Note: This is one of five reviews that will cover all surgical interventions for treating distal radial fractures in adults. Each review will provide updated evidence for one of the several surgical categories that are presented together in the currently available review (Handoll 2003a). Following publication of the five reviews, Handoll 2003a will be converted to an 'umbrella' review summarising the evidence for surgical treatment for these fractures.
Description of the condition: distal radial fracture in adults
Fractures of the distal radius, often referred to as "wrist fractures", are common in both children and adults. They are usually defined as occurring in the distal radius within three centimetres of the radiocarpal joint, where the lower end of the radius interfaces with two (the lunate and the scaphoid) of the eight bones forming the carpus (the wrist). The majority are closed injuries, the overlying skin remaining intact.
In this review, we consider the treatment of distal radial fracture in adults only, in whom they are one of the most common fractures, predominantly in white and older populations in the developed world (Sahlin 1990; Singer 1998; Van Staa 2001). In women, the incidence of these fractures increases with age, starting at around 40 years. Before this age, the incidence is higher in men (Singer 1998). In contrast, between 60 to 94 years of age, females predominate. A recent multi-centre study in the United Kingdom of patients aged 35 years and above with Colles' fracture (see below) reported an annual incidence of 9/10,000 in men and 37/10,000 in women (O'Neill 2001).
Young adults usually sustain this injury as a result of high-energy trauma, such as a traffic accident. In older adults, especially females, the fracture more often results from low-energy or moderate trauma, such as falling from standing height. This reflects the greater fragility of the bone, resulting from osteoporosis. It has been estimated that, at 50 years of age, a white woman in the USA or Northern Europe has a 15% lifetime risk of a distal radius fracture whereas a man has a lifetime risk of just over two per cent (Cummings 1985). More recent estimates (Van Staa 2001) of lifetime risk of radius or ulna fracture at 50 years of age are similar: 16.6% for women versus 2.9% for men.
Distal radial fractures are usually treated on an outpatient basis. It is estimated that around 20% of patients (mainly older people) require hospital admission (Cummings 1985; O'Neill 2001). This figure includes all people receiving surgery.
Surgeons have classified fractures by anatomical configuration and fracture pattern, to help in their management. Simple classifications were based on clinical appearance and often named after those who described them. In the distal radius, the term "Colles' fracture" is still used for a fracture in which there is an obvious and typical clinical deformity (commonly referred to as a 'dinner fork deformity') - dorsal displacement, dorsal angulation, dorsal comminution (fragmentation), and radial shortening. The introduction of X-rays and other imaging methods made it clear that the characteristic deformity may be associated with a range of different fracture patterns, which may be important determinants of outcome, and therefore the way in which treatment is conducted. For example, the fracture through the distal radius may be extra-articular (leaving the articular surface of the radius intact) or intra-articular (the articular surface is disrupted, sometimes in a complex manner). Numerous classifications have been devised to define and group different fracture patterns (Chitnavis 1999). One of the most commonly used is that of Frykman which distinguishes between extra-articular fractures and intra-articular fractures of the radiocarpal and radio-ulnar joints, and the presence or absence of an associated distal ulnar (ulnar styloid) fracture (Frykman 1967). Another commonly used system is the AO (Arbeitsgemeinschaft fur Osteosynthesefragen) system (Muller 1991) which divides the fractures into three major groups: group A (extra-articular), group B (simple/partial intra-articular), and group C (complex/complete intra-articular). These three groups are then subdivided, yielding 27 different fracture types. Other classification systems have attempted to link fracture type more directly with fracture management. For instance, Cooney 1993 proposed a 'Universal Classification' based on fracture displacement, articular involvement, reducibility (whether the fracture can be reduced; that is whether the bone fragments can be put back in place) and stability (whether, once reduced, the fragments will remain so).
Description of the intervention: external fixation
In the last century, most distal radial fractures in adults were treated conservatively, by reduction of the fracture when displaced, and stabilisation in a plaster cast or other external brace. The results of such treatment, particularly in older people with bones weakened by osteoporosis, are not consistently satisfactory (Handoll 2003b). This has resulted in attempts to develop other strategies involving surgery aimed at more accurate reduction and more reliable stabilisation.
One such strategy is external fixation (Capo 2006; Fernandez 1999; Pennig 1996). Typically this is a closed, minimally invasive method where, in contrast to open surgery, the fractured bone ends are not exposed to direct view. Metal pins or screws are driven into bone, generally via small incisions of the skin and after drilling, on either side of the fracture. These pins are then fixed externally, such as by incorporation into a plaster cast or securing into an external fixator frame. The external component stabilises or 'fixes' the reduced fracture. Fracture reduction or alignment of the bony fragments is generally achieved by closed means, often in the process of applying external fixation. Reduction may be assisted by the application of a percutaneously (through the skin) inserted wire as a 'joy stick' to move the bony fragments back into place. There is considerable variety in the techniques (such as for pin insertion and placement) and devices used for external fixation. Some devices are 'non-bridging' (of the wrist joint) in that the distal pins are placed in the distal radial fragment. In 'bridging fixators', the distal pins are placed in one or more metacarpal bones. Some fixators are linear or uniplanar, whereas others are multiplanar. In addition, some bridging fixators have an articulation (e.g. a ball joint) that allows limited wrist movement. The duration and extent of immobilisation with external fixation also vary. In some cases, external fixation may be augmented by additional methods of fracture fixation. In this review, we include only trials using supplementary percutaneous pinning. This involves extra pins or wires being inserted through the skin and used to fix or support distal radial fragments.
Complications from this injury are diverse and frequent (Altissimi 1986; Atkins 1989; Cooney 1980). Some are associated with the injury itself. As well as concomitant injuries to soft tissues, fracture displacement can further compromise blood vessels, tendons and nerves, with median nerve dysfunction being the most common complication (Belsole 1993). Late complications include midcarpal instability and post-traumatic arthritis which can occur several months or years after injury (Knirk 1986; Taleisnik 1984).
Complications can also result from treatment interventions and include residual finger stiffness, which may be due to faulty application of plaster casts (Gartland 1951), pin track infection, soft tissue injury including tendon rupture, and additional fractures. Complex regional pain syndrome type 1, often termed reflex sympathetic dystrophy (RSD), but also referred to as algodystrophy, Sudeck's atrophy and sometimes shoulder-hand syndrome (Fernandez 1996) is a major complication (Atkins 2003) requiring many months of physiotherapy to alleviate symptoms (pain and tenderness, impairment of joint mobility, swelling, dystrophy, vasomotor instability) in serious cases. The pathology of RSD remains unclear.
Why it is important to do this review
External fixation is one of the main methods for surgical fixation of distal radial fractures. A key question is whether it produces superior results, particularly in terms of function, to conservative treatment. The answer to this question is likely to depend particularly on fracture configuration and bone quality. The issue of what is the best method of external fixation will be addressed in a separate review (Handoll 2007).
We aimed to evaluate the evidence from randomised controlled trials for the use of external fixation for fractures of the distal radius in skeletally mature people.
We compared the relative effects (benefits and harms) of any method of external fixation versus conservative treatment involving plaster or brace use in adults with these injuries. Also considered was augmented external fixation where supplementary percutaneous (through the skin) pinning was used to fix or support distal radial fragments.
We considered these effects primarily in terms of patient-assessed functional outcome and satisfaction, and other measures of function and impairment, pain and discomfort, the incidence of complications, anatomical deformity and use of resources.
