Traditionally, nondisplaced and minimally displaced fractures have been considered by most surgeons to be stable. Several studies have demonstrated predictable rates of healing in association with conservative treatment for these types of fractures, ranging from 90% to 100%1-7. In contrast, displaced scaphoid fractures—defined by most authors as those with >1 mm of displacement—have been recognized as unstable fractures that are associated with a significant risk of nonunion if not treated surgically8. Currently, there is a trend in orthopaedic practice toward early open reduction and internal fixation for the treatment of fractures that traditionally have been treated conservatively9-11. Possible explanations for this trend are increased patient expectations and more complete follow-up, including the emphasis on patient-based outcome measures of function and health status12.
Recent reports have advocated surgical treatment for nondisplaced and minimally displaced scaphoid fractures13-15. Two previous meta-analyses on the treatment of acute scaphoid fractures indicated that current evidence failed to prove the superiority of operative as compared with conservative treatment16,17. However, since the publication of those reviews, additional data have been published, and the discussion about the indications for operative treatment is ongoing13,18. Even though the present meta-analysis has substantial overlap with previous ones, we believe that it adds methodological quality and incorporates recently published data to provide the best currently available evidence to support clinical decision-making19.
The aim of the present review was to evaluate the evidence from randomized controlled clinical trials comparing surgical and conservative treatment for acute nondisplaced and minimally displaced fractures of the scaphoid. Furthermore, we aimed to evaluate methodological limitations in current studies evaluating these therapeutic options for scaphoid fractures. The primary outcome measures of the two treatment groups were patient-rated and physician-rated wrist function.
The present study is reported following the QUOROM (Quality of Reporting of Meta-analyses) guidelines20.
Types of Studies, Participants, and Interventions
In order to be included in the present systematic review, a study had to be a randomized or quasi-randomized controlled trial that compared surgical and conservative treatment of acute nondisplaced or minimally displaced scaphoid fractures in adult patients. A quasi-randomized trial is one in which participants are allocated to treatment in a manner that is not strictly random (e.g., date of birth, hospital record number).
The present review excluded (1) trials that included children or patients with congenital deformities or degenerative conditions and (2) trials that focused on the treatment of delayed union (four weeks to six months after a fracture) or nonunion (more than six months after a fracture).
Surgical intervention included internal fixation by means of either an open or percutaneous approach. Conservative modalities included all types of cast immobilization regardless of the length of the cast, immobilization of the thumb, or position of the hand in the cast.
Types of Outcome Measures
The primary outcome measure of this review was functional outcome based on validated hand and wrist function scores, including the Disabilities of the Arm, Shoulder and Hand (DASH) score21, the Patient Evaluation Measure (PEM)22, the Patient-Rated Wrist Evaluation (PRWE)23, and the modified Green and O'Brien score24. The secondary outcome measures of this review were patient satisfaction, pain, physician-rated functional outcome (such as range of wrist motion and grip strength), time to union, time off work, return to previous activity, costs, and negative outcomes (including the rates of infection, malunion, and nonunion). Nonunion was defined as failure of the fracture to unite at more than six months after injury, with radiographic evidence of a fracture line.
Search Methods for Identification of Studies
Our methods for the identification of studies were similar to the Cochrane Bone, Joint and Muscle Trauma Group methods used in reviews. A systematic search, from inception of the database in 1966 to February 27, 2009, was performed with the following terms: scaph* and fractur* and random*. Databases included the Cochrane Bone, Joint and Muscle Trauma Group Specialised Register, the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, EMBASE, CINAHL, and reference lists of articles. Researchers in the field were contacted to inquire about any additional unpublished trials or trials in progress.
Selection of Studies
Two trained orthopaedic research physicians (G.A.B. and J.N.D.) assessed abstracts of all studies identified during the initial search. Full copies of the report of potentially relevant studies were independently assessed by those two physicians with use of the above criteria. Disagreements were resolved by means of discussion, with arbitration by an experienced orthopaedic surgeon with training in clinical epidemiology (R.W.P.) when differences of opinion remained.
