Exercise or manual physiotherapy compared with a single session of physiotherapy for osteoporotic vertebral fracture: three-arm PROVE RCT

Definition and diagnosis of osteoporosis

The World Health Organization describes osteoporosis as a progressive systemic skeletal disease ‘characterized by low bone mass and microarchitectural deterioration of bone tissue, leading to enhanced bone fragility and a consequent increase in fracture risk’. 6 Osteoporosis is defined in terms of bone mass, based on bone mineral density (BMD) assessment. It is diagnosed in postmenopausal women or in men aged ≥ 50 years when BMD lies ≥ 2.5 standard deviations (SDs) below the mean value for a healthy young adult (a T-score of ≤ –2.5 SD) at the same skeletal site. 7 The T-score is calculated from the measurement of BMD using central (hip and/or lumbar spine) dual-energy X-ray absorptiometry (DEXA) scanning. 8 Established or severe osteoporosis is a BMD T-score of < –2.5 SD plus the presence of one or more fragility fractures. 9 Although this operational definition of osteoporosis focuses on BMD, the wider definition of osteoporosis emphasises that skeletal fragility is the product of decreased BMD and poor bone quality (i.e. the deterioration in bone architecture and microstructure associated with osteoporosis). 5

Osteoporosis can be classified as either primary or secondary. Primary or idiopathic osteoporosis includes postmenopausal osteoporosis, juvenile osteoporosis and age-related, senile osteoporosis and is the most common type of osteoporosis. 5 It affects both sexes, but women are at greater risk of developing primary osteoporosis because, on average, they have lower peak bone mass than men, they live longer and the lack of oestrogen after the menopause accelerates the normal rate of bone loss and deterioration in bone structure that occurs with ageing. 5 In contrast, secondary osteoporosis occurs because of other conditions or diseases, such as rheumatoid arthritis, endocrine disorders (diabetes mellitus, hypogonadism, etc.), cancer and gastrointestinal diseases (coeliac disease, inflammatory bowel disease, etc.), or because of medications that affect bone metabolism (e.g. glucocorticoids4). The clinical significance of osteoporosis lies not in low bone mass itself, but in the fractures that may happen as a consequence.

Epidemiology of osteoporosis and osteoporotic fracture

Osteoporosis is a common disorder, affecting hundreds of millions of people worldwide and > 3.2 million people in the UK. 10 Because bone density reduces with age and the global population is ageing, the prevalence of osteoporosis is increasing. 10,11 For example, the number of women living with osteoporosis in the UK has been predicted to increase from 1.8 million in 2010 to 2.1 million by 2020 (+16.5%). 12 Osteoporosis itself is often called a ‘silent disease’ as patients cannot perceive that their bones are getting thinner and this process is not painful or limiting. 13 The clinical importance of osteoporosis arises from its association with bone fractures. 8

In the osteoporotic spine, vertebral fractures often happen without any noticeable stress. 14,15 Rates of osteoporotic fracture are increasing in line with the increase in osteoporosis. 12 In the UK in 2010, the estimated number of osteoporotic fractures was 536,000; this number included approximately 79,000 hip fractures, 66,000 new clinically diagnosed vertebral fractures and 69,000 forearm fractures. 10 This is expected to rise to 682,000 osteoporotic fractures by 2025, an increase of 27% over 15 years. 10

Costs of osteoporosis and osteoporotic fracture

Osteoporosis incurs substantial personal and economic costs. In Europe, the costs associated with osteoporosis are greater than those caused by all cancers (apart from lung cancer) and many other chronic diseases including asthma, stroke, heart disease and rheumatoid arthritis. 3,13 Osteoporotic fractures result in significant excess mortality and morbidity. 8,10,12 The consequences vary by fracture site, with the greatest excess mortality and morbidity attributed to hip and vertebral fractures. 13 Hip fractures are associated with the highest mortality in the year after fracture, with vertebral fractures being the second most important site for fracture-related deaths. 13 Both hip and vertebral fracture result in marked reductions in quality of life (QoL) that persist for ≥ 18 months after fracture. 16