Our plan to study the outcomes in different age groups and for different fracture types, especially whether they are extra-articular or intra-articular, was thwarted by the variation in the trial characteristics.
Criteria for considering studies for this review
Types of studies
Any randomised or quasi-randomised (method of allocating participants to a treatment which is not strictly random e.g. by date of birth, hospital record number, alternation) controlled clinical trial comparing external fixation with conservative methods for treating distal radial fractures in adults was considered.
Types of participants
Included were patients of either sex who have completed skeletal growth, with a fracture of the distal radius. External fixation may be considered as primary treatment or take place after the failure of initial conservative management, generally within two to three weeks. Augmented external fixation in the form of supplementary percutaneous pinning was also included. Trials with a mixed population of adults and children were included provided the proportion of children was clearly small (< 5%); otherwise they would have been excluded unless separate data for adults were obtained. We considered it unlikely that we would find trials comparing external fixation with conservative treatment for rarer fracture patterns such as the Barton's fractures (Smith 1988) that are inherently unstable and generally considered not to be amenable to conservative or external fixation. Nonetheless, if found, trials with these types of fractures would have been considered for inclusion, and separate subgroup data sought for mixed fracture populations.
Types of interventions
Randomised comparisons of surgical interventions involving external fixation by itself or with supplementary percutaneous pinning versus conservative interventions such as plaster cast immobilisation.
We excluded trials comparing different methods, including techniques and devices, of external fixation; or trials comparing external fixation with other methods of surgical fixation, such as percutaneous pinning, or trials evaluating the use of supplementary methods, such as bone grafts and substitutes, other than percutaneous pinning, to external fixation compared with conservative treatment. These comparisons will be covered in other reviews, including one covering the use of bone grafts and substitutes.
Types of outcome measures
Our primary outcome of choice would be the number of people with an uncomplicated and speedy restoration of a pain-free fully-functioning wrist and arm with acceptable anatomic restoration and appearance. However, compatible with the general assessment and presentation of outcome within the orthopaedic literature, we report outcome in the following four categories.
(1) Functional outcome and impairment
- Patient functional assessment instruments such as Short Form-36 (SF-36), the Disability of the Arm, Shoulder, and Hand questionnaire (DASH) and the Patient-Rated Wrist Evaluation (PRWE) (MacDermid 2000)
- Return to previous occupation, including work, and activities of daily living
- Grip strength
- Range of movement (wrist and forearm mobility): range of movement for the wrist is described in terms of six parameters: flexion (ability to bend the wrist downwards) and extension (or upwards); radial deviation (ability to bend the wrist sideways on the thumb side) and ulnar deviation (on the little finger side); and pronation (ability to turn the hand so that the palm faces downwards) and supination (palm faces upwards)
(2) Clinical outcome
- Residual soft tissue swelling
- Early and late complications associated with distal radial fractures or their treatment, including reflex sympathetic dystrophy (RSD) and post traumatic osteoarthritis
- Cosmetic appearance
- Patient satisfaction with treatment
(3) Anatomical outcome (anatomical restoration and residual deformity)
- Radiological parameters include radial length or shortening and shift, dorsal angulation, radial inclination or angle, ulnar variance, and for intra-articular fractures: step off and gap deformity of the articular surface (Fernandez 1996; Kreder 1996a). Composite measures include malunion and total radiological deformity. Definitions of four of the most commonly reported radiological parameters are presented in Table 1.
(4) Resource use
- Hospital stay, number of outpatient attendances and other costs.
Timing of outcome assessment
Results were usually collected for the final follow-up time for which these are available. However, we also noted interim results where a marked and important difference in the timing of recovery had occurred.
Search methods for identification of studies
We searched the Cochrane Bone, Joint and Muscle Trauma Group Specialised Register (September 2006), the Cochrane Central Register of Controlled Trials (in The Cochrane Library 2006, Issue 3) (see Appendix 1), MEDLINE (1966 to September week 1 2006), EMBASE (1988 to 2006 week 36), CINAHL (1982 to September week 2 2006). No language restrictions were applied.
Similar search strategies were used for EMBASE (OVID-WEB) and CINAHL (OVID-WEB) (see Appendix 3).
We also searched Current Controlled Trials at www.controlled-trials.com (accessed September 2006) and the UK National Research Register at www.update-software.com/national/ (up to Issue 3, 2006) for ongoing and recently completed trials.
Searching other resources
We searched the reference list of articles. We also included the findings from handsearches of the British Volume of the Journal of Bone and Joint Surgery supplements (1996 onwards) and abstracts of the American Society for Surgery of the Hand annual meetings (2000 to 2006: www.assh.org/), the American Orthopaedic Trauma Association annual meetings (1996 to 2005: http://www.hwbf.org/ota/am/) and American Academy of Orthopaedic Surgeons annual meeting (2004 to 2006: http://www.aaos.org/education/anmeet/libscip.asp). We also included handsearch results from the final programmes of SICOT (1996 & 1999) and SICOT/SIROT (2003), and the British Orthopaedic Association Congress (2000, 2001, 2002 and 2003), and various issues of Orthopaedic Transactions and Acta Orthopaedica Scandinavica Supplementum.
We also scrutinised weekly downloads of "Fracture" articles in new issues of 17 journals (Acta Orthop Scand; Am J Orthop; Arch Orthop Trauma Surg; Clin J Sport Med; Clin Orthop; Emerg Med Clin North Am; Foot Ankle Int; Injury; J Accid Emerg Med; J Am Acad Orthop Surg; J Arthroplasty; J Bone Joint Surg Am; J Bone Joint Surg Br; J Foot Ankle Surg; J Orthop Trauma; J Trauma; Orthopedics) from AMEDEO (www.amedeo.com).
Data collection and analysis
Selection of studies
All review authors independently assessed potentially eligible trials for inclusion using a pre-piloted form. Any disagreement was resolved by discussion.
Data extraction and management
Using a data extraction form, two of the review authors (HH and JH) independently extracted trial details and data for new trials, and one author (HH) repeated data extraction of trials already included in Handoll 2003a and checked for consistency with her previous data extraction. HH entered the data into RevMan. Any disagreements were resolved by discussion. When appropriate, extraction of results from graphs in trial reports was performed where data were not provided in the text or tables. We contacted trialists of trials not reported in full journal publications for additional information or data. Contact with other trial authors was dictated by the vintage of the publication, a general impression of the expected gain, and anticipated or known difficulty in locating trial authors.
Results were collected for the final follow-up time for which these were available. We also noted instances where clinically important differences had been reported at intermediate follow-up assessments.
Assessment of methodological quality
In this review, risk of bias is implicitly assessed in terms of methodological quality.
Two of the review authors (HH and JH) independently assessed methodological quality of the newly included trials using a pre-piloted form. One author (HH) repeated her assessment of the trials already included in Handoll 2003a. All disagreements were resolved by discussion. Titles of journals, names of authors or supporting institutions were not masked at any stage. A modification of the quality assessment tool used in the current 'umbrella' review was used. Instead of scores, each item was graded based on whether the quality criterion was met: 'Y' (met), '?' (possibly or only partially met) or 'N' (not met). The rating scheme covering 11 aspects of trial validity plus brief notes of coding guidelines for selected items are given in Table 2.
Measures of treatment effect
Where available, quantitative data, both dichotomous and continuous, for the outcome measures listed above (see 'Types of outcome measures') are presented in the analyses. Relative risks and 95% confidence intervals were calculated for dichotomous outcomes and mean differences and 95% confidence intervals calculated for continuous outcomes.
Unit of analysis issues
No unit of analysis issues arose in this review.