Assessment of Methodological Quality
The quality of the selected studies was independently assessed, without masking the source of authorship of the trial reports, by two physicians (G.A.B. and J.N.D.) with use of a quality-assessment tool derived from the Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group25. To evaluate the methodological quality of the selected studies, we scored the methods of allocation and blinding and the loss to follow-up. The GRADE criteria were used to evaluate the quality of evidence according to outcome. Disagreement was resolved by means of discussion and, if necessary, by means of scrutiny by an independent orthopaedic surgeon with a doctorate in study methodology (R.W.P.).
Data Collection
The two physicians independently extracted data from all eligible studies with use of a piloted data-extraction form. Disagreement was resolved as described above. All authors of the selected trials were contacted to obtain unpublished original raw data in order to complete the data-extraction form. For all continuous outcomes, means and standard deviations were extracted for analysis. If means and confidence intervals were reported instead, standard deviations were calculated from these values. If the required data could not be derived from the published data, as was the case for five of the eight selected studies, we referred to unpublished data obtained from the authors. Incomplete data (e.g., means without standard deviations) could not be used and were excluded from this meta-analysis.
Data Pooling Across Studies
Outcome measures in the eligible studies were reported with use of several different measurement tools. To increase the generalizability of our results where possible, comparable continuous outcome data were pooled across studies with use of the method of standard mean differences with random effects. This model calculates the error term for both within-study and interstudy variability in the meta-analysis. Data from studies using the same measurement tools were plainly pooled. If different measurement tools were used, data were standardized for comparison and were pooled with use of the following methods.
For our primary outcome measure, we pooled comparable scores from different patient-reported functional outcome instruments when these instruments scored disability on a 100-point scale, with 100 representing the least or most disability. If more than one outcome instrument was used, the average outcome was used for analysis. Using the same method, we pooled comparable scores of pain as reported on a 100-point scale, with 0 and 100 representing the least and most pain, respectively.
Range of motion was reported as flexion and extension and/or radial and ulnar deviation as a percentage of the value for the uninjured wrist. When range of motion was reported in more than one plane, outcomes were standardized by calculating the means of different planes, with equal weighting of the different planes. We then calculated the percentage lost in comparison with the uninjured side by subtracting values from 100% (e.g., 13% lost instead of 87% remaining).
Grip strength measurements were reported as a percentage of the value for the uninjured hand in most studies and were reported in kilograms in some studies. When the percentage compared with the uninjured side was provided, we calculated the percentage lost in comparison with the uninjured side by subtracting values from 100% as described above. When only the measurements in kilograms were provided, data were standardized according to the average values from the Guides to the Evaluation of Permanent Impairment26. The derived values of average grip strength were 47 kg for men and 24 kg for women. We then calculated the percentage lost in comparison with the average grip strength values, adjusted for sex.
In addition, three experienced orthopaedic surgeons (J.S.H., D.R., and M.B.) graded complications in terms of severity in order to pool the data across studies for comparison on the basis of patient-important end points27. Three subgroups of complications were created: low, moderate, and severe. Low-grade complications were defined as transient or short-term consequences and included superficial infection. Moderate-grade complications were defined as long-term consequences that did not require additional surgery and included osteoarthritis. Severe-grade complications were defined as (1) long-term consequences that required additional surgery, including screw malposition, hardware removal, malunion, nonunion, and osteonecrosis, and (2) complex regional pain syndrome.
Data Analysis
For both treatment groups in each study, standard mean differences and 95% confidence intervals were calculated for continuous outcomes and risk ratios (i.e., relative risk)28 and 95% confidence intervals were calculated for dichotomous outcomes. Treatment effect was defined as significant if p < 0.05. Heterogeneity between studies was tested with use of both the chi-square test (with significance defined as p < 0.1) and the I2 test (with substantial heterogeneity defined as values of >50%)29.
Source of Funding
Financial support for this study was received from the Netherlands Organisation for Scientific Research (NWO) and the Marti-Keuning-Eckhardt Foundation and the Anna Fonds Foundation. The funding was used for salaries.