The economic burden of osteoporosis is a combination of direct costs related to fracture care and osteoporosis management (e.g. surgery and hospitalisation, diagnostic scans, therapy and pharmaceuticals), as well as indirect costs (e.g. from lost work time and productivity for the individual and any caregiver) and more intangible costs in terms of the value of the quality-adjusted life-years (QALYs) lost. 13 In 2010, the cost of osteoporosis-related care to the NHS was estimated to be £4.4B, of which acute fracture care accounted for the largest proportion. 10 By 2025, it is estimated that this figure will rise to £5.5B per year. 10 These may be conservative estimates. Burge et al. 17 highlight the difficulties of analysing the total cost of osteoporosis, noting that indirect costs, such as unpaid care time or longer-term, ongoing care provided by Personal Social Services, may be unaccounted for and result in underestimating the total cost of osteoporosis. 17

Lifestyle recommendations

Guidelines recommend that people with osteoporosis and at high risk of osteoporotic fractures are provided with advice about osteoporosis and a range of lifestyle measures to promote bone health and prevent falls. 9,30 These include ensuring adequate calcium and vitamin D intake through diet or with supplementation, stopping smoking and avoiding excessive alcohol consumption. 8,9,30 They detail the value of physical activity, of undertaking regular weight-bearing exercise and muscle strengthening exercise. 4,8,9,30 Over the last 3 years, information accessible to patients (via osteoporosis patient societies etc.) to support exercise recommendations has become more comprehensive. 30,31

Pharmacological treatment

Pharmacological bone-protective medications are central to the management of osteoporosis. Initiated when a high risk of fracture has been identified, the main aim of treatment is to reduce fracture risk. 8 Most work by regulating bone remodelling and may increase BMD [i.e. by inhibiting or slowing down the resorption of old bone (antiresorptive drugs) or by stimulating the osteoclast cells that build new bone (anabolic drugs)]. 32 High-quality evidence exists to support the effectiveness of most major osteoporotic medications in preventing fractures in postmenopausal women, but few data are available about potential benefits or side effects in men. 4,8 Medication is often more effective at preventing OVF than other types of fracture. 2,8

Frequently prescribed, mainly antiresorptive drugs are the bisphosphonates [alendronate (Fosamax®; Merck Sharp & Dohme, Kenilworth, NJ, USA), risedronate (Actonel®; Procter & Gamble Pharmaceuticals, Cincinnati, OH, USA), ibandronate (Boniva®, Genentech Inc., San Francisco, CA, USA, a subsidiary of F. Hoffmann-La Roche AG, Basel, Switzerland) and zolendronate (Reclast® and Zometa®; Novartis International AG, Basel, Switzerland], denosumab (Prolia®; Amgen Inc., Thousand Oaks, CA, USA), which is a monoclonal antibody against the major mediator of bone resorption, and raloxifene (Evista®, Eli Lilly and Company, Indianapolis, IN, USA), which is a selective oestrogen receptor modulator. 32 Common anabolic drugs include parathyroid hormone treatments such as teriparatide (Forteo®, Eli Lilly and Company, Indianapolis, IN, USA). 32 Previously widespread, hormone therapies are also available but are now used only in limited circumstances because of concerns about side effects. 8 The different osteoporotic medications have varying routes of administration, different antifracture efficacy at different sites and different side effects and cost profiles. 4,8,32 Although medications are an essential part of management, they do not address pain or deficits in back extensor muscle strength, balance or posture to improve function or reduce the risk of fracture owing to falls.

Surgical interventions

Surgery is used for a minority of patients after acute OVF, predominantly when conservative treatment is unsuccessful (i.e. when patients have severe, unremitting pain and significant disability). 33,34 Vertebroplasty and balloon kyphoplasty are the two main, minimally invasive surgical techniques employed,33 with newer cement augmentation techniques developing. 34 Vertebroplasty involves the injection of acrylic bone cement into the fractured vertebral body under fluoroscopic imaging guidance. During kyphoplasty, a deflated balloon is inserted into the compressed vertebral body, the balloon is expanded to correct the vertebral deformity and create a cavity, then the void is filled with bone cement. 33 Initial reports about these techniques were controversial, with significant complications such as cement extrusion and fracturing of adjacent vertebrae being documented;33 however, more recent studies report more positive results. 34–36 Current reviews acknowledge the inconsistent messages about the benefits of surgery and highlight the need for robust criteria when selecting patients for these procedures and for further research in the area. 33–35,37

Inclusion criteria

• Patients who had a diagnosis of primary osteoporosis confirmed by a radiograph or DEXA scan (≥ 2.5 SD below the norm) at the lowest lumbar level.

• Patients who had a history of at least one symptomatic OVF.

• Patients aged ≥ 18 years.

• Patients who were postmenopausal (if female).

