Description of the condition
Calcaneal or heel fractures are fractures of the calcaneus, also called the heel bone or os calcis. They comprise 2% of all fractures (Ibrahim 2007; Koval 2006). Most calcaneal fractures occur in younger working-age men. The economic impact of this injury to both the patient and society is considerable and is a consequence of extended hospital stay, cost of treatment, residual pain, time to mobilisation and delayed return to work (Schepers 2007). Studies suggest that people with these injuries can be incapacitated for up to three years and partially impaired for several years subsequently (Clarke 2007).
Heel bone fractures can be broadly divided into intra-articular (where the articular or joint surfaces of the calcaneus are disrupted) and extra-articular fractures (where the articular surfaces remain intact). Approximately three-quarters of calcaneal fractures are intra-articular (Ibrahim 2007). The majority of displaced intra-articular fractures involve the posterior facet, which is the major weight-bearing surface of the sub-talar (ankle bone) joint (Koval 2006).
Displaced intra-articular calcaneal fractures (DIACFs) are typically the result of high energy trauma, such as a fall or jump from a height. Patients present with a painful, swollen and deformed heel. Some patients may be unable to walk properly or at all. Bruising around the heel extending into the arch of the foot is suggestive of calcaneal fracture; blistering may also result as a consequence of significant swelling. Approximately 15% of all calcaneal fractures are open, where the fractured bone is exposed (Essex-Lopresti 1952), and between 5% and 10% are bilateral, thus involving both feet (Sanders 2000).
Description of the intervention
It is generally agreed that undisplaced extra-articular fractures should be managed conservatively, with a combination of rest, analgesia, compression, splinting and non weight-bearing for six to eight weeks (Clarke 2007). However the treatment of displaced intra-articular calcaneal fractures (DIACFs) is more problematic. Historically, DIACFs were also treated conservatively with a combination of rest with elevation, ice and immobilisation with plaster cast splintage, followed by physiotherapy and gradual mobilisation (Sanders 2000). However, this often led to delayed reconstruction of the malunited fracture, leaving patients with a painful and stiff foot which delayed or permanently prevented return to work and previous activities. Until the 1970s, operative treatment was technically challenging, and often led to postoperative infection, malunion, nonunion and amputation (McLaughlin 1963).
In the last 20 years, improvements in anaesthesia, antibiotic prophylaxis, the Arbeitsgemanschaft für Osteosynthesefragen/Association for the Study of Internal Fixation (AO/ASIF) principles of fixation, advances in materials and implants, and computed imaging have led to improvements in outcome after operative repair. This has popularised fixation of most fractures, including those of the calcaneus. Operative treatments for DIACF include reduction (repositioning of the displaced bone fragments) with percutaneous pin fixation, open reduction and internal fixation (ORIF) or primary arthrodesis (joint fusion). In percutaneous fixation, Kirschner-wires (K-wires) and pins are inserted through only minimal skin incisions. ORIF entails a skin incision through which the fracture fragments are visualised, realigned and then held in position by plate and screws. Subtalar arthrodesis (joint fusion) is generally used as a last resort and reserved for the most severe fractures or when non-union and prolonged joint pain predominate. Postoperative complications, such as surgical site infection and delayed wound healing, can occur after surgical stabilisation of calcaneal fractures. One study, based upon retrospective record review, reported that 25% of patients developed wound complications, with an increased risk observed in those with open fractures, diabetes and amongst smokers (Folk 1999).
How the intervention might work
Surgical fixation aims to restore, or at least improve, the normal anatomy of the heel bone and congruity of the subtalar joint. Accurate and rigid internal fixation of intra-articular fractures aims to give the best possible range of movement by allowing early motion and optimising joint mechanics. In doing so it aims to speed recovery and return to previous activities. This in turn should reduce long-term risk of developing symptomatic arthritis.
In the past some have advocated subtalar arthrodesis as the initial fracture treatment for all DIACFs (Harris 1946), because the severity of joint disruption was not retrievable and early fusion allowed more rapid return to function than a persistently painful and stiff malunited fracture. However, more recently this approach has been restricted to those with more severe injuries (Buch 1996; Sanders 1991).
Although advances in surgical techniques may have improved functional outcome for many patients, surgical fixation of these fractures is still technically challenging and the risks of surgical complications, such as surgical site infection (SSI), and of treatment failure remain.
Why it is important to do this review
A previous systematic review of the treatment of calcaneal fractures (Bridgman 1999) concluded that "even where there is some benefit of operative compared with non-operative treatment, it remains unclear whether the possible advantages of surgery are worth its risks". This conclusion was based on the evidence available from three small trials (O'Farrell 1993; Parmar 1993; Thordarson 1996), all of which were reported to be of poor methodological quality. The review by Bridgman 1999 referred to a large then-ongoing trial, which has since been published (Buckley 2002). It is thus important to update the review in the light of the new evidence, including data from Buckley 2002. Additionally, although modern surgical intervention has improved post-fracture outcome for many patients, there remains a lack of consensus on the use of surgery for the optimal management of calcaneal fractures.
To assess the effects (benefits and harms) of surgical versus conservative treatment of displaced intra-articular calcaneal fractures.
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 evaluating surgical versus conservative management for displaced intra-articular calcaneal fractures were eligible for inclusion.
Types of participants
Individuals aged over 14 years, considered as skeletally mature, with displaced intra-articular calcaneal fractures (DIACFs). Participants with unilateral or bilateral fractures were eligible for inclusion. We included all severities of fracture, both open and closed.
Types of interventions
Trials comparing surgical versus conservative treatment of DIACF were considered eligible for inclusion.
- Closed manipulation with percutaneous pin fixation
- Open reduction with internal fixation (ORIF) with or without bone graft
- Primary arthrodesis
- Ice and elevation
- Plaster cast immobilisation
- Early or delayed mobilisation
Rehabilitation, such as physiotherapy, is likely to be provided after definitive treatment, thus after either surgical or conservative treatment. Some trials may incorporate multiple interventions, such as surgery with mobilisation and physiotherapy. In such cases, patients having surgery were considered as the surgical treatment group and patients having non-surgical interventions were considered as the conservative treatment group.