Dealing with missing data
Where appropriate, we have performed intention-to-treat analyses to include all people randomised to the intervention groups. To a limited extent, we have investigated the effect of drop outs and exclusions by conducting worse and best scenario analyses. We were alert to the potential mislabeling or non identification of standard errors for standard deviations. Unless missing standard deviations could be derived from confidence interval data, we did not assume values in order to present these in the analyses.
Assessment of heterogeneity
Heterogeneity was assessed by visual inspection of the forest plot (analysis) along with consideration of the test for heterogeneity and the I² statistic (Higgins 2003).
Assessment of reporting biases
There were insufficient data to assess publication bias; for example, by preparing a funnel plot.
Data synthesis (meta-analysis)
When considered appropriate, results of comparable groups of trials were pooled. Initially we used the fixed-effect model and 95% confidence intervals. We also considered using the random-effects model, especially where there was unexplained heterogeneity.
Subgroup analysis and investigation of heterogeneity
There were no data available to carry out our pre-specified subgroup analyses by age, gender and type of fracture (primarily, extra-articular versus intra-articular fractures). Presentation in separate subgroups was also considered where there was a fundamental difference in the method of external fixation (such as bridging versus non-bridging external fixation; the use of supplementary percutaneous pinning), but considered inappropriate. To test whether the subgroups are statistically significantly different from one another, we tested the interaction using the technique outlined by Altman and Bland (Altman 2003).
There were no data available to carry out our pre-specified sensitivity analyses examining various aspects of trial and review methodology, including the study quality (specifically allocation concealment, outcome assessor blinding and reportage of surgical experience).
Interpretation of the evidence
We graded the findings of the treatment comparison(s) according to the six categories of effectiveness used by contributors to Clinical Evidence (BMJ 2006) (see Table 3) to assist our interpretation of the evidence.
Description of studies
Results of the search
The search for trials predated the development of this review, which is essentially a reworked update of part of a previously published review (Handoll 2003a) covering all surgical intervention for these fractures. We have not documented the numbers of references retrieved by electronic searches; updates of MEDLINE, EMBASE and CINAHL are now generated on a weekly basis. Of 21 potentially eligible studies put forward for study selection, 15 were included, five were excluded and one is in 'Studies awaiting assessment'.
Twelve of the included trials were previously included in Handoll 2003a; this includes Young 2003, whose study ID has been changed to reflect the identification of a full report. The three additional trials (Hegeman 2004; Kreder 2006; Zheng 2003) have been included in the current review; the Kreder trial was listed as ongoing in Handoll 2003a.
All of the included studies were fully reported in medical journals. Six of the included trials were initially located by handsearching. The rest were located in the following ways: The Cochrane Bone, Joint and Muscle Trauma Group Specialist Register (2); MEDLINE (5); PubMed (1) and bibliography checking (1). A translation from Chinese was obtained for Zheng 2003.
Details of the methods, participants, interventions and outcome measures of individual trials are provided in the 'Characteristics of included studies' table.
The publication dates of the main reports of these trials span 18 years; Pring 1988 being the earliest. Each trial took place in one of nine countries: Canada (1 trial), China (1), India (1), Israel (1), New Zealand (1), Spain (1), Sweden (2), The Netherlands (2), and the UK (5). All were all single-centre studies apart from Kreder 2006, which had three centres. Most were conducted in teaching hospitals.
Age and gender
The 15 included trials involved a total of 1022 participants. Five trials (Horne 1990; Howard 1989; Jenkins 1989; Roumen 1991; Stein 1990) provided no information on the gender composition of their study populations. For the rest, the percentage of females ranged from 17% (Rodriguez-Merchan 92) to 91% (Hegeman 2004). The mean ages of the trial populations ranged from 36 years (Rodriguez-Merchan 92) to 72 years (Horne 1990). Aside from restricting trial participation to people over 55 years, Roumen 1991 provided no information on age. Only the age range (18 to 52 years) of the study population was provided for Zheng 2003. It is clear that the vast majority of participants in the included trials were skeletally mature: in particular, three trials (Kapoor 2000; Kreder 2006; Young 2003) excluded children either by explicitly setting a minimum age limit or stipulating that the participants should be adult or skeletally mature. Four trials (Hegeman 2004; Horne 1990; Lagerstrom 1999; Roumen 1991) further restricted the trial population to more mature adults: above 55, 60, 45 and 55 years respectively. Upper age limits were applied in eight trials (Abbaszadegan 1990: 75 years; Hegeman 2004: 80 years; Howard 1989: 75 years; Jenkins 1989: 65 years; Kreder 2006: 75 years; Lagerstrom 1999: 75 years; Rodriguez-Merchan 92: 45 years; Young 2003: 75 years).
Overall the trials in terms of age and gender of their study populations appear to fall into three categories:
- Seven trials (Abbaszadegan 1990; Howard 1989; Kreder 2006; McQueen 1996; Pring 1988; Stein 1990; Young 2003) included a general adult population, albeit some with an upper age limit (generally 75 years). The majority of participants were women in five of these trials; no data on gender mix were provided for Howard 1989 and Stein 1990.
Types of fractures
Two trials (McQueen 1996; Roumen 1991) recruited people whose fractures had redisplaced by two weeks, whereas the other 13 trials involved primary treatment of people with acute fractures. No trial referred to people with open fractures and it is likely that all fractures were closed; this was explicit in Kreder 2006 and Zheng 2003.
Several descriptions were used to define the types of fracture in the included trials. Fourteen trials gave explicit reference to the inclusion of dorsally displaced or Colles' fractures (including the use of Older's classification: Older 1965) or presented radiological results indicating that fractures were dorsally displaced. Kapoor 2000 included both dorsally and volarly displaced intra-articular fractures. Six trials (Horne 1990; Jenkins 1989; Kapoor 2000; Pring 1988; Rodriguez-Merchan 92; Zheng 2003) provided no criteria of the extent of the displacement required for trial entry. Kreder 2006 set an upper limit of the extent of dorsal angulation and metaphyseal comminution as well as stipulating joint congruity; this reflected the inclusion criteria of another concurrent trial involving more displaced fractures (Kreder 2005). Only intra-articular fractures were included in six trials (Hegeman 2004; Kapoor 2000; Lagerstrom 1999; Rodriguez-Merchan 92; Roumen 1991; Stein 1990). Both intra-articular and extra-articular fractures were clearly included in the remaining trials except Howard 1989; however, some, if not most, of the severely displaced and comminuted Colles' fractures of this trial were intra-articular. Of the 13 trials applying or reporting fracture type according to an established classification system, two trials grouped fractures according to more than one classification system. Seven trials referred to the Frykman system, four trials to the AO system and two trials to Older's system. Zheng 2003 categorised their fractures according to the universal system (Cooney 1993), which also includes whether the fractures were reducible.
All 15 trials compared external fixation with plaster cast immobilisation. There was, however, considerable variety in the devices used for external fixation and in surgical techniques such as methods of reduction, pin insertion and pin placement. The duration and extent of immobilisation in both the external fixation and conservative treatment groups also varied among trials. One trial (Pring 1988) used pins incorporated into plaster. One trial (Jenkins 1989) applied a (multiplanar) fixator that did not bridge the wrist joint. Linear or uniplanar fixators which bridged the wrist joint were clearly used in eight trials (Abbaszadegan 1990; Horne 1990; Lagerstrom 1999; McQueen 1996; Rodriguez-Merchan 92; Stein 1990; Young 2003; Zheng 2003) and non-linear or multi-planar fixators used in three trials (Howard 1989; Kapoor 2000; Roumen 1991). The configurations of the external fixator frames used in Hegeman 2004 and Kreder 2006 were not clear, but were probably uniplanar. Explicit use of optional percutaneous K-wire fixation to stabilise the fracture was referred to in three trials (Kreder 2006; Rodriguez-Merchan 92; Zheng 2003). In Jenkins 1989, the distal wires were orientated so that they transfixed comminuted fractures (such as die punch fractures) of the distal fragment before the external fixator was assembled.