• Patients able to walk ≥ 10 m independently with or without an aid.

• Patients able to understand and participate in the physiotherapy programme.

Exclusion criteria

• Patients aged < 18 years.

• Patients who had osteoporosis secondary to other metabolic bone disorders or disease (e.g. rheumatoid arthritis, cancer and osteomalacia), experienced lower-limb joint surgery, or fracture, in the previous 6 months.

• Patients whose primary problem was back pain with pain radiating into the lower limb.

• Patients who had undergone vertebroplasty, facet joint injection or any physical therapy (e.g. chiropractic, osteopathic or physiotherapeutic treatment for back pain in the previous 12 weeks).

• Patients with severe unstable cardiovascular or pulmonary disease or significant psychiatric or neurological conditions that would preclude participation in the physiotherapy treatment arms.

Screening and recruitment

Patients were approached by clinical staff during routine clinic visits. They were provided with a patient information sheet (see Appendix 13) and asked if they would consider participating. A research physiotherapist would then recontact them at a later date to book an appointment at a research clinic. Here, the trial information was discussed, the patient had an opportunity to ask questions, eligibility was rechecked and, if appropriate, patients were consented. Participants were asked to give written informed consent in accordance with the principles of Good Clinical Practice41 and the Declaration of Helsinki. 42 At the time of consent, outcome assessors collected baseline measures.

Settings and locations

The trial was run in 21 NHS trusts (names at time of participation):

1. Cambridge University Hospitals NHS Foundation Trust

2. Countess of Chester NHS Foundation Trust

3. East Sussex Healthcare NHS Foundation Trust

4. King’s College Hospital NHS Foundation Trust

5. Nottingham University Hospitals NHS Foundation Trust

6. Oxford University Hospitals NHS Trust

7. Portsmouth Hospitals NHS Foundation Trust

8. Royal Devon and Exeter NHS Foundation Trust

9. Royal Orthopaedic Hospital NHS Foundation Trust

10. Royal Surrey County Hospital Foundation Trust

11. Royal United Hospitals Bath NHS Foundation Trust

12. Sheffield Teaching Hospitals NHS Foundation Trust

13. Solent NHS Trust

14. South Devon Healthcare NHS Foundation Trust

15. Southend University Hospital NHS Foundation Trust

16. Staffordshire and Stoke-on-Trent Partnership NHS Foundation Trust

17. The Ipswich Hospitals NHS Trust

18. The Royal Bournemouth and Christchurch Hospitals NHS Foundation Trust

19. University College London NHS Foundation Trust

20. University Hospitals of Leicester NHS Trust

21. Western Sussex Hospitals NHS Foundation Trust.

Sequence generation

The variable block randomisation schedule was prepared by the trial statistician. Randomisation was stratified by centre to control for any confounding factors evident at local recruitment sites.

Allocation concealment

The research clinician telephoned the randomisation service, and only once the participant was registered in the trial was the random allocation generated; hence, allocation was concealed. Allocation was sent directly to the therapist and the participant was informed at the appointment or when their first appointment was arranged.

Blinding

The outcome assessor was blind to the allocation arm of the participant and was independent of intervention delivery. Participants were requested to not disclose their allocation to the outcome assessor. If an outcome assessor was unblinded, this was recorded. If they remained unblinded at each research visit, the assessors were asked to guess which allocation the participant had been given. The participants and therapists could not be blinded to the allocation arm.

Database and data processing

Project data were stored in a Microsoft SQL Server 2012 Enterprise Edition (version: 11.0.6251.0, Microsoft Corporation, Redmond, WA, USA); the database was hosted in the University of Warwick’s data centre and rules were imposed for data entry, which included valid range for responses, linked dates and participant identification numbers.

Data were single entered into the database by trial personnel. The trial statistician carried out checks of the plausibility of values, missing data and form return rates to enable further queries to be resolved prior to freezing data for scheduled DMEC reports and analysis.

For data quality assessment, 10% of all forms at all time points were randomly selected for data checking on a 2-monthly basis. All disagreements found when checking were corrected and any systematic faults found as a result of the checks were also corrected.

Data set access

The final data set was accessible to all trial members after data lock. The chief investigator assumed overall responsibility for the data report and had full access to the trial data set. There were no contractual agreements that limited access for investigators.

Scoring and missing items

The QUALEFFO-41 score and scores for individual domains were calculated in accordance with formulae given by the International Osteoporosis Foundation. 55 The total score was considered missing if all questions were missing, with domain scores considered missing if all questions were missing for that domain.