Types of outcome measures
Preference was given to validated, patient-reported outcome measures of function and pain.
- Function (e.g. walking ability). An example of an instrument commonly used to measure function (and pain) is the American Orthopaedic Foot and Ankle Society score (Kitaoka 1994)
- Chronic pain (e.g. pain lasting beyond expected healing time, three months)
- Health-related quality of life (QoL)
- Return to work and former activities, such as wearing usual footwear, returning to sporting activities, etc.
- Postoperative complications (e.g. surgical site infection (SSI)) and other postoperative morbidity or other serious adverse event
- Subsequent operation (after primary treatment)
- Objective measures of impairment (e.g. range of ankle movement)
- Time to union (bone healing)
- Radiological measurements (e.g. Bohler’s angle and signs of osteoarthritis)
Search methods for identification of studies
We searched the Cochrane Bone, Joint and Muscle Trauma Group Specialised Register (to July 2011), the Cochrane Central Register of Controlled Trials (CENTRAL)(The Cochrane Library, 2011 Issue 3), MEDLINE (1948 to July 2011) and EMBASE (1980 to 2011 Week 27). To identify ongoing trials, we searched the WHO International Clinical Trials Registry Platform and Current Controlled Trials. We also searched the Orthopaedic Trauma Association annual meeting archives from 1996 to 2011. No restrictions were applied to the language of publication.
In MEDLINE, we combined subject-specific terms with the Cochrane Highly Sensitive Search Strategy for identifying randomised trials (sensitivity- and precision-maximizing version) (Lefebvre 2011). In EMBASE, subject-specific terms were combined with the relevant Scottish Intercollegiate Guidelines Network randomised controlled trial search strategy (SIGN). The search strategies for CENTRAL, MEDLINE and EMBASE are shown in Appendix 1.
Searching other resources
We checked the reference lists of relevant articles and contacted researchers involved with ongoing trials.
Data collection and analysis
Selection of studies
Both review authors independently examined the titles and abstracts of articles identified in the search as potentially relevant trials. From this initial assessment, we obtained full versions of all potentially relevant articles.
Data extraction and management
Both review authors independently extracted data using a piloted data extraction form. Disagreements were resolved by discussion.
Assessment of risk of bias in included studies
Both review authors independently assessed risk of bias using The Cochrane Collaboration's 'Risk of bias' tool (Higgins 2009). Assessors were not blinded to the authors or source institution. Appraisal criteria included: sequence generation for randomisation, concealment of allocation, blinding, incomplete outcome data, selective outcome reporting and other potential sources of bias (for example, an extreme imbalance in baseline patient characteristics).
Each of these factors was recorded as yes ('low' risk of bias), no ('high' risk) or unclear with a brief summary provided in table format (see the Characteristics of included studies). Where data were unclear, we contacted authors for clarification, where possible. After this process, each paper was graded as being at low, unclear or high risk of bias. Appendix 2 gives more information about the 'Risk of bias' tool.
Measures of treatment effect
For each outcome, summary estimates of treatment effect (with 95% confidence intervals (CI)) were calculated where possible. For dichotomous outcomes, the risk ratio (RR) was used when appropriate. For continuous outcomes, the mean difference (MD) was used when appropriate. Standardised mean differences were planned for pooling continuous outcomes using different measurement scales. Where possible, intention-to-treat data were used in the analysis.
Unit of analysis issues
We anticipated that unit of randomisation and analysis in the included trials would be the individual patient. This was confirmed by initial pilot screen of the published literature. However, bilateral calcaneal fractures are quite common and trials including patients with bilateral fractures may present results for fractures or limbs rather than individual patients. Where such unit of analysis issues arose and appropriate corrections had not been made, we considered presenting data for such trials where the disparity between the units of analysis and randomisation was small. However, all trials included in the final review analysed by individual rather than by fracture site.
Dealing with missing data
Where data were missing or unsuitable for analysis (for example where intention-to-treat analysis was not presented), we attempted to contact study authors for further information and data. Where data were missing to the extent that the study could not included in the meta-analysis and attempts to retrieve data had been exhausted, results were presented and discussed in the context of the findings. We calculated missing standard deviations from other available data such as standard errors, confidence intervals or P values. We calculated P values for proportions where these were not reported. We used available case analysis, whereby data were only included for those whom the result was known, thus denominators relate to those with data for the particular outcome in question. We did not impute missing values for the analyses (except for calculation of missing SDs and P values).
Assessment of heterogeneity
When deciding whether meta-analysis was appropriate, we assessed the clinical diversity across studies. This included assessment of the comparability of participant characteristics (such as age and type of fracture), interventions, co-interventions, and outcomes. Where two or more studies were deemed to be clinically homogenous, the pooled data were assessed for statistical heterogeneity using RevMan. Heterogeneity was assessed by visual inspection of the forest plot (analysis) along with the test for heterogeneity and the I² statistic (Higgins 2003).
Assessment of reporting biases
Where sufficient trials and data were available, we were to attempt to assess publication bias by preparing a funnel plot. Our searches for trials listed in clinical trial registers and trial protocols should help to avoid publication bias. Study authors were contacted in an attempt to establish a full data set or obtain reasons for the non-reporting of certain outcomes. This was not possible for some studies because of the length of time since publication, difficulty in tracing original data and failure to trace authors.
Quantitative data were entered into RevMan and analysed using Cochrane MetaView. For each comparison, summary estimates of treatment effect together with 95% confidence intervals (CIs) were calculated for individual outcomes. For dichotomous outcomes, risk ratios (RRs) were calculated. For continuous outcomes, mean differences (MDs) or, where more appropriate, standardised mean differences (SMDs) were calculated. Initially we used the fixed-effect model. We also used the random-effects model in the event of unexplained heterogeneity.