Literature Search
The search resulted in 120 potentially eligible studies, eight of which met our inclusion criteria (Fig. 1)13,14,18,30-34. All included studies were randomized controlled trials. In total, 419 patients from eight trials were included. Two hundred and seven patients were treated surgically, and 212 were treated conservatively.
Description of Included Studies
The study characteristics are summarized in a table in the Appendix.
Adolfsson et al. reported on fifty-three patients (fourteen women and thirty-nine men) with a mean age of thirty-one years (range, fifteen to seventy-five years) who had a recent (less than fourteen-day-old) nondisplaced fracture of the waist of the scaphoid (a Herbert35 type-B1 or B2 fracture)30. The authors randomly allocated twenty-five patients to operative treatment (percutaneous screw fixation) and twenty-eight patients to conservative treatment (cast immobilization). Patients were followed for a minimum of sixteen weeks; 26% were lost to follow-up.
Arora et al. randomly allocated forty-seven patients with nondisplaced waist (Herbert type-B2) fractures of the scaphoid to percutaneous scaphoid fixation (twenty-three patients) and cast treatment (twenty-four patients)31. The follow-up group consisted of forty-four patients (twelve women and thirty-two men) with a mean age of thirty-three years (range, twenty to fifty-six years). Patients were followed for a minimum of twenty-four weeks; 6% were lost to follow-up.
Bond et al. reported on twenty-five patients (three women and twenty-two men) with a mean age of twenty-four years (range, eighteen to thirty-four years) who had sustained an acute nondisplaced fracture of the scaphoid waist (a Herbert type-A2 or B2 fracture)14. Eleven patients were randomized to percutaneous screw fixation, and fourteen were randomized to cast immobilization. The mean duration of follow-up for each group was twenty-five months; no patient was lost to follow-up.
Dias et al. reported on eighty-eight patients with a mean age of 29.5 years (range, sixteen to sixty-one years) who had a clear bicortical nondisplaced or minimally displaced fracture of the waist of the scaphoid (a Herbert type-A2, B2, or B5 fracture)32. The study group included nine female and seventy-nine male patients. Forty-four patients were allocated to each treatment group, consisting of internal fixation with a Herbert screw or cast immobilization. Minimal initial fracture displacement was present in association with three fractures in the conservative treatment group and eight fractures in the surgical treatment group. Patients were followed for a minimum of fifty-two weeks; 8% were lost to follow-up.
McQueen et al. randomly allocated sixty consecutive patients with a nondisplaced or minimally displaced fracture of the waist of the scaphoid (a Herbert type-B1 or B2 fracture) to percutaneous screw fixation or immobilization in a cast13. Each group consisted of thirty patients. Minimal initial fracture displacement was present in association with two fractures in the conservative group and five fractures in the surgical group. There were ten female and fifty male patients with a mean age of 29.4 years (range, seventeen to sixty-five years). Patients were followed for a minimum of fifty-two weeks; 8% were lost to follow-up.
Saedén et al. randomized sixty-one patients with sixty-two acute fractures of the scaphoid (AO type-C2 and C3 fractures) to surgical treatment with use of a Herbert screw (thirty-two fractures) and conservative treatment with use of a cast (thirty fractures)33. The mean age at the time of injury was thirty-three years (29 ± 13 years for the operative treatment group and 37 ± 20 for the conservative treatment group). Thirteen fractures were in females, and forty-nine were in males. Patients were followed for a minimum of twelve years; 18% were lost to follow-up.
In 2007, Vinnars et al.34 studied the data for a subgroup of fifty-two patients from the same randomized trial reported by Vinnars et al. in 200818. They included acute nondisplaced fractures of the scaphoid (Herbert type-A2, B1, B2, and B3 fractures). There were sixteen female patients and thirty-six male patients. The median age was thirty-two years for the casting group and twenty-nine years for the surgical treatment group. Both groups included twenty-six patients. Only additional data from this report (i.e., data not reported by Vinnars et al. in 2008) were used for analysis. The mean duration of follow-up was 10.2 years, and no patient was lost to follow-up.