A concise summary of the participants, fracture type, timing and details of the interventions for the 15 trials is given in Table 4.
Five studies were excluded for reasons stated in the 'Characteristics of excluded studies' table. Four studies were found not to be randomised trials and there was insufficient information on one quasi-randomised trial (Sprenger 1988), published only as an abstract. (The latter trial appeared as an included trial in Handoll 2003a.)
No ongoing studies were identified.
Studies awaiting assessment
The possible inclusion of Moroni 2004 requires a response from the lead author concerning the numbers allocated to each group.
Risk of bias in included studies
The quality of trial methodology, judged using the 11 quality criteria listed in Table 2, is disappointing. Associated with this is a high potential for the key systematic biases (selection, performance, assessment and attrition) leading to questions about internal validity, and issues of clinical relevance and applicability or external validity. These will be considered further in the 'Discussion'. The results, together with some notes on specific aspects, of the quality assessment for the individual trials are shown in Table 5. Information specific to the first three items of the quality assessment is given in the methods sections of the 'Characteristics of included studies' table. A summary of the results for individual items of quality assessment is given below.
Allocation concealment (item 1)
Only one trial (Kreder 2006), which used sealed opaque sequentially marked envelopes, was considered to have satisfied the criteria for secure allocation concealment. It was unclear whether allocation was concealed prior to randomisation in 11 trials. Three of these used envelopes (Hegeman 2004; McQueen 1996; Young 2003), one used a computer (Lagerstrom 1999) and seven trials provided no information (Abbaszadegan 1990; Horne 1990; Howard 1989; Kapoor 2000; Pring 1988; Rodriguez-Merchan 92; Roumen 1991). The three remaining trials used quasi-randomised methods based on dates of birth (Jenkins 1989), day of hospital admission (Stein 1990) and admission number (Zheng 2003).
Intention-to-treat analysis (item 2)
Clear statements of participant flow with evidence of intention-to-treat analysis were available for six trials (Abbaszadegan 1990; Hegeman 2004; Howard 1989; Kreder 2006; McQueen 1996; Young 2003). Failure to provide the numbers assigned to each group at randomisation, post-randomisation exclusions, inappropriate analyses and lack of accountability of losses were reasons for a 'N' rating for four trials (Horne 1990; Kapoor 2000; Pring 1988; Roumen 1991).
Blinding of outcome assessors (item 3)
No trial reported blind outcome assessment although four trials (Howard 1989; Kreder 2006; Pring 1988; Young 2003) referred to independent assessors for some outcomes. Total blinding of outcome assessment is impractical for trials testing surgical interventions but it is possible for some outcomes and more so at longer-term follow up.
Comparability of baseline characteristics (item 4)
Two trials (Hegeman 2004; Kreder 2006) provided sufficient information indicating the similarity in the baseline characteristics of gender, age and type of fracture. Potentially important imbalances between treatment groups in participant characteristics, where information was provided, were found in Jenkins 1989 (age), Lagerstrom 1999 (gender) and Zheng 2003 (gender).
Blinding of patients and treatment providers (items 5 and 6)
These are unlikely in these studies and none was claimed.
Care programme comparability (item 7)
We found it difficult to confirm comparability of care programmes other than the trial interventions, although we judged it highly likely in two trials (Howard 1989; Lagerstrom 1999). Systematic differences between the arms of a trial, such as the different methods of anaesthesia for closed reduction in Kreder 2006, can change the actual comparison under test.
Description of inclusion criteria (item 8)
Ten trials provided sufficient trial inclusion and exclusion criteria to define their study populations. This item was rated 'N' ("not met") in Pring 1988 where no details of the type of fractures included were available aside from "Colles'".
Definition of outcome measures and quality of outcome measurement (items 9 and 10)
The definition of outcome measurement was clear enough to give a good idea of what was recorded in all of the included trials except Stein 1990. Two trials (Horne 1990; Stein 1990) were considered to have inadequate outcome measurement, which included follow up of variable duration. Only Kreder 2006 was rated as having 'optimal' quality outcome measurement, which included use of validated patient assessed quality of life instruments and active follow up. Nonetheless, several other trials (including Hegeman 2004; McQueen 1996; Young 2003) had active follow up and applied clearly relevant measures of function. Of note is the grading or scoring of overall functional outcome according to non-validated scoring systems in several trials. These systems, which often included anatomical and clinical outcomes, included modifications of other scoring systems such as that of Gartland and Werley (Gartland 1951). Also, noted but not rated, were instances where adjustments were made for hand dominance. The variety of schemes used and other outcome measures reported by the trials is evident from inspection of the 'Characteristics of included studies' table.
Length of follow up (item 11)
Follow up ranged from a minimum of four months (Horne 1990) to 10 years (Howard 1989). Follow up of variable duration at times where participants are at different stages of recovery, may be a potential cause of bias in Horne 1990 (4 to 15 months) and Stein 1990 (6 months to 4 years).
Loss to follow up (not rated)
Loss to follow up was substantial in several trials (see 'Characteristics of included studies' table and Table 5). Nearly a third of participants were missing from the final analyses of Kapoor 2000 and Young 2003. The latter trial, however, gave a full account of the losses at seven years whereas losses were not explained in Kapoor 2000. For some of the trials appearing to have no losses, it may be the case that these were not reported.
Effects of interventions
In the following, the results are presented for the basic comparison, namely external fixation versus plaster cast immobilisation, despite the evident variation among the trials in patient characteristics and interventions (see Table 4), and in the methods and timing of outcome assessment and selection of reported outcomes (see below and 'Characteristics of included studies' table). Furthermore, few of the trials are sufficiently similar to fall neatly into specific sub-categories. For instance, the pooling of data from the two trials studying treatment of redisplaced fractures in older people (McQueen 1996; Roumen 1991) would appear valid but there remains clear dissimilarity in the known characteristics of the trials (e.g. different types of external fixator; the re-reduction of fractures in the conservative treatment group and the inclusion of extra-articular fractures in McQueen 1996 but not in Roumen 1991). Hence, whilst some exploratory analysis has taken place, we considered that, on the whole, subgroup analysis was not appropriate. Separate data for participants in the three trials with optional supplementary pinning were not available.
Functional outcome and impairment
Final overall functional outcome, or some aspect of function, was considered better in the external fixation group in eight studies (Abbaszadegan 1990; Howard 1989; Jenkins 1989; Kapoor 2000; Pring 1988; Rodriguez-Merchan 92; Stein 1990; Zheng 2003). No statistically significant differences in final functional outcome were reported by seven studies (Horne 1990; Kreder 2006; Lagerstrom 1999; McQueen 1996; Pring 1988; Roumen 1991; Young 2003). Though some statistically significant differences in some aspects of function were apparent at one year follow up in Young 2003, the trialists considered that the differences were not clinically significant; nonetheless we have presented these as well as the seven year results in the analyses.
