Background

Fragility fractures result from low-energy trauma, usually a fall from standing height or lower. Each year 300,000 people attend UK NHS hospitals with a fragility fracture related to bone insufficiency in older age [1]. This represents a major health, social and economic problem, with an estimated annual cost of £1.8 billion [2]. Lower limb fragility fractures can have a devastating impact, resulting in mobility problems and loss of independence [3].

A core component of rehabilitation after fragility fracture is exercise prescription. A previous systematic scoping review of exercise prescription for people with any type of fragility fracture included studies up to 2009 [4]. While the scale of that review provided a comprehensive overview of exercise interventions at the time, an updated and more focussed systematic review was indicated to inform the development of future interventions for this patient group.

To the best of our knowledge no reviews to date have examined the quality of intervention reporting in trials involving people with lower limb fragility fractures. In other areas of exercise rehabilitation, limitations in reporting that prevent replication in other trials or implementation into clinical practice have been identified [5, 6]. It is therefore important to identify not only what exercise interventions have been assessed but also to establish if reporting of lower limb fragility fracture trials have similar issues in reporting quality, and if so, what areas of reporting are in greatest need of improvement to enable replicability and implementation. Exercise targets improvement in mobility after lower limb fragility fracture and this is a core outcome domain in this patient group, [7] therefore it is also important to identify what outcome measures have been used.

The overall purpose of our review was to provide evidence to guide future exercise intervention development and evaluation for people with pelvic and lower limb fragility fractures and to highlight areas of study design and intervention reporting that could be enhanced to improve the quality, replicability and implementation of future trials. Our aims were to identify the types of exercise interventions that have been tested in randomised clinical trials, investigate the reporting quality of exercise interventions, describe which mobility outcome measures have been used, and evaluate the risk of bias in the trial design and conduct.

Methods

This systematic review was registered on the PROSPERO database (https://www.crd.york.ac.uk/prospero/display_record.php? ID = CRD42017060905) and reported according to PRISMA guidance [8].

Eligibility

Types of studies

Randomised controlled trials or quasi-randomised controlled trials were considered eligible.

Types of participants

Studies involving adults (50 years or older) within one year of a pelvic or lower limb fracture initially treated surgically or conservatively were included. Studies were excluded if participants were younger (aged under 50 years old), unless separate data for older adults were available, or the proportion of younger adults was small (less than 10%) and, preferably, numbers balanced between the groups.

Types of interventions

Trials comparing different prescribed exercise regimes against each other, or prescribed exercise versus a comparator intervention such as rest, immobilisation in a brace, cast or splint, advice only, or ‘usual care’ were eligible. Exercise prescription encompassed planned physical activity, exercise or active rehabilitation prescribed by a physician, physical therapist or occupational therapist, or other allied health professional [4].

Types of outcomes

We extracted data on which outcome measures of mobility were used in the trials both in terms of subjectively assessed measures of mobility (e.g. Lower Extremity Functional Scale) and objective clinical measures of mobility (e.g. timed walking tests). Duration and timing of follow-up were also extracted.

Search strategy for identification of studies

We searched the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, EMBASE, and the Physiotherapy Evidence Database (PEDro). We did not apply language restrictions to the searches. Studies published in 1996 or later were included. Searches were completed April 2019 and updated in MEDLINE and EMBASE in July 2019. Reference lists of included trials were checked for potentially eligible studies. An example search strategy is available in the online supplementary file.

Selection of studies

Two reviewers independently screened the titles and abstracts using Covidence software (Covidence, Australia). We obtained full reports of potentially eligible studies, and both reviewers independently performed study selection. If agreement was not achieved by discussion at any stage, a third review author adjudicated. Articles for inclusion were limited to those written in English and published in academic journals.

Data extraction

One author extracted data using a standard data extraction form and a second author checked the extracted data against the source while tabulating the data. The data extraction form was piloted and then modified. The following information was systematically extracted: sample size, sample demographics (age, sex, injury characteristics, time since injury), detailed descriptions of the interventions (including setting, timing, care personnel involved, training, equipment used, weight-bearing, prescription of walking aids, and the type and prescription of exercises used, and assessment of adherence), and the specified mobility outcome measures.

