0.13)Dias et al6 26 weeks 44 0.89 (0.13) 44 0.90 (0.17) !0.01 (!0.07–0.05)Dias et al6 52 weeks 44 0.94 (0.12) 44 0.93 (0.16) 0.01 (!0.05–0.07)Bond et al2 2 years 11 0.93 (0.12) 14 0.86 (0.12) 0.07 (!0.02–0.16)Saeden et al24 12 years 32 0.96 (0.13) 30 0.98 (0.13) !0.02 (!0.08–0.04)



specific subgroup analyses. Although different surgicaltechniques were used, we made only basic comparisons(eg, internal fixation versus cast immobilization). We be-lieve percutaneous internal fixation may provide bettertreatment compared with open reduction and internal fixa-tion. A fourth limitation is that the included studies, espe-cially the three studies that compared nonoperative treat-ments, had a variable level of methodologic limitations.These limitations included a lack of concealed allocation,lack of blinding of outcome assessment, and lack of in-tention-to-treat analysis. The quality of the included studyused for pooling analysis affects the results of a meta-analysis.21

Although the pooled result revealed a higher rate ofnonunion in the nonoperative group, this finding was notrobust because the sensitivity analysis (pooled results with

a random effects model, best-case and worst-case scenarioanalysis, or pooled results of high-quality studies only)showed no difference between the operative and nonop-erative group. Furthermore, if Dias et al6 had provided aslightly longer period of cast immobilization in the non-operative group, the pooled result would have shown nodifference in nonunion rate between the operative and non-operative groups. The pooled results revealed internalfixation resulted in faster return to work, but there wassubstantial heterogeneity. Furthermore, the sensitiveanalysis with pooled results of high-quality studies onlyshowed no difference between the operative and nonop-erative groups. The time of return to work can be influ-enced by a host of factors such as the patient’s occupationand the attitude of the patient, physician, and employer,which can explain the heterogeneity. Although the results

Fig 4. The weighted mean differ-ence for range of wrist motion of in-dividual studies with different timingof the outcome assessment did notshow a difference between the op-erative and nonoperative groups at12 weeks, 26 weeks, 52 weeks, 2years, and 12 years followup.

TABLE 8. Relative Risk for Complications

Citation

Operative Number ofParticipants with

Complication/Numberof Total Participants

Nonoperative Number ofParticipants with

Complication/Numberof Total Participants

Relative Risk(fixed effect)

(95% confidence interval)

Adolfsson et al1 1/24 0/26 3.24 (0.14–75.91)Bond et al2 1/11 0/14 3.75 (0.17–84.02)Dias et al6 13/39 0/42 29.03 (1.78–472.39)Saeden et al24 0/32 0/30 Not estimableTotal 15/106 0/112 12.23 (2.32–64.60)



of grip strength and ROM of the wrist could not be sta-tistically pooled because of different timing of outcomemeasures, there was a trend suggesting early internal fixa-tion can provide only transient benefits compared withnonoperative treatment. However, there were more com-plications after operative treatment than after nonoperativetreatment. The complications fell into the following cat-egories: screw problem, reflex sympathetic dystrophy, su-perficial wound infection, problems related to scarring,and sensory change in the region of the palmar cutaneousbranch of the median nerve. Although the majority of thesurgical complications were related to the scar and minorscar-related complications could be minimized by the per-cutaneous technique, nonoperative treatment was a saferprocedure. Only one study24 reported the incidence of os-teoarthritis. The result showed internal fixation resulted inmore scaphotrapezial osteoarthritis than with nonoperativetreatment (60.9% versus 25%), but the effect size was notsignificant. The rates of symptomatic scaphotrapezial andradiocarpal osteoarthritis were similar between operativeand nonoperative groups. However, the findings were notstrong because radiographic followup was incomplete (23of 32 in the operative group, 16 of 30 in the nonoperativegroup), and the results of the sensitivity analysis wereinconsistent.

