Description of the condition
The ankle joint is made up of three bones: the tibia ('shin bone'), the fibula (the other lower-leg bone) and the talus ('ankle bone'). The distal (lower) ends of the tibia and fibula form a 'saddle shaped' joint on top of and around the talus. The specific part of the tibia that curves down and articulates with the inside facing part of the talus is called the medial malleolus. The posterior malleolus, which is also part of the tibia, is at the back of the ankle joint. The lateral malleolus forms the distal end of the fibula and articulates with the lateral (or outside facing) part of the talus. Situated just above the ankle joint is the 'syndesmosis', which is the joint between the distal tibia and fibula held firmly together by ligaments. Together with the many ankle ligaments, the three malleoli help to keep the ankle joint stable during movement.
A broken ankle or ankle fracture is when one or more parts of the distal tibia or fibula that form the ankle joint are fractured or broken. There will often be associated soft-tissue injuries, particularly to the ankle ligaments. Most ankle fractures are closed injuries, in that the overlying skin remains intact. Around 2% are open fractures (Court-Brown 1998). An epidemiological study of 1500 ankle fractures in adults attending Edinburgh Royal Infirmary during 1988 to 1991 reported an annual incidence of 122 fractures per 100,000 persons (Court-Brown 1998). Of these, 52% occurred in men. The age distributions of fractures in men and women differed. The 58% of fractures that resulted from a simple fall or twisting injury tended to occur in elderly women. Conversely, sports injuries, mainly from soccer accidents, typically occurred in young men.
Court-Brown 1998 reported that the 70% of fractures were isolated malleolar fractures (predominantly of the lateral malleolus), 23% were bimalleolar (often of the medial and lateral malleoli) and 7% were trimalleolar (all three malleoli fractured). As well as categorising ankle fractures by location and the number of malleoli involved, various fracture classification systems have been devised to describe the different fracture patterns and help inform treatment decisions. Three commonly used classification systems for ankle fractures are the Lauge-Hausen system (Lauge Hansen 1942), the Weber system (Weber 1972) and the AO classification system (Müller 1969; Müller 1990). The Lauge-Hansen system classifies injuries by the position of the foot and direction of force at the time of injury (Lauge Hansen 1942). The anatomical classification system of Danis and Weber (Danis 1949; Weber 1972) subgroups fibular fractures as A (below the syndesmosis), B (at the syndesmosis) or C (above the syndesmosis) depending on the relationship of the fracture to the syndesmosis. The AO classification system is based on fracture patterns alone. Clinical decisions are, however, often made without considerations of these formal classification systems and will take into account also other aspects such as the damage to soft-tissues and the general health of the patient.
Description of the intervention
Following closed manipulation to achieve reduction of any displaced fractures, conservative management of ankle fractures generally comprises immobilisation in a below-knee cast for several weeks. This is to stabilise the fracture and allow it to heal. Various methods of immobilisation include casts (plaster or synthetic material), walking casts and functional braces. A Cochrane review on the rehabilitation of ankle fractures includes comparisons of different methods of immobilisation (Lin 2008).
Surgical treatment involves the reduction (if displaced) of the fractured parts and fixation using various devices such as metal plates, screws, tension bands or external fixation. These operative techniques aim to provide anatomical restoration and immediate stability, which facilitates earlier mobilisation. However, all surgery carries the risk of complications such as wound infection, pulmonary embolism, implant or fixation failure, mortality, amputation and reoperation (SooHoo 2009). For an evaluation of post-surgical rehabilitation interventions, see Lin 2008.
How the intervention might work
Healing of the bone takes at least several weeks and consists of five major phases: induction, inflammation, soft callus formation, ossification and remodelling (Koval 2002). For conservative treatment, immobilisation of the fractured parts is generally considered to be important for bone healing. However, immobilisation can lead to muscle atrophy, cartilage degeneration, and a stiff, painful and swollen joint. Also, conservative treatment can lead to secondary displacement, which generally requires surgery (Dietrich 2002), painful nonunion (Walsh 2004) and prolonged immobilisation.
With early mobilisation and postoperative exercises these adverse effects may be prevented. If surgical treatment can protect or accelerate the bone healing process by securely stabilising the fracture, it can also reduce recovery and 'back-to-work-time' and therefore indirect costs to society. This may not be the case for older people with osteoporosis because the porosity of their bones may increase the risk of fixation failure and thus preclude early mobilisation (Salai 2000a).