Where possible, time to return to work and time to bone healing were analysed as survival (time to event) outcomes, using the appropriate analytical method (as described in the Cochrane Handbook for Systematic Reviews of Interventions). Where it was not appropriate to pool data, results were presented using a narrative approach and in table format.
Subgroup analysis and investigation of heterogeneity
In our investigation of heterogeneity we planned to consider aspects of clinical and methodological diversity. Considerations of clinical diversity were to include assessment of differences in study location and setting, participant characteristics including co-morbidities, characteristics of the trial interventions and of other care provided. For methodological diversity, this included assessment of the randomisation process, study quality (explicitly in terms of risk of bias), outcome measurement and analytical method.
We intended to analyse studies according to differences in surgery or conservative management. We did not anticipate any specific subgroup analyses, except for where there may be marked differences in surgical approaches, or specific regimens of rehabilitation or physiotherapy. If suitable data were available, we intended to conduct a subgroup analysis by severity of fracture and test whether subgroups were statistically significantly different from one another (Altman 2003). However, this was not undertaken due to the lack of clarity of description about fracture severity on all participants in included studies.
We intended to conduct sensitivity analyses by examining various aspects of trial and review methodology, including the effects of missing data, and including trials at high or unclear risk of bias, such as selection bias arising from the lack of allocation concealment, and trials only reported in conference abstracts.
Description of studies
Results of the search
Both review authors independently read 305 titles and abstracts. Where discrepancies arose, the full paper was obtained for review. We identified 40 study reports for full assessment.
A total of four trials were included with data published across multiple (n = 20) publications (Buckley 2002; Chrintz 1993; Parmar 1993; Thordarson 1996). Nineteen study reports were excluded and one trial was placed in ongoing studies (see the Characteristics of ongoing studies). A study flowchart is presented in Figure 1.
|Figure 1. Study flow diagram.|
Data from the four included trials were published in a total of 20 conference abstracts and publications. The trials in the review were published between 1993 and 2002 although one trial recruited participants between 1977 and 1979 (Chrintz 1993). Parmar 1993 recruited patients between 1985 and 1992. The largest trial (Buckley 2002) recruited 426 participants between 1991 and 1997. Thordarson 1996 did not report on the timing of recruitment. All trials randomised by the individual rather than by fracture.
Studies were conducted in Canada (Buckley 2002), Denmark (Chrintz 1993), UK (Parmar 1993) and the USA (Thordarson 1996). Three trials were single centre studies (Chrintz 1993; Parmar 1993; Thordarson 1996) and one was a multi-centre trial (Buckley 2002). Overall, 602 participants were randomised to either surgery or conservative treatment of displaced intra-articular fracture and follow-up data were available for 443 (74%) participants. Three studies reported sample age and sex characteristics, either for those recruited (Buckley 2002) or for those followed up after treatment (Parmar 1993; Thordarson 1996). Overall, the age of participants ranged from 15 to 79 years, with estimates of mean age ranging from 35 years (Thordarson 1996) to 48 years (Parmar 1993). A total of 450 males and 56 females were included, reflecting the higher proportion of calcaneal injury in males (89% versus 11%).
The four studies used different methods of surgery (Buckley 2002: plate, screw or wire fixation; Chrintz 1993: Steinmann pin; Parmar 1993: K-wires; Thordarson 1996: plate fixation). There were also differences in the conservative and post-surgical treatment (e.g. Parmar 1993 used plaster cast mobilisation after surgical fixation; Thordarson 1996 prescribed early mobilisation), including timing of commencing weight-bearing (see the Characteristics of included studies).
The trials varied in timing of follow-up. Outcome data were reported at one to two years after treatment by three studies (Chrintz 1993; Parmar 1993; Thordarson 1996). Buckley 2002 reported follow-up data for two to eight years, with a mean follow-up of three years. Parmar 1993 also reported 15 year outcomes (Ibrahim 2007); however, this was for a limited subset of 33% of the original randomised cohort.
Details of the 19 excluded studies are presented in the Characteristics of excluded studies. Of these, seven were review articles and nine were non-randomised controlled trials. Of these, one study (O'Farrell 1993) was an uncontrolled clinical study whereby allocation was based upon consultant treatment preference in a sequential series of participants; this study was included within previous systematic reviews (Bridgman 2002; Gougoulias 2009). Two excluded studies, which were solely reported as registrations in the UK-NRR archive were either not completed or no further details were obtained.
Risk of bias in included studies
Overall, all trials had methodological flaws that put them at either unclear or high risk of bias for at least one domain (see Figure 2 and Figure 3). A full description of risk of bias assessment for individual studies is also provided in the Characteristics of included studies.
|Figure 2. Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.|
|Figure 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study.|
Parmar 1993, which used quasi-randomisation methods with allocation to treatment group by year of birth, was judged at high risk of selection bias. Chrintz 1993 was reported in two conference abstracts as being a 'randomized trial' but with no description of sequence generation or allocation concealment. Thordarson 1996 used sealed unmarked envelopes although no details were provided about sequence generation.
The large multi-centre Canadian trial (Buckley 2002) employed a more rigorous methodology, whereby random numbers were generated by a central administration site; these random number assignments were sent to recruiting study centres. This study described using a 'pre-randomized design' approach to minimise open discussion of uncertainty.
Given the nature of the intervention, it was not appropriate to assess blinding of participants or treatment providers. However, it was appropriate to assess blinding of outcome assessors. Two trials (Chrintz 1993; Parmar 1993) were graded at high risk of bias and the other two trials (Buckley 2002; Thordarson 1996) at unclear risk. Buckley 2002 administered a standardised questionnaire to assess quality of life and disease-specific symptoms, but clinical judgement about quality of assessment of fracture reduction was conducted by the main investigator who had undertaken 73% of all surgical procedures.
Incomplete outcome data
All four trials were at high risk of bias from incomplete outcome data. This was mainly due to attrition bias from a loss to follow-up or post-randomisation exclusions. Incomplete reporting of data, such as providing percentages only in the main report of Parmar 1993, was another source of bias.