The EQ-5D-5L was used mainly for economic analyses evaluation, the methods and results of which are described in Chapter 6.

Interim analysis

The interim analysis decision rule was based on a comparison of the estimated mean change from baseline in the QUALEFFO-41 score for each of the three trial arms.

Although the integrity of the trial in terms of type I error rate control does not require prespecification of the decision rule to be used at the interim analysis, the below rule was proposed prior to the start of the trial to decide which intervention arm(s) should continue to be included along with the SSPT in the second stage of the trial. This rule ensured that the interventions would continue only if they were sufficiently promising relative to the SSPT, and that the most promising would be chosen to continue alone along with the SSPT unless both were sufficiently promising and appeared to be of similar efficacy, in which case both would continue:

• If the mean change from baseline of the QUALEFFO-41 score for an intervention arm was not > 0.5 points greater than that of the SSPT, that intervention arm would be dropped from the trial. Note that under this rule both intervention arms might be dropped, in which case the trial would be stopped because of futility.

• If the mean change from baseline of the QUALEFFO-41 score for one intervention arm was > 2 points higher than for that of the other arm, the intervention with the lower mean change from baseline would be dropped from the trial.

Demographic and baseline measurements

Demographic characteristics, clinical characteristics and baseline measurements were presented to evaluate the comparability of intervention arms and generalisability to clinical settings. Count data were presented as percentages and continuous data were presented as means and SDs in each arm.

Primary end-point analyses

The analysis of the co-primary end points of change in QUALEFFO-41 score and TLS test time from baseline at 12 months was conducted to allow for the adaptive design used and to allow for the multiple comparisons arising from the two pairwise comparisons between the comparator SSPT arm and the two experimental intervention arms.

Hypothesis tests were conducted using the method proposed by Bretz et al. ,56 modified by Friede et al. ,57 to allow for the use of an early end point, in this case the change in QUALEFFO-41 score from baseline to 4 months, at the interim analysis. Details of the method are given in Appendix 24. The analyses of 12-month change in QUALEFFO-41 and TLS test were based on linear models to test for treatment effects, adjusting for age (dichotomised into ≤ 70 years and > 70 years), number of baseline spinal fractures (dichotomised into two or fewer and more than two), recruitment centre and baseline score for either the QUALEFFO-41 or the TLS test.

Unbiased estimates of differences between each intervention arm and a SSPT, adjusting for the adaptive design used, were based on the method proposed by Stallard and Kimani,58 with corrected confidence intervals (CIs) calculated using the method proposed by Magirr et al. 59 These were presented alongside means from each arm unadjusted for baseline or centre and uncorrected for the adaptive design used.

Secondary end-point analyses

For all secondary end points other than number of falls and falls frequency, changes in secondary end points from baseline to 4 and 12 months (including 4-month changes in the co-primary end points) for the SSPT arm were compared with each intervention arm. Hypothesis tests and estimated treatment differences were obtained using linear models adjusting for baseline value and centre, and these were presented alongside unadjusted means from each arm.

The number of falls in the previous year recorded at 4 and 12 months was analysed using a Poisson regression model, adjusting for baseline number of falls and recruitment centre. The frequency of falls was dichotomised into rarely (one or fewer per year) and sometimes (one or fewer per month) or frequently (one or more per week) and analysed using a logistic regression model, adjusting for baseline frequency and recruitment centre.

Subgroup analyses

Analyses were conducted to investigate subgroup effects specified prior to the data analysis being conducted. The subgroups investigated were sex, age (≤ 70 years vs. > 70 years) and number of spinal fractures at baseline (two or fewer vs. more than two). In each case, linear models adjusting for baseline QUALEFFO-41 or TLS test score and centre were used to test for an intervention*baseline interaction effect and then to estimate and test treatment effects within each subgroup.

Sensitivity analyses

As a sensitivity analysis, we compared demographic and baseline characteristics for participants who had complete baseline and 12-month QUALEFFO-41 or TLS test data and for those for whom this was missing at the 12-month time point. We also compared data from the different recruitment centres.

Data from participants recruited in stage 1 (i.e. with data included in the interim analysis) and data from those recruited in stage 2 were also compared.