In 2008, Vinnars et al. reported on eighty-five patients (nineteen women and sixty-six men) with a mean age of thirty-two years who had sustained an acute nondisplaced scaphoid fracture (a Herbert type-A2, B1, B2, or B3 fracture)18. Forty-three patients were allocated to receive surgical treatment, and forty-two were allocated to receive conservative treatment. Patients were followed for a mean of 10.2 years; 11% were lost to follow-up.
Outcome Measure Reporting
The reported outcomes of the included studies are summarized in a table in the Appendix.
Four of the eight studies evaluated the primary outcome measure, functional outcome based on validated function scores13,18,31,32. Two studies evaluated the DASH score18,31, including one study that evaluated the DASH and PRWE scores18, one evaluated the PEM score32, and one evaluated the modified Green and O'Brien score13.
One study evaluated patient satisfaction14, and two studies evaluated pain31,32. Seven studies evaluated range of motion and grip strength13,14,18,30-33. Three studies evaluated time to union13,14,31, seven evaluated infection13,14,18,30-33, six evaluated the rate of nonunion14,18,30-33, and seven evaluated the rate of malunion13,14,18,30-33. Six studies evaluated time off work13,14,31-34, three evaluated return to previous activity13,18,30, and two evaluated costs31,34. All studies evaluated complications13,14,18,30-34.
Methodological Quality
The methodological quality of the eligible trials was limited, as shown in a table in the Appendix. Even though all studies were randomized controlled trials, concealment of allocation was unclear in three studies30,31,33, no study was blinded, and outcome assessors were blinded in only one study13. The rate of loss to follow-up, reported in all studies, ranged from 0% to 26%. The authors of four of the eight studies clearly stated that their analysis was based on intention-to-treat principles13,18,32,34. The authors of three studies reported on cross-over to surgical treatment18,32,33; however, Saedén et al.33 did not clearly state whether they performed an intention-to-treat analysis. The quality assessments of all outcome measurements as well as the effect of surgical as compared with conservative treatment are described in Table I.
Functional Outcome
Four13,18,31,32 of the eight selected studies evaluated the primary outcome measure of this meta-analysis: functional outcome based on validated function scores. Their data could be successfully pooled, and the groups were homogeneous. Analysis revealed a significant difference in treatment effect, in favor of surgical treatment (standard mean difference = —0.62, 95% confidence interval = —0.89 to —0.36, p < 0.01, I2 = 5%) (Fig. 2).
Secondary Outcomes
Patient satisfaction was only reported in one study14. Within that study, there was a significant difference in treatment effect, in favor of surgical treatment (standard mean difference = 2.06, 95% confidence interval = 1.06 to 3.07, p < 0.01).
Measurements of pain could be pooled from two studies31,32. Analysis of these data showed significant heterogeneity and no significant differences in the pooled treatment effect (standard mean difference = 0.49, 95% confidence interval = —0.76 to 1.74, p = 0.44, I2 = 91%).
Seven studies evaluated range of motion and grip strength13,14,18,30-33, but the data from one of those studies30 could not be included because no standard deviations were provided. Thus, data for range of motion were pooled across six studies13,14,18,31-33. Analysis of those data revealed no significant difference in pooled treatment effect and considerable heterogeneity (standard mean difference = —0.16, 95% confidence interval = —0.70 to 0.37, p = 0.55, I2 = 81%). Data for grip strength were pooled across the same six studies13,14,18,31-33. Analysis of those data revealed substantial heterogeneity between treatment groups and a significant difference in pooled treatment effect in favor of surgical treatment (standard mean difference = —2.59, 95% confidence interval = —4.24 to —0.94, p < 0.01, I2 = 97%).