Anatomical restoration of displaced fractures is more likely to be achieved using surgical methods. A lateral talar dislocation of only one millimetre results in an average reduction of 42% of the contact area between talus and tibia (Ramsey 1976), which results in severe peak loads. It is assumed that such peak forces lead to secondary loss of cartilage, which in turn increases the risk of post-traumatic osteoarthritis. The better anatomical stabilisation achieved via surgery might reduce lateral talar dislocation more effectively than immobilisation alone and thus lessen the risk of long-term post-traumatic osteoarthritis. However, as well as the additional risks generally associated with surgery and anaesthesia, patients with conditions such as diabetes and peripheral vascular disease are at increased risk of complications and an unsatisfactory outcome. Moreover, there is evidence that not all fractures need perfect anatomical repair for a satisfactory outcome. Conservative treatment with closed reduction and cast immobilisation can yield good results for certain less severe fracture types (Bauer 1985a; Herscovici 2007; Kristensen 1985). Displaced fractures can be treated successfully with closed reduction and plaster cast (Rowley 1986a; Wei 1999). If immobilisation alone gains equal results easily for certain fractures types, surgical interventions should be considered as over-treatment.
Why it is important to do this review
In current practice, most Weber A fractures are treated conservatively and most Weber C fractures are treated by open, anatomical reduction and internal fixation. The remainder (roughly 50%) of all ankle fractures consists of Weber B fractures, which are treated both surgically and conservatively. Some clinicians think conservative measures are adequate in ankle fractures, but others consider exact anatomical reconstruction is essential to prevent predisposition for post-traumatic osteoarthritis. The rate of surgical interventions increases with the number of malleoli fractured, but depending on location, a wide range (14% to 72%) in the rate of surgical interventions has been reported in the USA (Koval 2005). There is also controversy in the treatment of older people with osteoporotic bones and other comorbidities that increase the risk of surgical complications (SooHoo 2009). Additionally, the Lauge-Hansen and Weber classification systems are not able to assess the intrinsic stability of all ankle fractures, which is considered an important determinant for the type of treatment. To compare the outcome of both treatment modalities for ankle fractures in adults, a systematic evaluation of benefits and harms is needed.
To assess the effects (benefits and harms) of surgical versus conservative interventions for treating ankle fractures in adults.
Criteria for considering studies for this review
Types of studies
Randomised and 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 studies comparing surgical and conservative treatments for ankle fractures in adults were included.
Types of participants
Studies including adult participants with ankle fractures who underwent a surgical or conservative intervention were included. Trials containing adults and children were included if the proportion of children was clearly small (< 5%), or if separate data for adults could be obtained. Similarly, we excluded studies in which participants with more extensive fractures of the distal third of the tibia (pilon or tibial plafond fractures) or studies that included more than 5% fractures with delayed presentation, unless separate data for acute ankle fractures could be obtained. Studies evaluating surgical revision of displaced fractures were excluded.
Types of interventions
Studies comparing any type of surgical treatment with any type of conservative intervention were included. The following procedures were compared:
- Any kind of fracture stabilisation with osteosynthesis (lag screw, plates, tension bands, bridge plating, external fixation or internal fixation)
- Any kind of fracture stabilisation with non-invasive interventions (plaster cast immobilisation, walking cast, orthosis, any kind of removable type of immobilisation).
Types of outcome measures
The primary outcome measures were functional outcome, pain and major adverse events. Preference was given to validated outcome measures, including visual analogue scale readings for pain. Examples of validated outcome measures included the physician-completed Olerud Molander Ankle Score (Olerud 1984), patient-rated functional outcomes such as the Lower Extremity Function Scale (Binkley 1999), and appropriate components of generic quality of life measures (e.g. SF36, SF12, and EQ-5D).
Major adverse events were generally considered as those related to the fracture or intervention that required secondary intervention (i.e. the need for surgery or further surgery; or prolonged intensive rehabilitation). Adverse events included insufficient primary osteosynthesis, soft tissue necrosis and any infection, osteitis, post-traumatic thrombosis, delayed union, nonunion, secondary fracture displacement, re-fracture, joint stiffness, muscular atrophy, tendinous insufficiency, sensory deficit, tarsal tunnel syndrome, and complex regional pain syndrome type 1 (also known as Sudeck’s dystrophy or reflex sympathetic dystrophy) (Stanton-Hicks 1995).
The secondary outcome measures sought included measures of recovery such as time to resume normal activities or return to work, range of motion, measures of functional impairment, anatomical result (x-ray) and radiologically-defined osteoarthritis.