Complier-average causal effect analysis

A prespecified CACE analysis was conducted to determine if the level of compliance with the exercise programme affected participant outcomes as measured by the QUALEFFO-41 and TLS test (primary outcome). The CACE estimates were calculated using the instrumental variable method described by Dunn et al. 60 The model adjusted for baseline and recruitment centre, with CIs calculated using a bootstrap approach. No adjustment was made for the adaptive design used.

The analysis was conducted using thresholds of four or more sessions (partial compliance) and the maximum of seven sessions (full compliance). Estimates were obtained when ≥ 50% of participants in each of the intervention arms were classified as compliers.

Formal approvals

The PROVE trial was approved by South Central Research Ethics Committee in June 2012 (Research Ethics Committee reference 12/SC/0411) and by the research and development departments of each participating centre. The final approved trial protocol has been published. 39 A first substantial amendment was granted on 4 January 2013, in which the participant consent and information leaflets were reviewed and approved. A second substantial amendment was approved on 28 March 2014 to update the participant information leaflet, the reconsent form and the letter to participants. A third substantial amendment was granted on 10 September 2014 to update the letter to participants from general practitioners (GPs). A fourth substantial amendment was granted on 22 April 2015 to update the participant reply slip.

Trial Steering Committee

A Trial Steering Committee (TSC) was responsible for monitoring and supervising the progress of the PROVE trial towards the interim and final milestones. The TSC consisted of three independent experts, a lay member and leading members of the Trial Management Group (TMG). Membership of the TSC is detailed in Acknowledgements.

Data Monitoring and Ethics Committee

The DMEC was independent of the trial and was tasked with monitoring ethics, safety and data integrity. The trial statistician provided data and analyses requested by the DMEC at each of the meetings. Membership of the DMEC is detailed in Acknowledgements.

Trial Management Group

A TMG was responsible for the day-to-day management of the trial; it consisted of the chief investigator, research fellow, statistician and trial co-ordinator. They ensured the overall integrity of the trial, compliance with the protocols, the welfare of all participants and that the trial was appropriately reported.

Rationale for single session of physiotherapy

Guidelines for the management of osteoporosis and fragility fracture focus on areas such as diagnosis and prescription of bone-protective medication, but also recommend that those at high risk of osteoporotic fractures receive education about osteoporosis and lifestyle advice to promote bone health. 8,78,79 Physiotherapy is acknowledged as an important component of rehabilitation after fracture but no specific recommendations about physiotherapy treatment are provided. 78

The commissioning brief from the HTA 10/99 stipulated that the trial should include a control or comparator arm that consisted of simple advice, an attention control and secondary prevention of fractures in line with National Institute for Health and Care Excellence (NICE) guidance (Medication). 80 Prior to the trial, discussions with members of the FRiSCy network confirmed that relatively few people with OVF were referred for physiotherapy; most would be seen in primary care by their GP and started on bone-protective medication, some were given lifestyle advice, some might be provided with written information and some were directed to the ROS for information. Differences in usual care and education provided between clinicians and between areas were apparent. Consequently, to ensure that all participants received recommended education and advice a standardised intervention was developed as a comparator arm.

Rationale for manual therapy

Previous research suggests that there are multiple factors that contribute to the development of back pain and spinal deformity, such as hyperkyphosis in people with osteoporosis. 23,81,82 Physiotherapists use manual techniques, such as spinal mobilisations, soft-tissue mobilisation or taping, to address modifiable elements, such as soft-tissue restrictions and decreased muscle length, with the aim of providing pain relief, improving spinal motion and posture and physical function. 63,64,75,83,84 Manual techniques have been examined in two small RCTs of people with osteoporosis and back pain and some positive effects have been reported on thoracic kyphosis,63 pain and QoL. 64 However, each of these trials was underpowered, and manual therapies were combined with other techniques, making it difficult to determine their effectiveness.

Rationale for exercise therapy

Exercise interventions for people with vertebral osteoporosis or osteopenia range from simple individual back extension exercises to a mix of weight-bearing exercises, balance exercises and progressive strength training interventions. 67,70,72,74,76,99,101 Common elements include exercises to improve cervical glide, scapula retraction, shoulder flexion, trunk extension, hip and knee extension and ankle dorsiflexion. Exercises and activities that increase the load on the vulnerable anterior vertebral body (i.e. exercises that involve excessive or dynamic spinal flexion or rotation), high-impact loading and overhead compression activities are not suitable for people at high risk of recurrent OVF. 38

Exercise interventions have been delivered individually,76 in a class format,65,67–69,99,100,102–106 as a HEP70,101 or as a combination of physiotherapist-led exercise and a HEP. 74,75