There was a significantly shorter time to union in the surgical treatment groups as reported in three studies13,14,31 (standard mean difference = —4.20, 95% confidence interval = —7.69 to —0.70, p = 0.02, I2 = 96%); however, heterogeneity was considerable. The rate of infection was not significantly different across seven studies13,14,18,30-33 (risk ratio = 3.25, 95% confidence interval = 0.35 to 30.34, p = 0.3, I2 = 0%). There were no significant differences in the rates of nonunion across six studies14,18,30-33 (risk ratio = 0.72, 95% confidence interval = 0.18 to 2.86, p = 0.64, I2 = 0%) or malunion across seven studies13,14,18,30-33 (risk ratio = 0.17, 95% confidence interval = 0.02 to 1.39, p = 0.1, I2 = 0%).
With regard to time off work, the data from one study32 could not be included because they were incomplete, even after contacting the authors. The analysis of time off work from five studies13,14,31,33,34 revealed a significant difference in favor of surgical treatment (standard mean difference = —1.69, 95% confidence interval = —2.70 to —0.68, p < 0.01, I2 = 89%) (Fig. 3). With regard to return to previous activity, one study had incomplete data30 and therefore only two studies13,18 could be analyzed. Analysis of those data revealed the groups to be perfectly homogeneous. There was no significant difference in treatment effect (risk ratio = 0.98, 95% confidence interval = 0.91 to 1.05, p = 0.54, I2 = 0%). The data from two studies on the total costs of treatment31,34 were pooled, and the analysis showed no significant difference between treatment groups (standard mean difference = —0.10, 95% confidence interval = —0.50 to 0.30, p = 0.2, I2 = 0%).
Data on complications were reported in all studies13,14,18,30-34. The data were graded in terms of severity in order to be pooled across studies and were analyzed in three subgroups. Low and severe-grade complications were reported in seven studies13,14,18,30-33, and moderate-grade complications were reported in two18,33. The low and moderate-grade subgroups and the combined group were homogeneous; the severe grade showed moderate heterogeneity. Analysis within the low-grade subgroup (risk ratio = 3.25, 95% confidence interval = 0.35 to 30.34, p = 0.30, I2 = 0%) and the severe-grade subgroup (risk ratio = 1.32, 95% confidence interval = 0.35 to 5.01, p = 0.68, I2 =51%) showed no significant difference in pooled treatment effect, and analysis of the moderate-grade subgroup (risk ratio = 3.73, 95% confidence interval = 0.98 to 14.16, p = 0.05, I2 = 42%) was nearly significant in favor of conservative treatment. Meta-analysis of complications for overall treatment effect did not significantly favor one treatment (risk ratio = 1.96, 95% confidence interval = 0.82 to 4.67, p = 0.13, I2 = 41%) (Fig. 4).
Key Findings
According to the best available evidence, this meta-analysis suggests that (1) surgical treatment of nondisplaced and minimally displaced fractures results in significantly better patient-reported functional outcome, greater patient satisfaction, better grip strength, a shorter time to union, and earlier return to work and that (2) there are no significant differences between surgical and conservative treatment with regard to pain; range of motion; the rates of nonunion, malunion, infection, or complications; or total treatment cost. However, the results of this meta-analysis should be interpreted with caution because of the following limitations.
The wide range of follow-up in the primary studies, ranging from sixteen weeks to twelve years, represents an important limitation and a general source of heterogeneity for this meta-analysis. Our primary outcome measurement, functional outcome, generally improves over time, as illustrated in two of the primary studies13,32 by a functional improvement trend between eight and fifty-two weeks of follow-up32. Post-immobilization active functional use of the hand and wrist plays a key role in improving function. Therefore, the fact that conservatively managed patients have had a relatively shorter mobilization time at the time of the latest follow-up evaluation (because of prolonged immobilization times) engenders bias with regard to functional outcome, in favor of the surgical treatment group.
Several secondary outcome measurements showed significant heterogeneity. The six studies pooled for grip strength13,14,18,31-33 showed a large variability, and none of the studies could be identified as causing major heterogeneity. Several authors have reported considerable variability in the relative strength of the two hands in a healthy population, and grip strength has been shown to be typically greater on the dominant side than on the nondominant side36-38. Therefore, the rate of dominant-hand injury, ranging from 34% to 64% across the five studies, could account for some of the grip strength variability among studies. We consider the physiological variability in grip strength within the population, the wide range in the rate of dominant-hand injury, and the wide range of follow-up times (as previously described) to be the main factors in this large variation among studies generating substantial heterogeneity.