Buckley 2002 had a high attrition rate (27%). Although an intention-to-treat analysis was described (using 'complete study group' data), the derivation of the missing data was not clear. Additionally, participants who went on to have a subtalar arthrodesis for osteoarthritis (7 surgical group patients and 37 conservative group patients) were reported as having been excluded from the analysis). Chrintz 1993 was inadequately reported but also had a 24% loss (16/68) to follow-up. The description of participant flow in Parmar 1993 was unclear, with a high attrition (24/80 = 30%) at two years follow-up (an earlier Cochrane review (Bridgman 2002) reported that these 24 participants were excluded because they were followed up for less than one year) and an even higher attrition at 15 years follow-up, with outcome data being available for 26 participants only. Thordarson 1996 lost four (13%) of 30 randomised participants by 15 month follow-up. Additionally, only 11/30 (37%) participants attended for follow-up clinical examination with an imbalance between the two groups (8 versus 3).
Two trials were at high risk of selective reporting because data for only one outcome were reported (Chrintz 1993) or because clinical outcomes were only reported for a subset of participants at follow-up (Thordarson 1996). The trial by Parmar 1993 excluded participants followed up for less than one year (24/80, 30%). For Buckley 2002, there was no apparent evidence of selective reporting, although the authors subsequently reported outcomes for selected participant subgroups.
Other potential sources of bias
Buckley 2002 was extensively published (nine publications) and although there appeared to be no differences in the main outcomes between treatment groups, subsequent publications found statistically significant findings within participant subgroups.
Effects of interventions
Data in the main report of Parmar 1993 were presented as whole percentages; some of which did not yield whole numbers upon calculation of the numerators for dichotomous data. In these cases we have rounded-up where the first decimal point is equal to or above 0.5.
Function e.g. walking ability
Three trials (Buckley 2002; Parmar 1993; Thordarson 1996) reported data for functional outcomes although based on different outcome measures. Chrintz 1993 stated that there were no differences between treatment groups in terms of functional outcomes but data values were not reported.
The measures used to capture function included: composite functional or gait scores, ability or distance walked without pain or difficulty, and other walking-related outcomes. The results are described below and presented in Analysis 1.1 (Figure 4) and Analysis 1.2.
|Figure 4. Forest plot of comparison: Function (e.g. walking ability)|
Higher scores indicate improved function, except for FFI where higher score indicates greater impairment (score reversed). For Buckley 2002, a higher composite VAS score is assumed to indicate greater improvement.
Buckley 2002 used a 'validated disease-specific' VAS scale comprising 12 patient-completed items for pain, physical function, difficulty walking, limp and overall result from heel fracture. This scale also incorporated items for completion by the surgeon and an independent assessor (three items each). The overall score produced ranges from 0 to 100, where higher scores indicated improved outcome. The trial found no statistically significant differences between the two groups in the mean VAS scores (424 participants; MD 4.30, 95% CI -1.11 to 9.71; P = 0.12). Values for the standard deviations were calculated from the reported P value (P = 0.13). These data, which were presented for the 'complete study group', are assumed to relate to participant follow-up at a mean of three years after treatment (Buckley 2002).
A subsequent publication from Buckley 2002 selected seven questions from the full VAS scale and reported these as 'gait satisfaction' scores on a subset of 319 participants, stratified by various patient characteristics (e.g. age, sex, workers compensation status), with two to eight year follow-up (O'Brien 2004). However, O'Brien 2004 did not report mean gait satisfaction scores for the whole sample or by treatment group.
Thordarson 1996 developed a composite functional outcome questionnaire based upon an early version of the AOFAS instrument. This questionnaire includes domains of pain, daily activity, shoe wear, walking, exercise and work. Scores range from 0 (worst result) to 100 (best result). Mean functional scores were significantly higher 15 months after surgery (15 participants) compared with conservative treatment (11 participants): mean 86.1 versus 55.0; MD 31.10, 95% CI 17.08 to 45.12; P < 0.0001; see Analysis 1.1).
The 15 year follow-up (Ibrahim 2007) of a limited subset of 26 participants from Parmar 1993 presented data for the AOFAS hindfoot scale (0 to 100), Foot Function Index questionnaire (0 to 100: higher scores indicate greater impairment) and a composite Calcaneal Fracture Scoring system (0 to 100: higher scores indicating greater improvement). There were no statistically significant differences between surgical (n = 15) and conservative participants (n = 11) with respect to any of these outcome measures (see Analysis 1.1).
Parmar 1993 found no significant differences between the two groups at two years in the proportion of participants with no walking difficulty (7/25 versus 6/31; RR 1.45, 95% CI 0.56 to 3.76); ability to walk unlimited distance (11/25 versus 16/31; RR 0.85, 95% CI 0.49 to 1.49); or walking without a limp (18/25 versus 23/31; RR 0.97, 95% CI 0.70 to 1.34) (see Analysis 1.2).
Chrintz 1993 reported without presenting data that there was no difference in pain between treatment groups. Pain was measured within composite functional or quality of life scales in the other three trials (Buckley 2002; Parmar 1993; Thordarson 1996). Pooled pain data reported in two trials (Parmar 1993; Thordarson 1996) for chronic pain (i.e. not minimal/infrequent pain) presented in Analysis 1.3 (Figure 5) showed no statistically significant difference between the two groups (19/40 versus 24/42; RR 0.79, 95% CI 0.53 to 1.18; I² = 44%). While Parmar 1993 presented percentages data for site and pattern of pain, the outcome definition and denominators used are unclear. Approximately half of the people had pain while walking on uneven or any surface in Thordarson 1996 (8/15 (53%) surgery versus 6/11(55%) conservative); three conservative group participants had pain when active and one had 'constant and bothersome' pain. The 15 year follow-up (Ibrahim 2007) of Parmar 1993 found different trends but no statistically significant differences between the two groups in the mean pain components within three composite functional scores (see Analysis 1.4).
|Figure 5. Forest plot of comparison: Chronic pain.|
The risk ratio indicates risk of developing chronic pain after treatment.