Exercise has also been found to positively affect bone mass and low BMD, particularly when intervention durations are prolonged. 107–113 Whether or not exercise can reduce the incidence of OVF remains uncertain. 114 Some studies report improvements in OVF risk and in factors associated with OVF risk but the lack of RCTs with appropriate measures and limited long-term follow-up in most studies make it difficult to determine the impact of exercise on vertebral fracture rates. 114,115

Single session of physiotherapy intervention

All trial participants had been prescribed bone-protective medication and all received a single outpatient appointment with a physiotherapist. This started with an assessment that included questions about the person’s current condition, any lifestyle changes that they had made to date to manage their osteoporosis and their educational needs in relation to their osteoporosis. The physiotherapist then provided education and advice following educational materials developed for the trial and considering the individual’s circumstances (e.g. including information about smoking or alcohol consumption and fracture risk). The education did not include any explicit exercise prescription, any manual therapy or any other physiotherapy (e.g. walking aid prescription).

Manual therapy intervention

The manual therapy intervention comprised:

• An initial assessment.

• All SSPT education and advice.

• Low-velocity spinal mobilisations at each treatment session. The start position (prone, prone over pillow or forward lean sitting) was adapted as needed by the treating therapist. Central posterior–anterior accessory mobilisations were applied through the thoracic and lumbar spine at vertebral levels identified by the treating therapist from the assessment. The number of repetitions and grade (grade 2–4) were to be chosen by the treating therapist. Using pain as a guide, the therapist could choose whether or not to mobilise at the level of a known previous vertebral fracture.

• Postural training comprising:

  • Education – providing a rationale for increasing awareness of posture in daily life, discussing the relationship between posture, pain and function and highlighting that flexed posture is likely to have both fixed and modifiable elements.

  • Teaching self-correction of posture to find patients’ usual ‘best posture(s)’ that promote optimal skeletal alignment and comfort in different positions.

  • Postural taping with extension bias over the thoracic region. After specific screening for precautions and contraindications, tape was applied bilaterally from each anterior aspect of the acromioclavicular joint, over the muscle bulk of the upper trapezius and then diagonally towards the spinous process of thoracic vertebral level (T) 6, intersecting over T6 to promote thoracic extension. 91 Tape was applied for up to 3 days at one time, on up to three occasions, and guidelines for wearing and removing the tape were included in the manual therapy diary (see Appendix 1).

• Soft-tissue mobilisation for a maximum of 5 minutes at each treatment session. Treatment was directed at the soft tissues and muscle groups over the thoracic area. The specific area targeted was identified by the treating physiotherapist on assessment (e.g. over the trapezius, supraspinatus, rhomboid or thoracic erector spinae muscle groups). Physiotherapists might elect to use any of the soft-tissue or trigger-point mobilisation techniques or to combine techniques.

• A HEP of three passive stretches each day, one stretch each to promote trunk extension, shoulder flexion and abduction, each for a maximum duration of 5 minutes. The three stretches were selected from the graded set of stretches for each muscle group by the physiotherapist to be at an appropriate level for the individual.

• A manual therapy diary including all stretches, taping records plus strategies to promote adherence, including an agreed personal exercise plan, goal setting, HEP record sheets and participant confidence rating (see Appendices 2 and 3).

• Documentation of assessments and interventions using standard logs (see Appendix 4).

Exercise therapy intervention

The exercise therapy intervention comprised:

• An initial assessment.

• All SSPT education and advice.

• A multimodal, graded exercise programme with three parts – strength training, balance training and weight-bearing exercise. Each element of the programme progressed in difficulty and the overall aim was for participants to achieve 45 minutes of moderate-intensity exercise three to five times per week by the end of 12 weeks. Specifically, it included:

  • Three balance exercises, one each from a walking balance, tandem standing and single-leg standing balance set of exercises.

  • A community walking exercise programme, using a pedometer to determine baseline walking ability, structure progression and encourage self-monitoring with lower-limb stretches as a cool down.

  • Three strength-training exercises, one each from sets to promote cervical dorsal glide (upper-cervical flexion and lower-cervical extension), scapula retraction, shoulder flexion, trunk extension, lumbar stabilisation and functional lower-limb strength.

  • Advice about exercising safely and effectively (e.g. minimising the risk of falls and considering comfort and posture).

  • Specific selection and progression of exercise intensity at a self-perceived moderate to somewhat hard level of effort (rating 3 or 4) using the Borg RPE scale.