Time to union was pooled from three studies13,14,31 and showed substantial heterogeneity. The most plausible reason for this heterogeneity is the fact that time to union represents an inaccurate and imprecise measurement, especially for scaphoid fractures39. Moreover, the diagnosis of scaphoid union was ascertained with use of radiographs, which have poor reliability40. Computed tomography seems to be the most reliable imaging technique for predicting scaphoid fracture union and should therefore be used as a reference standard for healing41. Only one of the included trials31 involved the use of a computed tomography scan as a standard of care to determine scaphoid fracture-healing.
Measurements of time off work were pooled from five studies13,14,31,33,34, and we identified the study by Bond et al.14 as the most important source of heterogeneity. A plausible reason for this was their study population, which consisted solely of full-time military personnel. Time off work in that study was defined as time until the patients returned to a full-duty status, which could most likely not be performed while wearing an arm cast. In other studies, time off work was shorter than the time of cast immobilization in many cases. Thus, many patients (especially non-manual laborers) returned to work prior to cast removal.
With regard to complications, although none of the differences for the subgroups or the combined group were significant, there were still some clinically important differences that should be addressed. Regarding low-grade complications, there were only two cases of postoperative infection among all 207 patients who were allocated to surgical treatment in the seven primary trials. With regard to the risk of osteoarthritis after conservative treatment—traditionally used to support a recommendation for surgery—the meta-analysis of the best available evidence suggests the opposite to be true. According to data from two studies18,33, osteoarthritis develops in 40% of patients after surgery as compared with 10% of patients after cast immobilization, a nearly significant difference (p = 0.05). Regarding severe-grade complications, additional surgery was required in 7.7% of patients in the surgical treatment group as compared with 6.0% of those in the conservative treatment group, and complex regional pain syndrome was reported in four patients in the surgical treatment group and in one patient in the conservative treatment group. In the total study group, with all severity grades combined, complications were substantially more frequent in the surgical treatment group (23.7%) than in the conservative treatment group (9.1%), with osteoarthritis accounting for the largest difference between groups.
With regard to the conclusions from this meta-analysis and the recommendations for clinical decision-making, the above limitations and the low to very low quality of outcome factors (according to the GRADE system) need to be taken into account. Thus, short-term benefits, including faster return to function, seem to be superior following surgical treatment, but these benefits are transient and there is an increased risk for osteoarthritis with surgical treatment, as was concluded in two of the primary studies18,32.
Strengths and Limitations
The strengths of this meta-analysis include a comprehensive literature search for reports in any language, the inclusion of only randomized clinical trials, duplicate data extraction, and duplicate assessment of quality of evidence with use of the GRADE system. Furthermore, all authors were contacted to clarify areas of uncertainty and to provide unreported data, which consisted of unreported standard deviations in most cases. We successfully obtained required raw data from four of the seven groups of authors13,18,31,33,34, making it possible to pool comparable outcome measurements from more studies and therefore to augment the quantity of evidence.
Nonetheless, this meta-analysis had some limitations, including the heterogeneity of secondary outcome measures and low-quality evidence. With regard to substantial heterogeneity in several secondary outcomes, we were able to identify important sources such as fracture type and displacement, study population, and large physiological variations. Even though heterogeneity was substantial in several secondary outcome measurements, we considered the pooling of comparable outcomes and the identification of sources of heterogeneity to be the best option for meta-analysis. However, conclusions from these substantially heterogeneous outcomes should be interpreted with caution.