Health-related quality of life (QoL)
Only one trial reported health-related quality of life data whereby the Short-Form-36 (SF-36) questionnaire was used at one and two years after treatment (Buckley 2002). Based upon follow-up data from 424 participants, there was no statistically or clinically significant difference between the two treatment groups respectively (MD 4.00, 95% CI -1.16 to 9.16; see Analysis 1.5; P = 0.13). Values for the standard deviations were calculated from the reported P value (P = 0.13). These data were manipulated by study authors in an attempt to identify differences by subgroups: SF-36 scores were dichotomised as above or below the sample mean and odds ratios were reported by age bands, sex, Bohler's angle, displacement classification, workload and involvement of injury. Given that sample sizes for these subgroup analyses were not reported and the high risk of multiple statistical testing, these data have not been presented in this review.
Return to work and former activities (ability to wear footwear, participate in sport, etc)
Return to work and other activities
The main report of Buckley 2002 did not present specific data on return to work. In a separate publication, Tufescu 2001 presented data on prediction of return to work; this was from a single surgeon who had participated in the wider study. These data were not used in this review. Chrintz 1993 reported, without presenting data, that there were no differences between the two groups in working ability, working capacity or leisure activity,
Data on return to work as before injury were pooled from two studies (Parmar 1993; Thordarson 1996) (see Analysis 1.6). The data for Parmar 1993 applied to the 35 men who were in work before their injury. A greater proportion of surgical participants returned to the same job compared with those having conservative treatment although this was not statistically significant and the data were heterogeneous (27/34 versus 15/27; RR 1.45; 95% CI 0.75 to 2.81; I² = 55%). The difference between the two groups was less for those returning to any employment (32/34 versus 22/27; RR 1.14; 95% CI 0.94 to 1.37) which included lighter and part time work. Seven of the remainder (2/34 versus 5/27) were unemployed and one in the conservative treatment group of Thordarson 1996 had severely impaired work capacity and was probably unemployed. At 15 years follow-up of Parmar 1993, there was no significant difference between the two groups in the work component of the Calcaneal Fracture Scale score (MD -0.60, 95% CI -8.27 to 7.07; see Analysis 1.7).
In Parmar 1993, a smaller proportion of surgical group participants returned to their previous recreation level (16/25 versus 23/31; RR 0.86, 95% CI 0.60 to 1.24; see Analysis 1.8). Thordarson 1996 found a greater proportion of surgical group participants had no limitations in their daily or recreational activities (13/15 versus 5/11; RR 1.91, 95% CI 0.97 to 3.75; see Analysis 1.8). Neither of these results were statistically significant. One conservatively treated participant in Thordarson 1996 had a severe limitation of activity.
Ability to wear usual shoes
Data for ability to wear usual shoes were available for two studies (Parmar 1993; Thordarson 1996) (see Analysis 1.9). These were not pooled given clearly statistical heterogeneity, which also may reflect differences in outcome measures. Parmar 1993 found no difference between the two groups in the ability to wear normal shoes (18/25 versus 22/31; RR 1.01, 95% CI 0.73 to 1.41). All participants of Thordarson 1996 were able to wear most shoes comfortably, although significantly more participants in the surgical group reported being able to wear all shoes comfortably (13/15 versus 4/11; RR 2.38, 95% CI 1.06 to 5.34).
Postoperative complications, other postoperative morbidity or serious adverse event
Three trials reported postoperative and other complications after treatment (Buckley 2002; Parmar 1993; Thordarson 1996) (see Table 1). Data from Buckley 2002 were reported over different publications: Howard 2003 found significantly more participants with major complications in the surgical group (57/206 versus 42/218; RR 1.44, 95% CI 1.01 to 2.04; see Analysis 1.10). These complications included late arthrodesis, thromboemboli, superficial and deep wound infections, malposition of fixation and compartment syndrome (see Table 1).
Parmar 1993 found no statistically significant difference between the two groups in sural nerve symptoms at one year follow-up (8/25 versus 6/31; RR 1.65, 95% CI 0.66 to 4.14; see Analysis 1.10). Notably, none of these 14 participants were found to have sural nerve hypoaesthesia when examined clinically.
Thordarson 1996 reported the only complication was a superficial site infection in surgical treatment group (1/15 (6.7%)).
Two studies (Chrintz 1993; Parmar 1993) reported on valgus deformity of the heel at follow-up. Chrintz 1993 reported without numerical data that this was found only in "conservatively treated patients". Three participants in each group of Parmar 1993 had valgus deformity (3/25 versus 3/31; RR 1.24, 95% CI 0.27 to 5.62; see Analysis 1.10).
Only Buckley 2002 reported data on reoperation or subsequent operation. In the surgical arm, two out of 206 participants (1%) required reoperation for hardware removal because screws had penetrated the posterior facet of the subtalar joint. Of those initially entered to the study, significantly fewer participants in the surgical arm went on to have subtalar arthrodeses for the development of subtalar arthritis (7/206 versus 37/218; RR 0.20, 95% CI 0.09 to 0.44; see Analysis 1.11). These patients were excluded from the final analysis. Additionally, the authors referred to subtalar arthrodeses to treat severe persistent pain within two years but these data were unavailable (Buckley 2002).
Range of movement (ROM)
Three trials (Chrintz 1993; Parmar 1993; Thordarson 1996) reported on range of movement but Chrintz 1993 stated only that there was no difference between the two groups in the "movement of the joints".
Parmar 1993 found no statistically significant differences at two year follow-up between the two groups in the numbers with normal ankle movement (19/25 versus 21/31; RR 1.12, 95% CI 0.81 to 1.56), normal subtalar movement (5/25 versus 4/31; RR 1.55, 95% CI 0.46 to 5.17) and those with more that 50% of the subtalar movement of the opposite side (13/25 versus 19/31; RR 0.85, 95% CI 0.53 to 1.36) (see Analysis 1.12). Two participants of the surgical group had no subtalar movement versus none in the conservative treatment group.