'Functional' scoring systems were used in 13 of the trials (Abbaszadegan 1990; Hegeman 2004; Horne 1990; Howard 1989; Jenkins 1989; Kapoor 2000; Kreder 2006; Pring 1988; Rodriguez-Merchan 92; Roumen 1991; Stein 1990; Young 2003; Zheng 2003); most were modifications of Gartland and Werley's scheme (Gartland 1951). Some consideration of deformity was present in the schemes used by Abbaszadegan 1990, Hegeman 2004, Kapoor 2000, Pring 1988 and Roumen 1991; whilst various complications were included in that used by six trials (Hegeman 2004; Horne 1990; Howard 1989; Kapoor 2000; Young 2003; Rodriguez-Merchan 92). The functional scores were usually graded into four categories ('excellent', 'good', 'fair' and 'poor'). However, some trials reported results for combined categories (such as fair or poor) only. Jenkins 1989 and Stein 1990 presented results split by subjective and objective gradings. The subjective end gradings from Jenkins 1989 and Stein 1990 were pooled with the overall functional gradings from the other studies in Analyses 01.01, 01.02 and 0.03. A higher proportion of people treated with external fixation achieved an excellent functional grading; in other words, a lower proportion of people treated with external fixation had a "not excellent" grading (see Analysis 01.01: 134/256 versus 166/265; RR 0.82, 95% CI 0.71 to 0.95). Analysis 01.02 shows a worst-case (for conservative treatment) and then a best-case analysis for this outcome. The result strongly favours external fixation when it is assumed that all participants lost to follow up or excluded (not possible for Horne 1990) in the external fixation group had an excellent result compared with none of those lost or excluded from the conservative treatment group. In the converse case, a neutral result is obtained (RR 1.09, 95% CI 0.95 to 1.25). More people in the external fixator group achieved excellent grades at 10 years in Howard 1989 but the data presented in the two abstract reports for long-term follow up were inconsistent. As shown in Analysis 01.03, the numbers with an unsatisfactory outcome (either fair or poor gradings) were statistically significantly fewer in the external fixation group (62/308 versus 83/304; RR 0.73; 95% CI 0.55 to 0.98; fixed-effect model). This was not statistically significant when using the random-effects model (RR 0.72; 95% CI 0.47 to 1.09: analysis not shown). The significant heterogeneity (P = 0.07; I² = 41.4%) is lost on the removal of the results of Roumen 1991 and the result is again statistically significant (RR 0.65, 95% CI 0.47 to 0.88; analysis not shown). In summary, these pooled results from non-validated 'functional' scoring schemes favour external fixation in terms of more people with excellent results (depicting a good recovery) and fewer people with an unsatisfactory outcome (either 'fair' or 'poor' result). However, as shown by the results of sensitivity analyses (e.g. Analysis 01.02) such findings are not robust.
Analysis 01.04 presents the separate subjective and objective results reported in Jenkins 1989 and Stein 1990: here it is notable that the trial participants of Jenkins 1989 rated the end functional result less favourably than the result using objective measures. Kreder 2006 found non-significant trends to better functional results for the surgical group in the upper extremity function part of the Musculoskeletal Function Assessment tool (see Analysis 01.05), and in the results for the bodily pain domain of the SF-36 and Jebsen-Taylor (Jebsen 1969) hand function scores (data reported as standard deviations from the norm for these outcomes are not presented in the Analyses). The mean 'demerit' Scheck score (Scheck 1962) for function was better, but not statistically significantly so, in the external fixation group of Pring 1988 at six months (3.30 versus 4.19).
Activities of daily living
At three months, more people in the external fixation group of Hegeman 2004 had problems with some activities of daily living; this was significant for fine hand co-ordination (see Analysis 01.06). None of the differences were significant at one year for Hegeman 2004. Young 2003 found similar numbers of people in the two groups experiencing difficulties with specific functional tasks at seven years (see Analysis 01.06).
There was no significant differences between the two groups of Kreder 2006 in the numbers of people who changed their job because of their injury at six months, one year or two years (see Analysis 01.07). The time to return to work or normal activities averaged 70 versus 75 days in Young 2003 (statistical significance not reported).
The greater grip strength in the external fixation groups of Jenkins 1989 and Young 2003 (1 year data) was statistically significant (see Analysis 01.08). Kapoor 2000 found a marginally statistically significantly better restoration of grip strength in the external fixation group (70% versus 63% of unaffected arm; reported P = 0.05). The differences between the two groups in mass grip strength were found not to be statistically significant at final follow up in seven trials: Abbaszadegan 1990, Hegeman 2004 (see Analysis 01.08), Kreder 2006 (see Analysis 01.09); Lagerstrom 1999 (see Analysis 01.10), McQueen 1996 (see Analysis 01.08), Pring 1988 (mass grip strength as % of normal side: 67.6% versus 63.8%) and Young 2003 (7 year results: see Analysis 01.08). Due to the statistically significant heterogeneity (P = 0.004; I² = 77.6%) in the results of the four trials (Hegeman 2004; Jenkins 1989; McQueen 1996; Young 2003) providing data for grip strength, these were not pooled. Lagerstrom 1999, who measured isometric grip strength, found that the recovery in grip strength was slower in the external fixator group and considered there was some indication for specific physiotherapy for patients treated by external fixation. One person in the external fixator group in Young 2003 at seven years had persistent wrist weakness necessitating a wrist splint.
Pain was reported as being significantly less in the external fixator group in Abbaszadegan 1990 (final mean visual analogue score (0 to 10 severe): 0 versus 1; reported P < 0.002) but numbers with persistent pain were reported to be similar in Hegeman 2004, Lagerstrom 1999 (19% overall had persistent pain after two years; separate data by intervention group were not available) and Young 2003 (see Analysis 01.11). Roumen 1991 reported the numbers of patients experiencing pain: at rest, during movement and ulnar compression pain. There were no statistically significant differences for any of these three measures (see Analysis 01.12). Young 2003 also gave details of the circumstances for persistent pain at seven years; notably the constant pain on movement of the wrist in two people in the external fixation group was sufficient for them to adapt their activities to avoid using the wrist.
Range of movement
The available data for range of movement are presented in Analyses 01.13 to 01.16. The difference in the flexion/extension arc values were reported as being statistically significant, in favour of the external fixator group, in Abbaszadegan 1990 (% of normal wrist: 95% versus 83%; reported P = 0.0002), but not in McQueen 1996 (see Analysis 01.13). McQueen 1996 also found no difference in the overall range of motion. Only wrist extension was better in the external fixation group at one year in Hegeman 2004. There were no statistically significant differences in range of motion outcomes at one year for Jenkins 1989 (see Analysis 01.13). There was no indication in Kapoor 2000 as to whether the superior range of motion in the external fixator group was reflected in statistically significant differences in the losses in the following three parameters: dorsi-palmar flexion (19 versus 37 degrees); radial-ulnar deviation (13 versus 16 degrees); pronation-supination (23 versus 40 degrees). Kreder 2006 reported there was no statistically significant differences in wrist and forearm range of motion at any of the follow ups, on the application of the Bonferroni correction for multiple comparison testing. The clinical implications of the significant findings, when not using the Bonferroni correction, for Kreder 2006 at two years in favour of external fixation in flexion and radial deviation, and in favour of plaster cast for supination, are not known but are likely to be minor (see Analysis 01.14). Though statistically significant, the differences between the two groups in ulnar deviation and pronation at one year follow up in Young 2003 are not clinically important (see Analysis 01.15); there were no statistically significant differences in range of motion outcomes at seven year follow up of this trial (see Analysis 01.16).