Search methods for identification of studies
The search was conducted in two stages. We initially searched the Cochrane Bone, Joint and Muscle Trauma Group Specialised Register (to 4 June 2010), the Cochrane Central Register of Controlled Trials (The Cochrane Library, 2010 Issue 2), MEDLINE (PubMed,1965 to 4 June 2010), EMBASE (Elsevier, 1974 to 4 June 2010) and CINAHL (EBSCO, 1981 to 4 June 2010). We also searched the WHO International Clinical Trials Registry Platform (to 4 June 2010) and Current Controlled Trials (to 4 June 2010) for ongoing or recently completed trials. We then updated our search up to 6 February 2012, setting the initial date of search from 1 January 2010, for the following: the Cochrane Bone, Joint and Muscle Trauma Group Specialised Register, the Cochrane Central Register of Controlled Trials (The Cochrane Library, 2012 Issue 1), MEDLINE, EMBASE (searched using the Ovid interface), CINAHL and the WHO International Clinical Trials Registry Platform and Current Controlled Trials.
The MEDLINE strategy was developed in accordance with the guidelines outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Lefebvre 2009a). The subject-specific search was combined with the Cochrane Highly Sensitive Search Strategy for identifying randomised trials in MEDLINE: sensitivity- and precision-maximising version (Lefebvre 2009b). This strategy was adapted to the syntax and capacities of the other databases (see Appendix 1). There were no restrictions based on language.
Searching other resources
We looked for additional relevant trials by checking the reference lists of identified randomised trials. All authors of included trials were asked for additional information on any published, unpublished, or ongoing trials.
Data collection and analysis
Selection of studies
The titles, abstracts and descriptor terms of all downloaded material from the electronic searches were read by CD, who discarded clearly irrelevant reports. The remaining citations were then screened independently by CD and HA to establish the need for obtaining full articles. Full articles were also obtained where there was any uncertainty about the relevance of the study. Subsequently, CD and HA independently applied the inclusion criteria. Any differences in study inclusion were resolved by discussion with a third review author (CL).
Data extraction and management
Two review authors (CD and HA) independently extracted the data using a pre-piloted data extraction form. Any disagreement was resolved by discussion, if necessary approaching a third author. Where necessary, trial authors were contacted for further information on their trials. Extracted data were stored and managed using Review Manager. If required, interim statistical calculations were performed by CD and CL.
Assessment of risk of bias in included studies
Two review authors (CD and HA) independently assessed the risk of bias of each included study. Any differences were resolved by discussion, with arbitration by a third review author (CL). We used The Cochrane Collaboration's 'Risk of bias' tool (Higgins 2008). Each study was graded for risk of bias in each of the following domains: sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting and 'other' (for instance, extreme baseline imbalance). We also assessed performance bias, specifically in terms of surgeon's experience.
Measures of treatment effect
Treatment effect was measured using risk ratios for binary data, and mean differences or, where different outcome measures were used, standardised mean differences for continuous data. Ninety-five percent confidence intervals were used throughout.
Unit of analysis issues
As expected, the unit of randomisation was the individual patient in the included trials. There were no trials with a cluster-randomised design.
Dealing with missing data
Trial investigators were contacted for additional data if necessary. Where appropriate, we performed intention-to-treat analyses to include all people randomised to the intervention groups. We investigated the effect of drop-outs and exclusions by conducting sensitivity analyses. We were also alert to the potential mislabelling or non identification of standard errors and standard deviations. In case of missing data, we investigated whether they were missing at random, in which case the missing data was to be regarded as not having an important influence on outcome, or missing not at random. If data were deemed to be missing not at random, we stated that replacement values would not be imputed but sensitivity analyses would be considered.
Assessment of heterogeneity
We judged the appropriateness of pooling by assessing clinical diversity in terms of participants, interventions and outcomes of the included studies. Statistical heterogeneity was assessed by visual inspection of the forest plot and by using the I² and Chi² statistical tests.
Assessment of reporting biases
Had sufficient data for the primary outcomes been available, we planned to assess publication bias using funnel plots.
Where judged appropriate, we planned to pool data using both the fixed-effect (DeMets 1987) and random-effects models (DerSimonian 1986). Where there was no important difference between the two results, the results from the fixed-effect model were presented. Otherwise, depending on the results of heterogeneity tests, results from both models would have been presented.
Subgroup analysis and investigation of heterogeneity
If sufficient pooled data were available, we intended to conduct subgroup analyses to compare the effects of the interventions according to the risk of bias of the trials (low risk versus unclear or high risk), type of fracture (Weber A, B and C; displaced versus non-displaced), age (under 65 years; 65 or over), comorbidity (diabetic; non-diabetic), surgical experience (resident versus surgeon), different definitions of union (clinical versus radiological), and types of surgical (plate versus other fixation) or conservative treatment (e.g. early mobilisation versus immobilisation).