Thordarson 1996 found no significant difference in the mean subtalar range of motion (20º versus 17º) in the subset of 11 participants (8 surgery versus 3 conservative) who underwent physical examination. Ankle range of motion was recorded as normal or mildly restricted in these 11 patients and no differences were observed in ankle/hindfoot stability (data values not reported).
Time to union (bone healing)
None of the trials reported data on time to bone union. Thordarson 1996 reported that all fractures had healed in both treatment groups at follow-up.
All four trials recorded radiological measurements, including Bohler's angle, residual displacement of posterior facet and radiographic signs of subtalar arthrosis. These data are presented in Table 2. Data for radiological signs of subtalar arthritis presented in Analysis 1.13 show no difference between the two groups at either 80 weeks follow-up (Chrintz 1993) or 15 years follow-up (Parmar 1993).
Buckley 2002 reported mean Bohler's angle values; however, it was unclear whether these values were measured pre- or post-treatment. Subgroup data for quality of fracture reduction (anatomic reduction, step-off of < 2 mm, comminuted reduction) were reported by mean VAS score in the operative group only. Another publication for this trial, reporting interim data for 88 participants, found no differences in Bohler's angle between treatment groups.
Chrintz 1993 found no differences in proportion with radiographic signs of subtalar arthrosis (12/26 versus 12/26) at median of 80 weeks follow-up. The authors stated that the degree of dislocation and Bohler's angle were significantly improved after surgical treatment; however, values were not reported.
Thordarson 1996 reported radiological results for a subgroup of 11 patients. They found a statistically significant improvement in the Bohler's angle in surgical participants between baseline and post-treatment (mean angle improved from 11º to 26º postoperatively; P = 0.001), compared with a decrease from 9º to 8º in those treated conservatively (P value not reported). Although mean residual displacement of the posterior facet values were reported, it was unclear at what time point radiographs were assessed (or whether on the subgroup of 11 participants assessed by an independent examiner at follow-up).
Radiological outcome data were only available for the 15 year follow-up of Parmar 1993. At 15 years after treatment, authors reported a trend to better Bohler's angle after surgery compared with conservative treatment (16.9º versus 10.4º; P = 0.07). No statistically significant differences were found between the two groups in calcaneal height (mean 38.2 mm versus 38.2 mm respectively; P = 0.57) nor in those with a moderate or severe grade of subtalar arthritis (14/15 versus 10/11; RR 1.03, 95% CI 0.82 to 1.29) (see Analysis 1.13).
The findings of this review, based on four studies involving 602 participants, found no robust evidence of improved functional or clinical outcomes in adults managed with surgery or conservative treatment for displaced intra-articular calcaneal fractures. To date, only one large multi-centre study has compared surgery with conservative treatment; this moderate quality Canadian trial recruited 424 participants and found no difference between clinical and patient-reported outcomes. Other smaller poor quality studies demonstrated a trend for improved outcomes after surgery but this, drawing from the results of the large trial, will be at the expense of a higher risk of postoperative complications.
Summary of main results
Four trials comparing surgical with conservative interventions for displaced intra-articular calcaneal fractures were included in the review. Overall, the trials were small with intermediate follow-up ranging from one to two years after treatment. Chrintz 1993 was incompletely reported in conference abstracts. Long-term follow-up at 15 years after treatment was reported by Parmar 1993. However, this trial reported late outcomes on a small subset (33%) of the original randomised cohort, also used different questionnaires at follow-up from earlier timepoints (Parmar 1993). These data should be interpreted with caution.
The largest trial (424 participants) found no difference in functional outcomes (composite functional score including daily activity, walking an other items) between treatment groups (plate or wire fixation versus conservative treatment) (Buckley 2002). The small study by Thordarson 1996 (30 participants), reported improved walking ability and distance walked after plate fixation and early mobilisation compared to conservative treatment. For the two remaining studies, using K-wires and plaster cast mobilisation (Parmar 1993) or Steinmann pin fixation (Chrintz 1993), there were no significant differences in functional outcomes between surgical and conservative treatment.
Data were pooled for two outcomes: chronic pain and return to work. Data from two small studies showed no difference in the risk of developing chronic pain after treatment. Pain is an important outcome as it will impact upon return to work and overall satisfaction with clinical treatment. Although a trend was observed whereby a greater proportion of surgical patients returned to the same work as before injury compared to conservative treatment, this was not statistically significant. These studies were small (total 61 participants) and for one study (Parmar 1993), data were only reported for employed men rather than the full sample. Similarly there were no differences in the proportion of participants returning to any type of work. Although used inconsistently across the studies, return to work is an important outcome as it probably represents successful treatment from a functional perspective.
There is often a poor correlation between radiological and clinical outcomes but patient-centred outcomes (function, pain and return to work) are more important than radiological results. One trial attempted to address satisfaction using a composite VAS scoring system; however, the scale was interpreted differently in separate publications: e.g. full VAS scale (Buckley 2002) versus a subset of seven gait items from the full VAS scale, later referred to as the 'Gait Satisfaction' scale (O'Brien 2004). The largest trial by Buckley 2002 found no differences in health-related quality of life at follow-up after surgery compared with conservative treatment.
Another important patient-reported outcome is ability to wear the same shoes as before injury. No differences were found in wearing of same footwear as before injury, although in the small study (26 participants) by Thordarson 1996, a greater proportion of surgical participants could wear all shoes comfortably compared with after conservative treatment.
There was good evidence to suggest a higher rate of major complications after surgery compared to conservative treatment, with a 44% increased risk observed in surgical participants (Buckley 2002). Surgical site infections were the most common postoperative complication, with up to 17% developing either a superficial or deep SSI. Only one study assessed reoperation or subsequent operation. Buckley 2002 reported that a greater proportion of conservative participants went on to have subtalar arthrodeses for subtalar arthritis. However, these patients were incorrectly excluded from the analyses. Of the three studies reporting range of movement, no significant differences were found in degree of movement in the ankle or subtalar joint at follow-up (data not pooled).