• Provision of an exercise therapy diary including all prescribed exercises and strategies to promote adherence to the HEP, including an agreed personal exercise plan with goal setting, exercise practice record sheets, assessment of the participant’s level of confidence with their HEP and telephone support.

• Documentation of assessments and interventions using standard logs (see Appendices 5–7).

Intervention materials

Assessment forms were developed for each trial arm to help guide and standardise the initial assessment across sites and formed part of the clinical record. A pamphlet summarising the educational content was created for treating physiotherapists and a set of laminated information cards, illustrated and written in point form in accessible ‘patient-friendly’ language, was produced (see Appendices 8 and 9). The illustrated cards were for use within the education session, to support the discussion between the therapist and the patient and to ensure that relevant areas were covered in a standard way and that no area was forgotten. The content included background information about bone, osteoporosis and OVF, lifestyle factors that affect bone health, ways to protect the spine in daily life and strategies to reduce the risk of falls. The ROS leaflet ‘Healthy living for strong bones’ was provided to all trial participants to reinforce educational messages (see Appendix 10).

All trial stretches and strength and balance training exercises were photographed, described and categorised into sets; each set included exercises that addressed the same area [e.g. shoulder flexion, graded in difficulty (see Appendix 11)]. These were uploaded to the trial website. The website included an area for physiotherapists carrying out trial interventions [https://research.ndorms.ox.ac.uk/prove/therapists.php (accessed 6 February 2019)] and an online HEP prescription tool. It provided access to all trial exercises and participant documentation. Once the physiotherapist had selected the relevant level of exercise from the exercise set, this was recorded on an exercise record sheet, providing details of frequency, number and intensity. The chosen exercises were included in the participant therapy diary along with other documentation including goal setting and a plan to create an individualised HEP. A brief illustrated summary for physiotherapists for setting and progressing strength and balance training exercises was produced for use along with the Borg RPE scale. These were intended to supplement material in the PROVE trial treating therapist manual, to be easily transportable and useable in the clinical setting as needed.

Documents to support the adherence strategies were produced: for example, a goal-setting record, an exercise programme record and a confidence ruler. Each participant in the intervention arms received a therapy diary in accordance with allocation (see Appendices 2 and 5). These contained all prescribed exercises and a range of documents produced for the trial that were designed to monitor and support adherence to the HEP. Each participant in the exercise arm received a pedometer for use in the walking programme. The pedometer was issued with a wear guide, a daily step count record sheet and guide to increasing step count and walking distance. The physiotherapist provided guidance to the participant on the use of the pedometer and agreed the target increases in walking distance with the participant. Participants retained the pedometer at the end of the intervention.

Intervention providers and setting

The intervention was delivered by UK physiotherapists registered with the Health and Care Professions Council and working in musculoskeletal rehabilitation or specifically with people with osteoporosis. The trial was conducted at 22 hospital sites across the UK in physiotherapy outpatient departments. All therapists were NHS employees who treated trial participants alongside their normal caseload. The treatments delivered as part of the PROVE trial did not require enhanced skill sets but were within the normal scope of physiotherapists working within the NHS. In most centres, therapists delivered each of the trial interventions (i.e. SSPT or manual or exercise therapy programmes) as necessary. In other centres, therapists might split these roles (e.g. one therapist provided SSPT, another conducted manual therapy intervention and another conducted exercise therapy). To enhance consistency, physiotherapists delivering the trial interventions received training from the trial research team prior to seeing trial participants and were required to use trial materials and follow procedures. The integrity of the intervention was monitored during site visits (see Intervention monitoring and support).

Intervention structure and dose

Each intervention was delivered via individual, face-to-face sessions of outpatient physiotherapy. The key reason for selecting individual, face-to-face sessions was that, although exercise and educational interventions can be delivered in class formats or individually, manual techniques require individual ‘hands-on’ treatment. In addition, many older adults prefer individual exercise over group exercise and find their home setting most convenient for exercise (e.g. avoids transport issues). 128 We also thought that individual outpatient physiotherapy was feasible in more settings and, thus, could be more easily implemented on a wider scale.

For those allocated to SSPT, the intervention was delivered in a single appointment, of up to 60 minutes in length. To investigate the different content of each intervention arm, it was decided to keep the underlying programme structure and level of attention in each arm as consistent as possible. As a result, both interventions consisted of a maximum of seven sessions of individual outpatient physiotherapy over 12 weeks. The first session was an assessment lasting up to 60 minutes, with subsequent sessions lasting for 30 minutes.