We intended to perform sensitivity analysis of pooled data examining various aspects of trial and review methodology, including the effects of missing data, and the inclusion of trials at high risk of bias (primarily, lack of allocation concealment) and trials only reported in abstracts.
Description of studies
Results of the search
After performing the first search up to 4 June 2010 (see Appendix 1) and removing duplicates, 1352 titles and abstracts were reviewed. Our subsequent search (2010 to 6 February 2012) resulted in 399 reference citations after the removal of duplicates (MEDLINE (48); The Cochrane Library (16); EMBASE (262); CINAHL (115)). Overall, of 18 potentially eligible studies, four were included, nine were excluded and five are ongoing or yet to be published trials (as of February 2012). All contact authors of the included trials (and the contact author of Salai 2000) were approached for additional information and clarification. While we were successful in contacting the authors of Phillips 1985 and Salai 2000, neither was able to provide additional information.
The four included studies (Bauer 1985; Makwana 2001; Phillips 1985; Rowley 1986) involved a total of 292 participants. These are summarised below, with a full summary for each trial detailed in the Characteristics of included studies.
All four trials were hospital based. Bauer 1985 was conducted in two hospitals in Sweden. The other three were single centre trials carried out in the UK (Makwana 2001; Rowley 1986) and the US (Phillips 1985). The first year of patient recruitment spanned from 1968 in Bauer 1985 to 1995 in Makwana 2001.
Information on gender was available for three trials: Bauer 1985 (64 female, 44 male); Makwana 2001 (31 female, 12 male); and Phillips 1985 (54 female, 42 male). In all, participant age varied between 15 and 91 years but Makwana 2001 set a lower age limit of 55 years and thus recruited an older population than the other three trials.
Bauer 1985 included patients with a displaced type A or B malleolar fracture. Makwana 2001 included patients with a displaced ankle fracture. Phillips 1985 evaluated patients with a closed supination-external rotation grade-4 or a pronation-external rotation grade-4 ankle fracture, classified according to the modified Lauge Hansen system (Lauge Hansen 1942), for whom a satisfactory closed reduction had been achieved. Rowley 1986 included patients with a displaced ankle fracture. Closed reduction was performed in all four trials.
Essentially, all four trials compared open reduction and internal fixation (ORIF) versus closed reduction and plaster cast immobilisation (conservative treatment).
In Bauer 1985, Makwana 2001 and Rowley 1986, surgical treatment was in accordance with AO/ASIF principles (Müller 1979). Phillips 1985 described the use of surgical techniques based on those of the Association for the Study of Internal Fixation (ASIF) (not referenced).
Bauer 1985 compared ORIF (followed by bed rest for five days, partial weight bearing from six weeks and full weight bearing at nine weeks) versus closed reduction and a plaster cast for six weeks (followed by partial weight bearing from six weeks and full weight bearing at nine weeks).
Makwana 2001 compared ORIF followed by a below-knee plaster cast for six weeks with protected weight bearing versus closed reduction followed by a below-knee plaster cast for six weeks with protected weight bearing.
In Phillips 1985, all participants had had a satisfactory closed reduction before randomisation to ORIF (followed by a below-knee plaster cast for one week; walking on crutches without weight bearing started a few days after surgery until the tenth week) versus a long-leg plaster cast for six weeks without weight-bearing and a below-knee for a further four weeks.
Rowley 1986 compared ORIF (followed by a below-knee backslab and active ankle movement for up to five days, then plaster cast for six weeks) versus closed reduction and a long-leg plaster cast for six weeks. Early weight bearing was encouraged in both groups.
The criteria for successful closed reduction were not clearly/precisely defined in Bauer 1985 and Rowley 1986. Both Makwana 2001 and Phillips 1985 used the same set of five criteria to define a satisfactory closed reduction.
Length of follow-up duration varied between 20 weeks (Rowley 1986) and an average of seven years (Bauer 1985). All four trials used different approaches to measuring function and clinical outcome. Bauer 1985 and Phillips 1985 used non-validated tools, whereas Makwana 2001 used the Olerud and Molander ankle score (Olerud 1984) as well as a visual analogue score to assess pain. The composite scoring scheme devised by Phillips 1985 assessed clinical and anatomical outcomes and arthritis. Rowley 1986 reported on the presence of foot deformity and foot print analysis for detecting abnormal foot angles. All four trials reported on complications and, to various extents, radiological outcomes.