Although all four trials reported radiological outcomes, variation in timing and type of assessment (e.g. measurement of Bohler's angle, anatomical reduction, signs of subtalar arthrodesis, degree of residual displacement) meant it was not possible to pool data. None of the studies reported statistically significant differences between radiological outcomes after surgery compared with conservative treatment.
Overall completeness and applicability of evidence
Of the 602 participants randomised to treatment, outcome data were available for 443 (74%) participants. Losses to follow-up were relatively low for studies conducting follow-up within two years of treatment although this varied by outcome. Few studies achieved high rates of return for clinical examination.
The aim of surgical management for displaced intra-articular fractures is anatomical reduction, stable fixation and early joint mobilisation (Ibrahim 2007). Given the timeline of the conducted studies, it is very likely that surgical technique may have changed over time. The earliest trial was conducted over 40 years ago (participants recruited from 1977 to 1979; Chrintz 1993) and the last participant recruited was in 1997 (Buckley 2002), over 15 years ago. Two studies used similar surgical procedures: open reduction and internal fixation with plates and screws, using a lateral approach (Buckley 2002; Thordarson 1996). Surgical technique was well described and also standardised within the multi-centre Canadian trial (Buckley 2002). However, one surgeon conducted 73% of all procedures within this study. These authors also reported subsets of data across multiple publications, one of which found that when workers compensation cases were excluded from the sample, analysis suggested that outcomes were improved after surgery in particular subgroups: e.g. young men, aged 20 to 29 years, women, those with a lower Bohler's angle or comminuted fracture preoperatively. However, these benefits may be at the expense of higher complication rates.
As with other surgical procedures, there may be a relationship between volume of surgery conducted and outcome. It would be expected, as noted in an online commentary to the Buckley 2002 paper (Sangeorzan 2002), that higher volume surgeons might have better surgical results, but even so Buckley 2002 only reported benefits of surgery in specific groups of patients within multiple subgroup analysis.
Quality of the evidence
Overall the quality (and quantity) of the evidence was poor. Three trials were small with overall sample sizes of less than 80 participants; data were only available from two of these. Particular methodological issues related to randomisation, inadequate concealment of allocation (Parmar 1993 was quasi-randomised), lack of intention-to-treat analyses and lack of blinding amongst outcome assessors.
Bias associated with incomplete outcome data, reflecting a large loss to follow-up or post-randomisation exclusions, was a major problem. In general, data were not fully reported, relevant outcomes were not measured or were often misinterpreted (e.g. health-related quality of life was referred to as 'patient satisfaction' in Buckley 2002). Additionally, straightforward presentation of primary outcome values (e.g. numerators and denominators, SDs) was lacking. For Buckley 2002, subgroup analyses were conducted and presented across multiple publications, resulting in differing sample sizes, times to follow-up, and reanalysis of primary outcomes. Authors were contacted to identify variance data for key primary outcomes but unfortunately these were unavailable.
Potential biases in the review process
The search strategy was comprehensive and we identified two abstracts that had not been retrieved by previous systematic reviews (Chrintz 1993). For our analyses, we did not pool estimates for data where there was evidence of clinical heterogeneity or risk of attrition bias. Unavaoidable bias may have arisen from our interpretation of trial methods and imputation of incompletely reported results (such as percentages in Parmar 1993); however, none of the analyses contradict the findings in the trial reports.
Agreements and disagreements with other studies or reviews
The findings from our review concur with findings from previous reviews. The original Cochrane review concluded there was scant trial-based evidence for the effectiveness of surgical treatment of calcaneal fractures and the possible complications of surgery (Bridgman 2002). We excluded one non-randomised study (O'Farrell 1993) that was included in the previous review. At that time, findings from the large Canadian trial were anticipated (Buckley 2002). In addition to data from Buckley 2002, our review identified two additional abstracts (Chrintz 1993) that have not been included in other reviews (Bridgman 2002; Gougoulias 2009). We await data from the recently closed UK multi-centre study (UK heel fracture trial).
Implications for practice
Although uncommon, calcaneal fractures typically occur in younger adults, often of working age, and recovery is prolonged with considerable associated disability and delay in return to work or enforced change of occupation. There is clinical controversy over whether displaced intra-articular fractures should be treated surgically. The evidence from this updated systematic review, largely driven by one large multi-centre trial of moderate quality, suggests there is little difference between surgical and conservative treatment in terms of long-term function and quality of life. Based on data from two small poor quality trials, patients may be more likely to return to the same job as they had pre-injury after surgery. However, there is a greater risk of major complications after surgery; with the exception of subtalar arthrodeses for the development of subtalar arthritis, which favoured surgery. Overall, there is insufficient high quality evidence relating to current practice to establish whether surgical or conservative treatment is better for adults with displaced intra-articular calcaneal fracture.
Implications for research
It is possible that surgical management of displaced intra-articular calcaneal fracture may offer improved function and quicker return to work, but the evidence from published trials is insufficient to confirm or refute this hypothesis. The shortcomings of the present evidence base include small study bias, selection bias, performance bias, detection bias, and attrition bias. Additionally, reporting of the methods and results in the included trials did not meet best contemporary standards, nor current standards (CONSORT statement).
Further good quality research is clearly justified. There is one UK multi-centre study that has recently closed (UK heel fracture trial; 150 participants). Results from this study are expected in 2013. A Dutch multi-centre trial was discontinued in January 2012 because of difficulties with hospital and participant recruitment (Schepers 2012). Despite conducting a national survey to estimate incidence of heel fracture in order to accurately estimate patient throughput, the Dutch research group found that recruitment was much lower than expected with some hospitals preferring surgery over conservative treatment (personal communication; July 2012). This shows the challenge of implementing a clinical trial to evaluate the effectiveness and cost-effectiveness of alternative treatment strategies when a particular intervention(s) becomes widely accepted and ingrained within clinical practice. It would thus be prudent to assess the need for further trials, and what form they should take, after an update of this review that incorporates new evidence from the current UK heel fracture trial.