Assessment of risk of bias in included studies

Two review authors independently assessed the risk of bias using Cochrane’s Risk of Bias tool [9]. We used the following domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting. Disagreements were resolved by discussion.

Intervention reporting

Reporting quality for the interventions was based on the TIDieR [10] guidance for reporting complex interventions. The quality of intervention reporting was assessed by one reviewer and verified by a second reviewer. Disagreements were resolved by discussion. The criteria for the assessments are shown in Table 1.

Table 1 Intervention reporting assessment criteria based on TIDieR [10] recommendations

Data synthesis

A narrative synthesis was undertaken and interventions were grouped by exercise and fracture type. Characteristics of studies were summarised as counts and percentages for categorical data and medians with interquartile ranges for continuous data.

Changes to protocol

The review focussed on intervention content and reporting quality as these have not been previously assessed in sufficient detail to inform the design and conduct of future trials. The originally planned focus on effectiveness and quantitative meta-analysis was not conducted as this became beyond the scope of resources for the study, and effectiveness meta-analyses are available [11].

Results

Study selection and characteristics

Figure 1 outlines the identification, screening, and inclusion of studies. Searches identified 6308 records. After removal of duplicates, the titles and abstracts of 6016 records were screened. Of these, 184 full-text articles were assessed, and 66 articles reporting 37 trials were eligible.

Fig. 1
figure 1

PRISMA flow diagram

Characteristics of included studies

Of the 37 included trials, most were conducted in Australia or the USA (18/37, 49%). Trial designs were mostly parallel group (35/37, 95%) with two intervention groups (31/37, 84%), see Table 2 for detailed study characteristics. In total, 3565 participants were randomised across the 37 trials, with a median sample size of 81 (IQR 48 to 124). In 32 trials that provided adequate baseline characteristic data, participants were aged a median of 81 years (IQR 79 to 82) and 76% (2536/3356) were female. All trials focussed on people with a hip fracture except one ankle fracture trial [12] that reported results for a subgroup of participants aged more than 50 years.

Table 2 Study characteristics

Interventions

A range of exercise types were assessed (see Tables 2 and 3), including 14 focussing on resistance exercise, five on weight bearing exercise, 13 varied the dose of sessions with health professions, and two each focussed on treadmill training, timing of weight bearing, or aerobic exercise. These main types of intervention were often combined with other types of exercise, and compared to diverse control interventions (see Table 3).

Table 3 Exercise interventions and comparators across included studies

The setting of exercise intervention delivery was 11 for inpatients, six for outpatients, 13 for community, six were a combination, and for one trial it was unclear what the setting was.

Outcomes

Subjectively assessed mobility outcome measures were used in 22/37 (59%) studies and 29/37 (78%) used an objective mobility measure. There were no common outcome instruments used across the trials. The most frequently used instruments were the Timed Up and Go test (11 trials) and gait speed (11 trials). The length of follow-up was a median of 6 (IQR 2.5 to 12) months.

Risk of bias within included studies

Risk of bias assessments are shown in Table 4. Within the limitations of reporting, it was judged that 30/37 (81%) trials had adequate sequence generation and 25/37 (68%) had sufficient allocation concealment. 10/37 (27%) of trials were at high risk of bias due to a lack of outcome assessor blinding.

Table 4 Risk of bias assessments

Reporting quality of interventions

Of the 37 included trials there were 65 different exercise intervention groups and 16 non-exercise or inactive control comparator groups (see Table 5 for reporting quality assessments). Of the 65 exercise interventions, reporting was judged as being clearly described for 33 (51%) when treatment started after injury, 61 (94%) for where it was delivered, 49 (75%) for who delivered it, 47 (72%) on whether delivered as group or individual, 29 (45%) for the duration of the intervention, 46 (71%) for session frequency, 8 (12%) for the full prescription details to enable the intervention to be reproduced, 32 (49%) clearly reported tailoring or modification, and 23 (35%) reported exercise adherence in the trial. Of the six comparator usual care exercise interventions, only one had more than half of the intervention reporting criteria assessed as being clear.