There were nine excluded studies. Strömsöe 1995 did not compare surgical versus conservative treatment. The other eight studies compared surgical versus conservative treatment in ankle fractures, but were excluded because they were not randomised controlled trials (see the Characteristics of excluded studies). In particular, the claim to be a randomised trial in the report of Salai 2000 was not substantiated either through contact with the trial author or by the inappropriate presentation of study results.
We identified five ongoing trials (Gray; Harris; Pakarinen; Sanders; Willett), details of which can be found in the Characteristics of ongoing studies. Three trials are multi-centre studies. Two studies will be finished in 2014 and 2016. The publication of three studies was delayed; although they were due to finish in 2010, no published results were identified by our search in February 2012.
Risk of bias in included studies
|Figure 1. Risk of bias summary: review authors' judgements about each risk of bias item for each included study.|
|Figure 2. Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.|
Due to the high level of missing data, all authors were approached for additional information. Only one attempt was successful (Phillips 1985), but no additional information was obtained.
Generation of the allocation sequence was considered of low risk of bias in three trials (Bauer 1985; Makwana 2001; Phillips 1985) and high risk in Rowley 1986, the allocation of which was based on the patient's record number. Allocation concealment was judged at low risk of bias in Phillips 1985, at high risk in Rowley 1986, and of unclear risk in the remaining two trials (Bauer 1985; Makwana 2001) which provided no details of this aspect.
We acknowledge blinding can be difficult due to the nature of the intervention, but nevertheless none of the studies provided information about blinding and thus all were judged at being at high risk of both performance and detection bias.
Incomplete outcome data
The risk of attrition bias was considered high in Phillips 1985 (large (49%) loss of follow-up) and Makwana 2001 (exclusion of eight conservatively treated patients with secondary dislocation; lack of intention-to-treat analysis). Bauer 1985 was judged at unclear risk of bias: while intention-to-treat-analysis was performed and few were lost to follow-up, the effect of the exclusion of type A fractures at follow-up is unknown. In Rowley 1986, two patients were excluded from the analysis because they required secondary surgery after failed manipulation. The authors' intention was to analyse these patients separately but this was not done and intention-to-treat analysis was not performed. However, this was unlikely to alter the findings of the trial, which was thus judged to be at low risk of attrition bias.
As we did not have access to the study protocols, we judged this to be at unclear (unknown) risk of bias.
Other potential sources of bias
In Makwana 2001, the imbalance in the number of smokers (0 versus 6) between the two groups was considered to be another potential source of bias. No information was available to assess potential other sources of bias in any of the included studies.
Effects of interventions
All authors were contacted to obtain the original data, but none was forthcoming. The lack of these data prevented our plans to study the outcomes in different subgroups.
Based on the results of a non-validated questionnaire, Bauer 1985 found no difference between the two groups at seven years in the incidence of people with self-assessed significant 'troubles' from their ankle (11/43 versus 14/49; risk ratio (RR) 0.90, 95% CI 0.46 to 1.76; see Analysis 1.1). With the exception of a marginally statistically significant difference in favour of the conservative treatment group in the numbers of participants reporting swelling, there were no statistically significant differences between the two groups in participants reporting problems of pain, restricted range of motion, unsteadiness and 'passing stiffness' (see Analysis 1.1). There were also no statistically significant differences between the two groups in walking difficulties on rough or even ground (see Analysis 1.1). Four participants, two from each group, formerly employed as heavy manual labourers, changed occupation or quit working because of their ankle fracture. Makwana 2001 found a significantly better mean Olerud score (Olerud 1984) in the surgically treated group after a mean follow-up of 27 months (mean difference (MD) 17.00, 95% CI 0.64 to 33.36; see Analysis 1.2). At 3.5 years of follow-up, Phillips 1985 found a significant difference in favour of the surgical group for a non-validated combined score (0 to 150: best outcome) that rated clinical, anatomical and arthritis outcome (MD 10.30, 95% CI 0.93 to 19.67; see Analysis 1.2); however, there was no difference between the two groups in the clinical scores (subjective + objective assessment: 100 possible points): 88.8 versus 84.3, difference reported as non significant. Makwana 2001 found no difference in pain scores at 27 months (see Analysis 1.3).