Clinical outcome and complications
The various complications suffered by the participants of these trials are presented in Analysis 01.17. Null events have also been entered when reported. Redisplacement resulting in secondary treatment mainly occurred in the conservative treatment groups, the exceptions being three participants whose external fixator was distracted to combat radial shortening in Howard 1989, one participant given open reduction and internal fixation in Kreder 2006 and three participants who underwent remanipulation in Young 2003. Similarly, redisplacement or, as in McQueen 1996, recurrent instability of already redisplaced fractures were significantly more common in conservatively treated participants.
Pin-site infections in 10 of the 69 people with pin-track infections (61/444 versus 1/402; RR 12.02, 95% CI 5.07 to 28.49) did not resolve with cleansing and antibiotics or, in Kreder 2006, with the early removal of supplementary percutaneous pins. Of the three serious infections in Jenkins 1989, two infections were recurrent and resulted in fixator removal and one developed into the only reported case of osteomyelitis. Surgical curettage of the pin tracks was required for two people in Kreder 2006. Two infections resulted in fixator removal in McQueen 1996, and the third required curettage. The two infections in Stein 1990 resolved on fixator removal; it is not clear whether this was premature. Some reference to care of pin sites was made in six trials (Howard 1989; Jenkins 1989; Kapoor 2000; Kreder 2006; McQueen 1996; Rodriguez-Merchan 92). There are 10 other cases of pin loosening or other pin site problems (see Analysis 01.17). In addition, Pring 1988 reported several complications for external fixation including thumb pain and various pin-related problems (see 'Characteristics of included studies' table) but failed to distinguish between primary and secondary external fixation.
Radial nerve neuropathy, often a short term sensory disturbance of the superficial radial nerve, was more common in the external fixator group; radial nerve symptoms in the plaster cast group were reported in a few cases of Howard 1989 and Young 2003. There was a marked excess of nerve related complications in the conservative treatment group in Howard 1989 that was not so apparent in the other trials.
The majority of cases of reflex sympathetic dystrophy (RSD) were reported as serious, requiring many months of intensive physiotherapy to resolve. There is no statistically significant difference between the two groups (25/384 versus 17/347; RR 1.31, 95% CI 0.74 to 2.32) in RSD. Previously, in Handoll 2003a, it was noticed that the two trials (McQueen 1996; Roumen 1991) including redisplaced, and hence unstable, fractures as well as older participants had more cases of RSD in the external fixator group, in contrast to most of the other included trials. As before, an exploratory analysis (see Analysis 01.18) was set to examine the difference in the direction of effect of the two subgroups (primary displaced fractures versus redisplaced fractures). As the difference between the two results is not statistically significant (P = 0.191) more evidence is required to establish if external fixation has a greater risk of RSD in older people with proven unstable fractures. It should also be noted that the differences between the two treatment groups in either subgroup of trials did not reach statistical significance.
Most cases with arthritis at seven years had mild cases of joint space narrowing in Young 2003 (11 versus 12); only one was symptomatic. The severity of the arthritic changes at 10 years in Howard 1989 was not reported (see Analysis 01.17).
There were no significant differences between the two groups in overall cosmetic deformity (Kapoor 2000) or various features of cosmetic deformity described in Hegeman 2004 (see Analysis 01.19). Howard 1989 reported that the pin-track scars were accepted by the patients; the only complaint being that the fixator made it difficult to use the hand. In Young 2003, 13 of the 36 people available at seven year follow up had "unsightly" forearm scars at the proximal pin insertion sites. Young 2003 found there was no difference between the two groups in the few people who were dissatisfied with their wrist at seven years (see Analysis 01.20).
Anatomical outcome (anatomical restoration and residual deformity)
Overall anatomical results at final radiological follow up were better, often significantly so, in the external fixation group in all trials except Horne 1990. The results of Horne 1990 may in part reflect the choice of radiological parameters (Van der Linden 1981) reported in Horne 1990 and the small numbers of participants (Axelrod 1991). Differences between the one year and seven year results for radial shortening (see Analysis 01.23) and dorsal angulation (see Analysis 01.24) in Young 2003 probably reveal a trend for deterioration over time but other factors, including the reduction in the sample size at seven years, may also influence these results. All the other trials indicated that external fixation held the reduced position better than plaster cast immobilisation. Redisplacement requiring secondary treatment occurred in 15% of conservatively treated fractures over nine trials (see Analysis 01.17: 7/356 versus 51/338; RR 0.17, 95% CI 0.09 to 0.32). Exceptionally, there were similar numbers of people in the two groups with redisplacement and treated redisplacement in Young 2003. Secondary treatment for redisplaced conservatively treated fractures consisted of remanipulation and external fixation in Abbaszadegan 1990, Lagerstrom 1999 and Pring 1988; these participants were then analysed as a separate group in these three trials. External fixation group participants with serious infection and plaster cast group participants requiring remanipulation were excluded from the rest of the analyses in Jenkins 1989. It is of concern to note that four people were recorded as requiring remanipulation in the full report of this trial, but six in another report (Jenkins 1988). Howard 1989 (external fixator group: distraction of the fixator increased for loss in radial length; plaster cast group: remanipulation), Kreder 2006 (external fixator group: open reduction and internal fixation; plaster cast group: external fixation if remanipulation was unsuccessful), Stein 1990 (plaster cast group: remanipulation) and Young 2003 (both groups: remanipulation) all retained those participants confirmed as having secondary treatment for redisplacement. Most of the 27 redisplacements occurring in the conservative treatment group in Rodriguez-Merchan 92 were probably remanipulated, but this was not stated explicitly in the trial report. Kreder 2006 reported that similar numbers had distal radius ulnar joint instability at six weeks (3/54 versus 2/59; RR 1.64, 95% CI 0.28 to 9.44); one person in the external fixation group had stabilisation of this joint by ulnar styloid repair. Statistically significantly fewer people in the external fixator group had recurrent instability in McQueen 1996 (14/60 versus 16/30; RR 0.44, 95% CI 0.25 to 0.77).
Four of the six trials rating overall anatomical results used the same rating scheme (Stewart 1984; Stewart 1985) which was derived from the scheme used by the two other trials (Roumen 1991; Stein 1990). The pooled results from these six trials give a consistent picture of the significantly superior anatomical results for external fixation (see Analysis 01.21 Anatomical grading: not excellent. 90/202 versus 148/169; RR 0.53, 95% CI 0.45 to 0.61; and Analysis 01.22 Anatomical grading: fair or poor. 17/214 versus 96/186; RR 0.17, 95% CI 0.11 to 0.27). It should be noted that the removal of the highly favourable results for the external fixation group of Stein 1990 shows these to be the basis of the statistically significant heterogeneity (P = 0.02; I² = 67%) in Analysis 01.21. This may be associated with the significantly better post-reduction results in the external fixation group of this trial, despite closed reduction being used in both groups.