Given the often poor correlation between radiological and clinical outcomes, key patient-centred outcomes should be measured in future studies: e.g. validated scales to capture function, quality of life, patient satisfaction, return to work and complications of treatment. These should be captured in the short, medium and long term after treatment (e.g. immediately after injury, one year, five years). A key requirement of future studies is the simple presentation of key descriptive data for primary outcomes to allow data pooling (mean, standard deviation or median, inter-quartile range). Clear reporting of number of events by time to follow-up would also allow calculation of risk and rate ratios.
The authors would like to thank Helen Handoll, Nigel Hanchard, Mario Lenza, John McKinley and Janet Wale for their hard work in editing and reviewing the protocol and review. We would like to thank Joanne Elliott and Lesley Gillespie for developing the search strategies, and Lindsey Elstub and Amy Kavanagh for editorial support.
We would like to thank Dr Santosh Baliga for his help with the protocol.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
Appendix 1. Search strategies
The Cochrane Library (Wiley Online Library interface)
#1 MeSH descriptor Calcaneus explode all trees (95)
#2 MeSH descriptor Subtalar Joint (17)
#3 (#1 OR #2) (109)
#4 MeSH descriptor Fractures, Bone explode all trees (3129)
#5 (#3 AND #4) (39)
#6 (calcan* near/2 fracture*):ti,ab (46)
#7 (calcis near/2 fracture*):ti,ab (1)
#8 (#6 OR #7) (47)
#9 (#5 OR #8) (62)
MEDLINE (Ovid interface)
1. Calcaneus/ (4930)
2. Subtalar Joint/ (844)
3. or/1-2 (5566)
4. exp Fractures, Bone/ (122426)
5. and/3-4 (1697)
6. ((calcan$ or calcis) adj5 fracture$).tw. (1385)
7. or/5-6 (1991)
8. randomized controlled trial.pt. (311155)
9. controlled clinical trial.pt. (82831)
10. randomized.ab. (217312)
11. placebo.ab. (126412)
12. clinical trials as topic.sh. (155333)
13. randomly.ab. (157067)
14. trial.ti. (93004)
15. or/8-14 (722471)
16. exp animals/ not humans.sh. (3614789)
17. 15 not 16 (666825)
18. 7 and 17 (68)
EMBASE (Ovid interface)
1. Calcaneus Fracture/ (1189)
2. Subtalar Joint/ (1285)
3. Intraarticular Fracture/ or Joint Fracture/ (1106)
4. and/2-3 (41)
5. ((calcan$ or calcis) adj5 fracture$).tw. (1666)
6. or/1,4-5 (1992)
7. Clinical trial/ (810400)
8. Randomized controlled trial/ (282042)
9. Randomization/ (53175)
10. Single blind procedure/ (13680)
11. Double blind procedure/ (99051)
12. Crossover procedure/ (29993)
13. Placebo/ (180511)
14. Randomi?ed controlled trial$.tw. (61130)
15. RCT.tw. (7099)
16. Random allocation.tw. (1025)
17. Randomly allocated.tw. (15063)
18. Allocated randomly.tw. (1674)
19. (allocated adj2 random).tw. (682)
20. Single blind$.tw. (10785)
21. Double blind$.tw. (115798)
22. ((treble or triple) adj blind$).tw. (237)
23. Placebo$.tw. (155955)
24. Prospective study/ (165114)
25. or/7-24 (1119553)
26. Case study/ (12455)
27. Case report.tw. (202029)
28. Abstract report/ or letter/ (782726)
29. or/26-28 (993346)
30. 25 not 29 (1086632)
31. limit 30 to human (990368)
32. and/6,31 (127)
Appendix 2. Assessment of risk of bias
Each of the following aspects were recorded as yes, no or unclear.
Was the allocation sequence adequately generated: e.g. coin toss, random number tables, computer generated, other?
Was allocation adequately concealed in a way that would not allow both the investigators and the participants to know or influence the intervention group before an eligible participant is entered into the study: e.g. central randomisation, or sequentially numbered, opaque, sealed envelopes?
Incomplete outcome data
Were incomplete outcome data adequately addressed? Incomplete outcome data essentially include: attrition, exclusions and missing data. If any withdrawals occurred, were they described and reported by treatment group with reasons given? Whether or not there were clear explanations for withdrawals and dropouts in treatment groups was recorded. An example of an adequate method to address incomplete outcome data is the use of intention-to-treat analysis (ITT).
Selective outcome reporting
Are reports of the study free from suggestion of selective outcome reporting? This was interpreted as no evidence that statistically non-significant results might have been selectively withheld from publication: e.g. selective under reporting of data, or selective reporting of a subset of data.
Other sources of bias
Was the study apparently free of other problems that could put it at a high risk of bias: e.g. baseline imbalance, or the use of an insensitive instrument to measure outcomes?
Details of blinding participants, personnel (surgeons) and outcome assessors were assessed. This was recorded as: yes, no or not possible, or unclear.
Quality assessment/internal validity
Quality assessment criteria were categorised as low, unclear or high risk of bias according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2009).
Contributions of authors
The review was conceived by AGS. The review was undertaken by JB and AGS. JB is the guarantor of the review. Both authors have contributed to the writing and revision of the review.
Declarations of interest
Sources of support
- University of Aberdeen, UK.
- University of Warwick, UK.Warwick Clinical Trials Unit
- No sources of support supplied
Differences between protocol and review
1. There was a change in authorship.
2. An additional secondary outcome was added: Objective measures of impairment (e.g. range of ankle movement).
Medical Subject Headings (MeSH)
Calcaneus [*injuries]; Fracture Fixation [*methods]; Intra-Articular Fractures [*therapy]; Pain Measurement; Quality of Life; Randomized Controlled Trials as Topic; Recovery of Function; Return to Work; Shoes; Walking
MeSH check words
* Indicates the major publication for the study