While Bauer 1985 and Phillips 1985 retained for the purposes of intention-to-treat analyses those patients who did not receive their allocated intervention or were otherwise early treatment failures (i.e. failure of closed reduction), the other two trials explicitly excluded these. There were significantly more early treatment failures in the conservative treatment group (2/116 versus 19/129; RR 0.18, 95% CI 0.06 to 0.54; see Analysis 1.4). Skin damage precluded operations for two participants allocated surgery in Bauer 1985. As shown in the rest of the analyses in Analysis 1.4, there were no significant differences between the two groups in any of the specified complications (deep infection, superficial infection, surgical wound closure problem, skin ulcer, deep vein thrombosis, complex regional pain syndrome type 1 (Stanton-Hicks 1995), need for removal of internal fixation, surgical scar tenderness). It is noteworthy that all four conservative treatment group patients with wound infections in Bauer 1985 were patients who received surgery after closed treatment had failed.
Bauer 1985 showed no differences between the surgical and conservative groups in range of motion parameters at seven years (see Analysis 1.6). In Makwana 2001, the loss of dorsal range of motion compared with the contralateral ankle at follow-up was significantly less in the surgically treated group (9 versus 16 degrees, reported P = 0.044). Rowley 1986 found no statistically significant differences between the two groups regarding restriction of dorsiflexion or abnormal foot angle at 20 weeks follow-up (see Analysis 1.5). Time to independent weight bearing by all patients in Rowley 1986 was 16 weeks for the surgical group and 12 weeks for the conservative group.
At 3.5 years of follow-up, Phillips 1985 found a significant difference in favour of the surgical group in the anatomical scores (26.7 versus 22.1 (out of a maximum score of 35); reported P < 0.05). Pooled results for participants with radiological signs of osteoarthritis from two trials (Bauer 1985; Phillips 1985) showed no between-group differences (44/66 versus 50/75; RR 1.05, 95% CI 0.83 to 1.31; see Analysis 1.8).
In Bauer 1985, patients receiving surgical treatment left the hospital significantly later than conservatively treated patients (mean 9.5 versus 5.0 days, reported P < 0.05); 17 (30%) participants of the conservative treatment group were treated as outpatients. The median time of sick leave of fully employed participants was 14 weeks in both groups of Bauer 1985. In Makwana 2001, patients receiving surgery also left the hospital significantly later than the conservatively treated patients (mean 6.7 versus 2.6 days, mean difference 4.10 days, 95% CI 2.62 to 5.58; see Analysis 1.9).
Summary of main results
Four controlled trials (three randomised and one quasi-randomised) involving a total of 292 adults with displaced ankle fractures were included in this review.
Meta-analyses for functional outcome and pain were impossible due to the incompatible outcome measures used by the trials. The largest trial (Bauer 1985), following up 92 of 111 randomised participants, found no statistically significant differences between surgery and conservative treatment in the number of patients reporting symptoms and walking difficulties at seven years follow-up. One trial (Makwana 2001), reporting data for 31 of 43 randomised participants, found a significantly better mean Olerud score in the surgically treated group but no difference between the two groups in pain scores after a mean follow-up of 27 months. A third trial (Phillips 1985), reporting data for 49 of 96 randomised participants at 3.5 years follow-up, reported no difference between the two groups in a non-validated clinical score.
Early treatment failure, generally reflecting the failure of closed reduction (criteria not reported in two trials) probably or explicitly leading to surgery in patients allocated conservative treatment, was significantly higher in the conservative treatment group. Such patients were incorrectly excluded from the analyses of two trials. Otherwise, there were no significant differences between the two groups in any of the reported complications.
Bauer 1985 found no differences between the surgical and conservative groups in range of motion at seven years, whereas Makwana 2001 reported a better range of motion result for the surgically treated group. One trial (Rowley 1986) reporting data for 40 participants at 20 weeks follow-up found no statistically significant differences between the two groups regarding restriction of dorsiflexion or abnormal foot angle. Phillips 1985 found better anatomical scores in the surgical group at 3.5 years of follow-up. Pooled results for participants with radiological signs of osteoarthritis from two trials showed no between-group differences.
Two trials found that patients receiving surgical treatment stayed in hospital on average four days longer than conservatively treated patients.
Overall completeness and applicability of evidence
This review contains only four trials with a total of 292 participants. However, functional outcome data were available for far fewer participants (170 patients in three trials) and, moreover, these could not be pooled.
There were important differences between the four trials, including the types of participants (e.g. Makwana 2001 only recruited people over 55 years old), the surgical techniques (e.g. in Bauer 1985, ligamentous injuries were sutured, whereas this was not done in Makwana 2001), conservative treatment (e.g. types and duration of plaster cast immobilisation), the post-surgical treatment regimens (type and duration of plaster cast, early weightbearing or not, early active movements or not), and in the duration of follow-up and assessment of outcome. This variation, as well as lack of information on these trial characteristics (including the criteria for a satisfactory closed reduction), hinder the assessment of the applicability of the already limited and flawed evidence.