The meta-analysis showed low-quality to very-low-quality evidence (according to the GRADE system) for treatment effect with any of the selected outcome measurements and, therefore, we can only give "weak" recommendations with regard to the best treatment option27. Required data on our primary outcome measurement, patient-rated functional outcome, was only reported in four of the eight studies. Nevertheless, data were homogeneous and significantly favored surgical treatment. In the GRADE system of rating quality of evidence, the highest quality for outcome is only obtainable if there is no serious methodological limitation. Otherwise, for each limitation, quality of evidence will be downgraded stepwise to moderate, low, or very low.
For randomized controlled trials investigating the effect of surgical and conservative treatment of scaphoid fractures, some limitations seem to be to a great extent inevitable. First, there is no possibility of blinding patients and physicians in any trial comparing surgical and conservative treatment. Second, as both types of treatment have both important advantages and disadvantages, and the benefits and harms of most outcomes are minimal, confidence intervals of nearly every outcome include possible benefits from both types of treatment. This generates a limitation in the form of imprecision of treatment effect.
Some studies in this meta-analysis met the highest possible methodological quality for a surgical trial. Nonetheless, if one or more studies reporting on the same outcome have limited methodological quality, the GRADE system still scores the overall quality of that outcome as low. High-quality evidence within the GRADE framework is possible but difficult to obtain as all safeguards need to be achieved. However, when determining the strength of recommendation with regard to treatment, quality of evidence is one of the four determining factors. One also needs to consider the balance between the desirable and undesirable consequences, the variability in values and preferences, and costs27. Finally, only outcomes that are considered to be critical are the primary factors influencing a given recommendation and should be used to determine the overall quality of evidence supporting the recommendation.
Previous Literature
Traditional arguments in favor of surgical treatment for acute scaphoid fractures include (1) incomplete healing or nonunion in association with conservative treatment35,42, (2) reduced range of motion following cast immobilization35,43, (3) muscle weakness and reduced grip strength35,43 after prolonged immobilization, and (4) osteoarthritis at the time of long-term follow-up43. In addition, the socioeconomic argument of delay in return to work or sports is becoming more important in current algorithms to decide between conservative and surgical treatment of suspected scaphoid fractures as well as other fractures. Recently, Modi et al.44 performed a systematic review of surgical and conservative treatment of scaphoid fractures. The authors concluded that both treatments result in good outcomes and that surgical treatment should be reserved for patients desiring faster return to work or athletics.
Bhandari and Hanson17 and Yin et al.16 previously performed meta-analyses on this subject. Both of those meta-analyses indicated that there was no evidence from randomized trials to determine whether surgical treatment was superior to conservative treatment. However, since the publication of those studies, several randomized controlled trials have been conducted and published13,18,31,34. Furthermore, our literature research and meta-analysis added methodological rigor as we used guidelines from the QUORUM and GRADE working groups. In contrast to previous meta-analyses, we used functional outcome as a primary end point.
Implications for Future Research
All randomized trials that were included in this meta-analysis involved mainly comparable treatment methods for both groups, although outcome measures and measurement instruments differed to a great extent. Only four13,18,31,32 of the seven primary trials involved the use of validated instruments for the evaluation of functional outcome. In current algorithms for decision-making, there is a trend to favor patient-rated functional outcome and return to function over more traditional outcomes such as time to union and measurement of radiographic evidence of fracture union40.
Therefore, future randomized trials should use functional outcomes that are evaluated with use of validated outcome instruments as the primary end point. Substantially heterogeneous outcomes in this systematic review, such as grip strength and range of motion of the injured wrist, should be reported in comparison with the value for uninjured wrist. A large trial of high methodological rigor (e.g., adequate concealment of randomization and blinding of outcome assessors) would be needed to allow conclusions to be drawn regarding the treatment effect on grip strength and range of motion. Finally, we encourage future research groups to report their outcome data with use of means and standard deviations to increase the generalizability of results.
Note: The authors thank Bert Berenschot for performing the literature search and the authors of the primary trials for their cooperation in providing the additional data required for this study. Geert A. Buijze, MD, thanks the Netherlands Organisation for Scientific Research (NWO) and Job N. Doornberg, MD, PhD, thanks the Marti-Keuning-Eckhardt Foundation for supporting this research.