The differences, in favour of the external fixation group, between the two treatment groups in the mean values of individual radiological parameters were reported to be statistically significant, often without confirmatory data, in eight trials: Abbaszadegan 1990 (dorsal angulation; radial shortening); Hegeman 2004 (dorsal angulation, radial angulation); Howard 1989 (dorsal angulation; radial shortening); Jenkins 1989 (dorsal angulation; loss in radial angulation; radial shortening), McQueen 1996 (dorsal angulation), Pring 1988 (radial angulation; from graph), Young 2003 (dorsal angulation (at one year); radial shortening); Zheng 2003 (dorsal angulation; ulnar angulation; radial shortening. Rodriguez-Merchan 92 also reported superior results in the external fixator group; for example, dorsal angulation (mean: 0.2 versus 9.5 degrees) and loss of radial length (mean: 0.5 mm versus 4.1 mm). Losses in various radiological parameters in those trials providing sufficient data are presented in Analysis 01.23. This shows that all losses in these parameters were less in the external fixation group, with two exceptions (Horne 1990: dorsal displacement; McQueen 1996: radial shortening). Since statistically significant heterogeneity (I² = 94.9%) was evident for radial shortening, pooling was not performed. The data presented in the analyses for Jenkins 1989 pertain to the time of fracture union and removal of cast and fixator at around four weeks. Later data at one year for this trial showed a further loss in mean radial length in the external fixator group compared with a slight gain in the cast group (mean loss: 1.24 mm versus -0.70 mm). Radial length shortening had worsened in both groups of Young 2003 at seven year follow up and the difference between the two groups was no longer statistically significant (see Analysis 01.23); however, the difference was reported to be statistically significant in both the draft and published reports (P < 0.05). Kreder 2006 reported the trend for better restoration in radial length and palmar tilt (volar angulation) in the external fixation group was not statistically significant. Superior anatomical values for dorsal and radial angulation and radial length were usually evident and claimed for the external fixation group (see Analysis 01.24) but the variation in the definitions of dorsal angulation and "normal" values makes the results hard to interpret.
Several other measures of structural deformity are presented in Analysis 01.25. Trialist-defined 'malunion' (based on dorsal angulation and radial shortening) reported by McQueen 1996 and Young 2003 was significantly less in the external fixation group at one year (36/108 versus 47/90; RR 0.58, 95% CI 0.41 to 0.81). In contrast, the incidence of malunion in those patients available at seven years follow up in Young 2003 was similar in the two groups (RR 0.96; 95% CI 0.63 to 1.47). Jenkins 1989 found a better reduction and fixation of die-punch fractures in the external fixation group; this may reflect the particular design of non-bridging fixator used in this trial where the distal pins also transfixed the fracture fragments.
No trial provided quantitative data on resource use and costs. It is not known how many people receiving surgery were day cases, thus avoiding the costs of a hospital bed. Overnight hospital admission was routine for surgical patients in Rodriguez-Merchan 92 whereas Horne 1990 referred to a day-care facility. Stein 1990 referred to admission to hospital for 24 hours or to a day-care facility. There was no report of the cost implications of remanipulation of redisplaced fractures or of the routine care of pin sites or additional care resulting from pin site infection.
This review of a common surgical intervention for one of the most common fractures has 15 included trials involving a total of only 1022 participants. Though essentially these trials evaluated one basic comparison, namely external fixation versus plaster cast immobilisation, there was considerable variation in the trials especially in terms of patient characteristics and interventions (see Table 4); and in their methods, including outcome assessment.
Limitations of the review methods
As this review abided by the criteria and methods set out in a published protocol, we have restricted our comments to two issues. The first is the question of whether trials have been missed or inappropriately excluded in our search and selection processes. The second issue concerns decisions made regarding pooling and subgroup analysis.
Our search was comprehensive and built on searches carried out over many years prior to the separate development of this review (Handoll 2003a). It has included the handsearch of conference proceedings and checks for ongoing trials. An inclusive and benefit-of-doubt approach during trial searches has been maintained throughout by the lead author (HH). Additionally, trial authors of unpublished trials have been sent requests for information and trial reports. It is possible that we have missed some potentially eligible trials but, if so, these may still not be suitable for inclusion, particularly if unpublished and inadequately reported. We guarded against study selection bias by the independent selection of eligible trials by all three review authors.
We decided to pool the results of evidently heterogeneous trials in an attempt to address the basic question of whether external fixation of any kind produces significantly different results compared with conservative treatment (plaster cast) in adults with distal radial fractures. We considered grouping trials by participant age and gender according to the three categories described in 'Description of studies'; in particular to examine the results in good quality bone in the younger age group compared with osteoporotic bone in the older age group. However, we decided that any findings would be impossible to interpret given the other differences in characteristics (types of fracture and interventions, assessment of outcome, methodology) and quality of the studies in the different age groups. Thus we concluded that the subgrouping the 15 trials into smaller separate categories was not viable. We, however, made one exception by subgrouping according to primary and secondary displaced fractures for RSD (see Analysis 01.18). This reflected a previously set up hypothesis in Handoll 2003a. As discussed below, our overall approach has important implications in terms of the interpretation of review findings.
Limitations of the review evidence
Overall, the available evidence is limited in scope and quantity, and is of uncertain validity. The usual reservations of the reliability of evidence from small and underpowered trials apply. Especially, we were careful to avoid mis-interpreting inconclusive evidence as 'evidence of no effect'. Systematic bias, in the form of selection, performance, exclusion or assessment bias, or a combination of these could not be ruled out for any trial. However, this was much less a concern with Kreder 2006, which was the only trial with clearly concealed treatment allocation. Another limitation was the inadequate assessment of outcome, particularly of function and in the long term. Non-validated outcome measures, such as those based on the Gartland and Werley scoring system (Gartland 1951), that combine aspects of function, pain, deformity and complications are particularly crude indicators of outcome.
Considerable caution is needed when interpreting these, especially when the scores have been reduced into categories such as excellent, good, fair or poor. Many trials predated the development of validated patient functional assessment instruments such as Short Form-36 (SF-36), the Disability of the Arm, Shoulder, and Hand questionnaire (DASH) and the Patient-Rated Wrist Evaluation (PRWE) (MacDermid 2000). These help to standardise functional assessment in a meaningful way and assist interpretation (Amadio 2001). Again, Kreder 2006 proved an exception.
Applicability of the review evidence
Generalising the findings of the included trials, should these be valid, is hampered by inadequate reporting of study details: in particular, the type and severity of the fracture, and quality of bone, all of which are key determinants of treatment and prognosis. The variety of fracture classification systems, with associated issues of reliability and validity further complicates this area (Jupiter 1997) and hinders direct comparisons between trials. For example, the two fracture classifications used by trials in this review (the AO and Frykman) place different emphases on various fracture patterns and anatomical components. Studies have revealed unsatisfactory interobserver reliability and intraobserver reproducibility for both classification systems (Andersen 1996; Kreder 1996b). Moreover, a large retrospective study found that neither classification system was useful for predicting clinical outcome (Flinkkila 1998). Thus, both fracture classification systems have been shown to have serious limitations. Nine trials in this review stipulated criteria for anatomical displacement of the fracture for trial entry. However, Kreder 1996a found quite broad margins of error ("tolerance limits") for anatomical measurements in general. This with other factors, such as variations in anatomical reference points, again hinders treatment comparison.
Fracture instability was the inherent or explicit criterion for many of the included trials, and proven for the two trials of redisplaced fractures (McQueen 1996; Roumen 1991). Generally though it is not established how best to predict fracture instability. A study of 4024 patients concluded that the patient age, metaphyseal comminution of the fracture and ulnar variance were the most important factors in predicting instability of distal radial fractures (MacKenney 2006).
Summary of the evidence
We have summarized the evidence using the categories described in BMJ 2006 (see Table 3). The effectiveness of external fixation compared with conservative treatment is graded as 3 ('Trade off between benefits and harms') as defined in Table 3. Some supporting information is provided in Table 6 and below.