Duration of follow-up in Rowley 1986 of just 20 weeks is clearly unsatisfactory for a full assessment of outcome but it is also noteworthy that even a follow-up of seven years as in Bauer 1985 is insufficient to ascertain post-traumatic osteoarthritis. Horisberger 2009 showed that the latency time to end stage osteoarthritis after an ankle fracture can be as much as 20 years.
Quality of the evidence
Overall, the quality of the evidence is very limited. We judged the risk of bias was unclear or high in the majority of categories for the four trials. Rowley 1986 was quasi-randomised and thus at high risk of selection bias and all trials were at risk of bias due to lack of blinding, including of assessors. Of note is the high risk of attrition bias from incomplete outcome data in Makwana 2001, which incorrectly excluded patients after randomisation, and Phillips 1985, which had a large loss to follow-up. All trials were small and insufficient to confirm any lack of differences between the two groups.
Potential biases in the review process
The review was performed according to the unchanged published protocol. We tried to minimise publication bias by undertaking a comprehensive search strategy and checking non-English language citations (a translation of a Polish and German study resulted in their exclusion). Our search also included a search for ongoing and recently completed trials. However, it is still possible that potentially relevant trials have been missed. Additionally, although unsuccessful, we tried to obtain additional data and information from the included trialists.
Agreements and disagreements with other studies or reviews
The conclusion of this review is in line with another systematic review about the surgical versus conservative treatment of ankle fractures (Petrisor 2006). Petrisor 2006 also included Salai 2000, which we excluded because it did not appear to be randomised (and its data were unusable). Our search also located five ongoing or unpublished trials that may be included in a future update of this review.
Implications for practice
There is insufficient evidence to conclude whether surgical or conservative treatment produces superior long-term outcomes for ankle fractures in adults.
Implications for research
In future, adequately powered well designed and conducted, and appropriately reported clinical trials could provide more robust data. They should use validated outcome measures. Ideally, patients should be followed up for several years after randomisation and standard treatment regimens for surgical or conservative interventions should be used. Inclusion criteria and randomisation should take into account fracture type and displacement. However, before embarking on any new trials, it is important to note the existence of several ongoing studies, including a large multi-centre trial aiming to recruit over 600 patients. These trials also point to the importance of updating this review in due course.
The authors would like to thank Kate Bugler, William Gillespie, Helen Handoll and Caroline Hing for their valuable comments on drafts of the protocol and review. We also acknowledge the help of Lindsey Elstub, Joanne Elliott and Amy Kavanagh.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- Contributions of authors
- Declarations of interest
- Sources of support
- Index terms
Appendix 1. Search strategies
#1: Ankle Injuries[MeSH] OR Ankle[MeSH] OR Ankle Joint[MeSH]
#2: Fractures, Bone[MeSH] OR Fracture Healing[MeSH] OR Fracture Fixation[MeSH]
#3: #1 AND #2
#4: (Fracture*[TW] AND (ankle[TW] OR malleol*[TW] OR unimalleo*[TW] OR bimalleo*[TW] OR trimalleo*[TW] OR potts[TW] OR weber[TW] OR (distal[TW] AND (tibia*[TW] OR fibula*[TW]))))
#5: #3 OR #4
#6: Randomized Controlled Trial[PT]
#7: Controlled Clinical Trial[PT]
#10: Clinical Trials as Topic[MeSH: noexp]
#13: #6 OR #7 OR #8 OR #9 OR #10 OR #11 OR #12
#14: Animals[MeSH] NOT Humans[MeSH]
#15: #13 NOT #14
#16: #5 AND #15
The Cochrane Library (Wiley Online Library)
#1: MeSH descriptor Ankle explode all trees
#2: MeSH descriptor Ankle Injuries explode all trees
#3: MeSH descriptor Ankle Joint explode all trees
#4: (#1 OR #2 OR #3)
#5: MeSH descriptor Fractures, Bone explode all trees
#6: MeSH descriptor Fracture Healing explode all trees
#7: MeSH descriptor Fracture Fixation explode all trees
#8: (#5 OR #6 OR #7)
#9: (#4 AND #8)
#10: (fracture* NEAR/5 (ankle OR malleol* OR unimalleo* OR bimalleo* OR trimalleo* OR potts OR weber OR (distal AND (tibia* OR fibula*)))).ti,ab,kw
#11: (#9 OR #10)
1. Elsevier (1974 to June 2010)
#1 (('ankle'/exp OR 'ankle fracture'/exp OR 'ankle injury'/exp OR 'ankle dislocation'/exp) AND ('fracture'/exp OR 'fracture healing'/exp OR 'fracture fixation'/exp OR 'joint injury'/exp OR 'bone injury'/exp)) OR ((fracture* AND (ankle* OR malleol* OR unimalleo* OR bimalleo* OR trimalleo* OR potts OR weber OR (distal AND (tibia* OR fibula*)))) AND [<1950-2009]/py)
#2 (('crossover procedure'/exp OR 'double blind procedure'/exp OR 'randomized controlled trial'/exp OR clinical trial’/exp OR single blind procedure’/exp OR placebo’/exp) OR (random* OR factorial* OR crossover* OR ‘cross over’ OR placebo* OR 'double blind' OR 'single blind' OR assign* OR allocate* OR volunteer*)) AND [<1950-2009]/py)) AND ('human'/exp)
#3 #1 AND #2
2. Ovid (January 2010 to February 2012)
1 exp Ankle Dislocation/ or exp Ankle Injury/ or exp Ankle/ or exp Ankle Fracture/
2 exp Fracture Healing/ or exp Fracture/ or exp Fracture Fixation/ or exp Joint Injury/ or exp Bone Injury/
3 1 and 2
4 (ankle* or malleol* or unimalleo* or bimalleo* or potts or weber or (distal and (tibia* or fibula*))).tw.