Table 5 Intervention reporting assessment for each included study*

Discussion

A range of exercise types have been investigated for pelvic and lower limb fragility fractures, with most trials investigating resistance exercise or higher doses of sessions with a health professional. To date deficiencies in reporting of the exercise interventions hamper reproducibility of the interventions, especially in terms of the specific details on how exercises were prescribed. Reporting of usual care exercise comparator interventions was poor. Details on exercise prescription that were most often missed related to the movements performed in the exercises, sets and repetitions for resistance exercises, duration for aerobic exercises, and exercise loading or intensity. Adoption of the TIDieR [10] checklist for reporting complex interventions should improve reporting of future trials. TIDieR was published in 2014, prior to all but five of the 37 trials included in this review. Supplementary use of the Consensus on Exercise Reporting Template (CERT) [49] is also indicated as these guidelines additionally target the main deficiencies in reporting identified in our review. It is important to recognise that the problems with exercise intervention reporting in pelvic and lower limb fragility fracture trials are consistent with other fields of rehabilitation so these issues are not isolated [5, 6].

One key area of trial design and conduct that could be improved upon in future trials is the blinding of outcome assessors as this was inadequate in 27% of trials and this could be rectified without significant additional resource burden. Blinded outcome assessors are arguably crucial given that the nature of exercise makes it self-evident what intervention is being received, as reflected in our finding that no trial had a low risk of bias assessment for blinding of participants and personnel.

With one exception, all exercise trials for adults with a pelvic or lower limb fragility fractures have been focussed on hip fracture. There is a significant burden from other non-hip fragility fractures as they often require hospitalisation and result in long-term disability, [50] therefore further research for people with pelvic and other lower limb fragility fractures is also needed. Even though most trials have focussed on hip fracture, reflecting their proportionately greater health and socio-economic impact, Sheehan and colleagues [51] have highlighted that rehabilitation trials in this patient group have underrepresented participants with cognitive impairment and nursing home residents, therefore trials focussing on other populations are also indicated.

Previous reviews have included meta-analyses to assess the effectiveness of different exercise interventions [11]. The pooling of outcomes from these trials could be problematic in the context of the intervention heterogeneity and reporting quality limitations outlined in this review. Dealing with heterogeneity in intervention components is a common challenge in quantitative synthesis of complex interventions. One approach that enables an assessment of intervention components is meta-regression, as employed by Diong and colleagues in a review of hip fracture exercise trials, [52] however, there was heterogeneity in the comparator interventions in some of the pooled studies, and there is ongoing debate as to what extent these analytical approaches manage evident clinical variations in intervention components that can interact [53].

Mobility-specific subjective and objective outcome measures were included in 59 and 78% of trials respectively but it is evident within our review that there is inconsistency in the outcome instruments used. The degree of heterogeneity in outcome measure instruments would make quantitative synthesis problematic. Further consensus work towards a core outcome set for rehabilitation trials for people with pelvic and lower limb fragility fractures would therefore be valuable.

This review has some limitations. We included English language and published literature only, meaning that some relevant studies may have been missed. Data extraction and reporting quality was not completely repeated independently by a second reviewer due to the resource limitations of the study. However, a second reviewer did verify these data against the source and any discrepancies corrected in discussion. Finally, as there was no specific intervention reporting quality assessment tool, a review specific assessment was developed drawing on the TIDieR reporting guidelines. A tool for these purposes would be valuable for future research but findings from our assessments provided some clear areas of focus for improving reporting in future exercise trials.

Conclusion

All exercise trials for adults with a pelvic or lower limb fragility fractures have been focussed on hip fracture, apart from one ankle fracture trial. Research for people with pelvic and other lower limb fragility fractures is indicated. A wide range of exercise types have been investigated but to date deficiencies in reporting of the interventions hamper the reproducibility of the interventions, especially in terms of the specific details on how exercises were prescribed. Use of TIDieR and CERT reporting guidelines for future trials will likely improve intervention reporting. Trials of exercise interventions would also be improved by consistent use of blinded outcome assessors and with further consensus on which mobility outcomes should be assessed.