Superior anatomical results, both in maintaining the reduced fracture position and in final anatomy, in the external fixation group were evident in all trials except Horne 1990 and Young 2003. Putative reasons for these two exceptions are given in the Results. The functional results of the included trials presented a mixed and incomplete picture, albeit potentially favouring external fixation, that is to a great extent based on functional grades from non validated scoring systems that also rate radiological deformity. As explained above, there are insufficient data to explore the effects of different trial characteristics, including age, on functional results.
Secondary displacement, here often termed redisplacement, after primary reduction and recurrent instability after re-reduction in redisplaced fractures was a frequent complication of conservatively treated patients. Most of the redisplacements resulted in a second reduction and application of a second plaster cast. Although some redisplacements might be explained by operator error, this finding is a reasonable indication of the instability of fractures in the trial populations. However, though external fixation was more effective in holding a reduced fracture position it comes at a price of a high number of associated complications. Most of the complications related to external fixator pins were pin-track infections that resolved with cleansing and antibiotics and there was a low incidence of pin loosening and premature removal of the external fixator. Radial nerve injury was more common in the external fixation group and can be a consequence of poor pin placement resulting from suboptimal surgical technique (Ahlborg 1999; Seitz 1993); the use of stab incisions in Horne 1990 rather than small incisions to visualise and retract soft-tissue during pin insertion was criticised (Axelrod 1991). RSD occurred in nearly six per cent of patients (42/731); we could not confirm that RSD was more common in patients with redisplaced fractures that had been treated by external fixation. Data for long-term complications and overall outcome were too limited for comments, aside from their general absence.
Sufficient data were not available to undertake meaningful subgroup analysis; e.g. fracture pattern, patient characteristics, variations in the interventions. The considerable variation in the trials means that only the basic question of whether external fixation confers some benefit is addressed here. Some clues as to the best methods of external fixation may arise from the pending systematic review of the evidence from direct comparisons of different methods of external fixation (Handoll 2007).
Implications for practice
There is some evidence from a set of 15 heterogeneous and generally methodologically weak trials to support the use of external fixation for dorsally displaced fractures of the distal radius in adults. While the evidence shows that external fixation maintains reduced fracture positions and prevents late collapse and malunion compared with plaster cast immobilisation, there was insufficient evidence to confirm a superior overall functional or clinical result. External fixation was associated with a high number of complications but many of these were minor and some would probably have been avoided using a different surgical technique for pin insertion. There was not enough evidence to prove or disprove a difference in more serious complications between the two groups.
Implications for research
Given that a distal radius fracture in adults is a common injury and given that there is limited knowledge about the best method of treatment, either conservative or surgical, further research is called for (Handoll 2003a). However, rather than embark on yet more small single-centre trials, particularly those with inadequate methodology that are unlikely to provide the good quality generalisable evidence required, the identification of the priority questions for the management of these fractures is required (Handoll 2003c). The updating of the evidence summaries of other surgical interventions for these fractures is likely to inform this process.
We thank Lesley Gillespie for her help with the search strategy. We thank the following for helpful comments and input at the editorial and external review of the protocol: Joanne Elliott, Bill Gillespie, Jesse Jupiter, Lindsey Shaw and Janet Wale. We thank the following for helpful comments at the editorial and external review of the review: Lesley Gillespie, Jesse Jupiter, Vicki Livingstone and Janet Wale. We thank Joanne Elliott and Lindsey Shaw for their help during editorial processing of the review.
We thank Xiaoyan Chen for a translation from Chinese. We thank Jean Goodman, at the British Orthopaedic Association, for facilitating the perusal of the Jenkins' trial thesis. We are also very grateful to those trialists, in particular Claire Young for extensive feedback, who provided clarification and further information on their trials.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Index terms
Appendix 1. Search strategy The Cochrane Library (Wiley InterScience)
#1 MeSH descriptor Radius Fractures explode all trees in MeSH products
#2 MeSH descriptor Wrist Injuries explode all trees in MeSH products
#3 (#1 OR #2)
#4 ((distal near radius) or (distal near radial)) in Title, Abstract or Keywords in all products
#5 (colles or smith or smiths) in Title, Abstract or Keywords in all products
#6 wrist* in Title, Abstract or Keywords in all products
#7 (#4 OR #5 OR #6)
#8 fractur* in Title, Abstract or Keywords in all products
#9 (#7 AND #8)
#10 (#3 OR #9)
Appendix 2. Search strategy for MEDLINE (OVID-WEB)
1. exp Radius Fractures/
2. Wrist Injuries/
3. (((distal adj3 (radius or radial)) or wrist or colles or smith$2) adj3 fracture$).ti,ab.
Appendix 3. Search strategies for CINAHL and EMBASE (OVID-WEB)
Inclusion of 'pins and plaster' external fixation
Comment from Mr David L Shaw (04.12.07):
The authors are to be congratulated on this extensive review of a common problem in current trauma practice. The validity of their conclusions are let down however by a glaring error in the abstract. The use of "pin and plaster" as a form of external fixation was only ever routinely applied to unstable tibial fractures or possibly in third world or battlefield situations. The mechanical construct which one can achieve with wires or screws inserted into the bone and then wrapped in plaster is in no way comparable to the use of a device specifically designed to stabilise a distal radial fracture.
In relation to the choice between fracture fixation and manipulation, only one of the most relevant questions relates to the risks of redisplacement and the benefit of remanipulation as opposed to defaulting to operative stabilisation if fracture reduction has been lost at some time after manipulation has been used as the primary treatment.
We thank Mr Shaw for his interest in our review and for his feedback.
Cochrane reviews are intended for a world-wide audience. It is thus appropriate that lower-cost methods such as 'pins and plaster' are considered. Although its results were consistent with the review conclusions, the actual contribution of quantitative evidence from the only trial testing pins and plaster external fixation to the review and the conclusions was minimal. Mr Shaw may be interested to read our review on "Different methods of external fixation for treating distal radial fractures in adults", which includes a comparison of external fixation versus pins and plaster fixation. Our conclusion of "unknown effectiveness" for this comparison reflects the inadequate evidence from two trials available to address this comparison but it is still notable that the evidence from neither trial condemned the use of plaster and pins fixation.
We agree there is a distinction between primary and secondary (upon redisplacement) fixation and consider in our review that both situations represent fracture instability.
Comment from Mr David L Shaw (04.12.07)
Response from HHG Handoll and WJ Gillespie (18.12.07)
Last assessed as up-to-date: 16 May 2007.
Protocol first published: Issue 4, 2006
Review first published: Issue 3, 2007
Contributions of authors
This review was initiated by Helen Handoll (HH) who prepared the first draft of the protocol. This was critically reviewed by the other two authors, Rajan Madhok (RM) and Jim Huntley (JH). HH searched for trials and contacted trial authors. All three authors performed study selection. HH and JH reviewed those trials that had not been included in a previous review covering all surgical interventions. HH repeated her review of the other included trials that had been quality assessed previously by RM and HH. HH completed the first draft of the review in RevMan. All versions were scrutinised by the other two authors. Helen Handoll is the guarantor of the review.
Declarations of interest
Sources of support
- University of Teesside, Middlesbrough, UK.
- No sources of support supplied
Some of the wording in each of several sections of this review (in particular: Synopsis, Background, Methods, Discussion and Implications) is taken either entirely or in only a slightly modified form from a related review on Percutaneous pinning for distal radial fractures in adults. This has been done to make the review self-contained and to ensure consistency between related reviews without requiring unnecessary cross-referring by readers.
Medical Subject Headings (MeSH)
MeSH check words
Adult; Aged; Female; Humans; Male; Middle Aged
* Indicates the major publication for the study