6 4 and 5
7 3 or 6
8 Randomized Controlled Trial/
9 Clinical Trial/
10 Controlled Clinical Trial/
12 Single Blind Procedure/
13 Double Blind Procedure/
14 Crossover Procedure/
16 Prospective Study/
17 ((clinical or controlled or comparative or placebo or prospective$ or randomi#ed) adj3 (trial or study)).tw.
18 (random$ adj7 (allocat$ or allot$ or assign$ or basis$ or divid$ or order$)).tw.
19 ((singl$ or doubl$ or trebl$ or tripl$) adj7 (blind$ or mask$)).tw.
20 (cross?over$ or (cross adj1 over$)).tw.
21 ((allocat$ or allot$ or assign$ or divid$) adj3 (condition$ or experiment$ or intervention$ or treatment$ or therap$ or control$ or group$)).tw.
24 Case Study/ or Abstract Report/ or Letter/
25 23 not 24
26 7 and 25
27 (2010$ or 2011$ or 2012$).em.
28 26 and 27
29 limit 28 to human
S1: (MH "Fracture Healing") OR (MH "Fracture Fixation") OR (MH "Ankle Fractures") OR (MH "Fibula Fractures")
S2: (MM "Ankle") OR (MH "Ankle Injuries") OR (MH "Ankle Joint")
S3: S1 AND S2
S4: (fracture* AND (ankle* OR malleol* OR unimalleo* OR bimalleo* OR trimalleo* OR potts OR weber OR (distal AND (tibia* OR fibula*))))
S5: S3 OR S4
S6: (MH "Clinical Trials")
S7: (MH "Evaluation Research")
S8: (MH "Comparative Studies")
S9: (MH "Crossover Design")
S10: PT clinical trial
S11: S6 OR S7 OR S8 OR S9 OR S10
S12: TX ((clinical OR controlled OR comparative OR placebo OR prospective OR randomi*ed) AND (trial OR study))
S13: TX (random* AND (allocate* OR allot* OR assign* OR basis* OR divid* OR order*))
S14: TX ((singl* OR doubl* OR trebl* OR tripl*) AND (blind* OR mask*))
S15: TX (cross*over* OR (cross AND over*))
S16: TX ((allocat* OR allot* OR assign* OR divid*) AND (condition* OR experiment* OR intervention* OR treatment* OR therap* OR control* OR group*))
S17: S12 OR S13 OR S14 OR S15 OR S16
S18: S11 OR S17
S19: S5 AND S18
Protocol first published: Issue 4, 2010
Review first published: Issue 8, 2012
Contributions of authors
CD produced the first drafts of the protocol, search strategy and review. HA assisted with the search strategy. CL provided statistical feedback. All authors contributed to subsequent drafts of the protocol and review and approved the final versions. CD is the guarantor of the review.
Declarations of interest
Sources of support
- Luton and Dunstable NHS Trust, London, UK.Salary for healthcare activities and training medical students and residents
- St. Elisabeth Hospital, Tilburg, Netherlands.Salary for healthcare activities and training medical students and residents
- St. Radboud University Medical Centre, Nijmegen, Netherlands.Salary for healthcare activities and training medical students and residents
- No sources of support supplied
Medical Subject Headings (MeSH)
MeSH check words
Adult; Female; Humans; Male
* Indicates the major publication for the study