Three studies3,9,11 comparing different cast treatmentsrevealed the type of cast did not influence the union rate,but the Colles-type cast with the wrist in slight extension

resulted in better grip strength and ROM than the cast withthe wrist in flexion. These studies had substantial meth-odologic limitations, and their results could not be pooledbecause they involved different comparisons. Given thelarge numbers involved (292 patients) and the point esti-mate of relative risk approaching 1.0, which means theeffect would not be modified with larger numbers, theconclusion of the study comparing a below-elbow thumbspica cast with a below-elbow Colles type cast is, perhaps,the strongest among the three studies.3 Only limited con-clusions can be drawn from the study comparing an above-elbow thumb spica cast followed by a below-elbow thumbspica cast with a below-elbow thumb spica cast only be-cause of the small numbers involved (51 patients) and thepoint estimate of relative risk substantially deviating from1.0,9 which means the effect might be modified with alarger sample size.

Early internal fixation of acute nondisplaced or mini-mally displaced fractures of the scaphoid waist does notprovide greater benefits in terms of union and functionthan cast immobilization; however, it causes more com-plications and, perhaps, a higher risk of scaphotrapezialosteoarthritis. Although earlier return to work may be ob-tained with internal fixation, there is inconsistency acrossstudies. Because time of return to work is influenced bymany factors, it may not be a precise outcome measure forthe effectiveness of treatment. Additional rigorously con-ducted randomized trials are needed to establish whether

Fig 5. When the four studies com-paring operative with nonoperativetreatment were pooled, the opera-tive treatment produced more com-plications because the pooled rela-tive risk for complications was 12.23(95% confidence interval, 2.32–64.60).

TABLE 9. Standardized Mean Difference of Patient Evaluation

Citation


Standardized Mean Difference(95% confidence interval)





Bond et al2 11 !3.80 (0.20) 14 !3.10 (0.40) !2.06 (!3.07–!1.06)Dias et al6 44 3.90 (6.40) 44 5.20 (11.33) !0.14 (!0.56–0.28)Total 55 58 !1.04 (!2.92–0.84)



operative treatment is superior to nonoperative treatmentregarding acute proximal pole fractures of scaphoid bonesand which type of cast should be used in nonoperativetreatment of nondisplaced scaphoid fractures.

AcknowledgmentWe thank Ma Ning for literature collection.

References1. Adolfsson L, Lindau T, Arner M. Acutrak screw fixation versus cast

immobilisation for undisplaced scaphoid waist fractures. J HandSurg Br. 2001;26:192–195.

2. Bond CD, Shin AY, McBride MT, Dao KD. Percutaneous screwfixation or cast immobilization for nondisplaced scaphoid fractures.J Bone Joint Surg Am. 2001;83:483–488.

3. Clay NR, Dias JJ, Costigan PS, Gregg PJ, Barton NJ. Need thethumb be immobilised in scaphoid fractures? A randomised pro-spective trial. J Bone Joint Surg Br. 1991;73:828–832.

4. Cochrane Controlled Trials Register. Available at: www.mrw.interscience.wiley.com/cochrane/cochrane_search_fs.html. Ac-cessed December 10, 2005.

5. Cohen AP, Shaw DL. Focused rigidity casting: a prospective ran-domised study. J R Coll Surg Edinb. 2001;46:265–270.

6. Dias JJ, Wildin CJ, Bhowal B, Thompson JR. Should acute scaph-oid fractures be fixed? A randomized controlled trial. J Bone JointSurg Am. 2005;87:2160–2168.

7. Egger M, Smith GD. Bias in location and selection of studies. BMJ.1998;316:61–66.

8. Egger M, Zellweger-Zahner T, Schneider M, Junker C, Lengeler C,Antes G. Language bias in randomised controlled trials published inEnglish and German. Lancet. 1997;350:326–329.

9. Gellman H, Caputo RJ, Carter V, Aboulafia A, McKay M. Com-parison of short and long thumb-spica casts for non-displaced frac-tures of the carpal scaphoid. J Bone Joint Surg Am. 1989;71:354–357.

10. Guzman J, Esmail R, Karjalainen K, Malmivaara A, Irvin E, Bom-bardier C. Multidisciplinary rehabilitation for chronic low backpain: systematic review. BMJ. 2001;322:1511–1516.

11. Hambidge JE, Desai VV, Schranz PJ, Compson JP, Davis TR, Bar-ton NJ. Acute fractures of the scaphoid: treatment by cast immo-bilisation with the wrist in flexion or extension? J Bone Joint SurgBr. 1999;81:91–92.

12. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring in-consistency in meta-analyses. BMJ. 2003;327:557–560.

13. Higgins JPT, Green S, eds. Highly sensitive search strategies foridentifying reports of randomized controlled trials in MEDLINE.Cochrane Handbook for Systematic Reviews of Interventions 4.2.5

[Updated May 2005]; Appendix 5b. Available at: www.cochrane.org/resources/handbook/hbook.htm. Accessed May 31, 2005.

14. Horii E, Nakamura R, Watanabe K, Tsunoda K. Scaphoid fractureas a ‘puncher’s fracture.’ J Orthop Trauma. 1994;8:107–110.

15. Juni P, Witschi A, Bloch R, Egger M. The hazards of scoring thequality of clinical trials for meta-analysis. JAMA. 1999;282:1054–1060.

16. Kalainov DM, Lee Osterman A. Diagnosis and management ofscaphoid fractures. In: Watson HK, Weinzweig J, eds. The Wrist.Philadelphia, PA: Lippincott Williams & Wilkins; 2001:187–202.

17. Krasin E, Goldwirth M, Gold A, Goodwin DR. Review of thecurrent methods in the diagnosis and treatment of scaphoid frac-tures. Postgrad Med J. 2001;77:235–237.

18. Maher CG, Sherrington C, Herbert RD, Moseley AM, Elkins M.Reliability of the PEDro scale for rating quality of randomizedcontrolled trials. Phys Ther. 2003;83:713–721.

19. Mayr E, Rudzki MM, Rudzki M, Borchardt B, Hausser H, Ruter A.Does low intensity, pulsed ultrasound speed healing of scaphoidfractures? Handchir Mikrochir Plast Chir. 2000;32:115–122.

20. Moher D, Cook DJ, Eastwood S, Olkin I, Rennie D, Stroup DF.Improving the quality of reports of meta-analyses of randomisedcontrolled trials: the QUOROM statement. Lancet. 1999;354:1896–1900.

21. Moher D, Pham B, Jones A, Cook DJ, Jadad AR, Moher M, Tug-well P, Klassen TP. Does quality of reports of randomised trialsaffect estimates of intervention efficacy reported in meta-analyses?Lancet. 1998;352:609–613.

22. Papaloizos MY, Fusetti C, Christen T, Nagy L, Wasserfallen JB.Minimally invasive fixation versus conservative treatment of undis-placed scaphoid fractures: a cost-effectiveness study. J Hand SurgBr. 2004;29:116–119.

23. Phillips TG, Reibach AM, Slomiany WP. Diagnosis and manage-ment of scaphoid fractures. Am Fam Physician. 2004;70:879–884.

24. Saeden B, Tornkvist H, Ponzer S, Hoglund M. Fracture of the carpalscaphoid: a prospective, randomised 12-year follow-up comparingoperative and conservative treatment. J Bone Joint Surg Br. 2001;83:230–234.

25. Scherer RW, Dickersin K, Langenberg P. Full publication of resultsinitially presented in abstracts: a meta-analysis. JAMA. 1994;272:158–162.

26. Shestak K, Ruby LK. An unusual fracture of the scaphoid. J HandSurg Am. 1983;8:925–928.

27. Simonian PT, Trumble TE. Scaphoid nonunion. J Am Acad OrthopSurg. 1994;2:185–191.

28. Sjolin SU, Andersen JC. Clinical fracture of the carpal scaphoid:supportive bandage or plaster cast immobilization? J Hand Surg Br.1988;13:75–76.

29. The Physiotherapy Evidence Database. Available at: www.pedro.fhs.usyd.edu.au/scale_item.html. Accessed December 20, 2005.

30. Toby EB, Butler TE, McCormack TJ, Jayaraman G. A comparisonof fixation screws for the scaphoid during application of cyclicalbending loads. J Bone Joint Surg Am. 1997;79:1190–1197.

Fig 6. When the two studies com-paring operative with nonoperativetreatment and included patientevaluation as an outcome werepooled, no difference was found be-cause the pooled standardizedmean difference for patient evalua-tion was !1.04 (95% confidence in-terval, !2.92–0.84).



Treatment of Acute Scaphoid Fractures

Documents