Individual sessions were supplemented with a HEP. Home-based strength and aerobic exercise programmes in older populations report positive outcomes but maximising adherence is a key consideration. 128 Common strategies to promote adherence were used across the intervention arms (e.g. clear illustrated exercise instructions, goal setting, planning practice individually, systematically reviewing confidence with prescription, recording practice and telephone calls). Previous trials of home exercise that include telephone support from a physiotherapist70,101,128 report higher adherence and more positive outcomes than those without telephone support.

Within the NHS, current physiotherapy provision is limited to a ceiling of an average of around six sessions of individual physiotherapy. Thus, a factor considered when deciding the intervention dose was to determine if a positive treatment effect could be seen at levels of intervention deliverable within current NHS resources. It was felt that markedly increasing the duration of treatment and number of sessions would affect the generalisability of trial findings and increase the NHS excess treatment costs. The consultation with physiotherapy service managers had highlighted that increasing session numbers would be likely to reduce centre recruitment. Thus, the overall package of treatment was a maximum of seven physiotherapist contacts over 12 weeks. Exact session timings were agreed individually between the participant and the physiotherapist but with the expectation that there would be more sessions towards the start of the treatment period (as would happen in routine clinical practice).

Intervention prescription, progression or regression

To prompt an increase in physical capacity, sufficient training stimulus (overload) is needed to elicit musculoskeletal system adaptation and achieve motor learning, then, as an individual’s capacity improves, exercises need to progress to promote further improvement. 124 For each strength, balance or stretch exercise set in the home programmes, the physiotherapist estimated the appropriate start level based on their clinical judgement and initial assessment and then assessed this, asking the participant to perform up to three repetitions of the exercise, judging safety, form, comfort and rating intensity (Borg RPE scale). The exercise selected within the group was adjusted, if needed, based on feedback. Throughout the 12-week intervention, the physiotherapist reviewed and updated the HEP. This was to identify and address any problems and to allow for tailored progression as people adapted to the initial exercise prescription.

For the strength programme, the physiotherapist and the participant together selected an exercise from the designated set(s) that could be performed well and without pain for 10 repetitions (at least 8 and up to a maximum of 12) at a moderate intensity (rating 3–4 Borg RPE scale). 129 The number of sets was progressed up to a maximum of three sets of 10 repetitions, then the exercise level was reviewed (e.g. moving from prone trunk extension with arms to assist to prone trunk extension without arm assistance). The same principle applied to exercise levels that used resistance bands to increase the load. For the step-up and sit-to-stand exercises, the physiotherapist used the number of possible repetitions in 30 seconds at baseline to guide the initial number of repetitions per 30-second set, progressing up to a maximum of four sets.

In the walking programme, participants wore a pedometer for 1 week to allow the physiotherapist to evaluate the individual’s baseline walking level by calculating the average daily step count. Pedometers have been successfully used to structure, set intensity and progress self-monitored walking programmes for older people, including people with vertebral osteoporosis. 130 Participants were asked to increase their average daily step count by 15–30% by the end of 12 weeks in a graded fashion. 122 Progress was monitored mid-way, at 6 weeks, using the pedometer.

The balance exercises were made progressively more difficult in a standard way across the three exercise sets. Participants progressed from upper-limb support to no upper-limb support, from eyes open to eyes closed, from bilateral-leg to single-leg activity and from no distractors to mental distraction. 72,100,104,120

If a physiotherapy treatment, home exercise or taping prescription aggravated a pre-existing condition or resulted in new pain or injury, then physiotherapists were required to follow the adverse events protocol. Aside from this, if a participant reported difficulty with an element of the intervention, the physiotherapists would review the intervention and consider whether to adapt or reduce its intensity. If one element of the exercise programme was not tolerated (e.g. exercises in four-point kneeling), physiotherapists were advised to document this and miss out this element from the programme.

Intervention tailoring

We expected trial participants to have a wide range of pain, mobility and activity levels, various comorbidities, different levels of experience and confidence with exercise and a mix of home environments and support systems. To accommodate these factors and prescribe safely and effectively, it was necessary for the treating physiotherapists to tailor the PROVE trial intervention.

In the manual therapy arm, the therapist tailored spinal mobilisation treatments by selecting start position, spinal segment(s) treated, number of repetitions and mobilisation grade. The physiotherapist used their clinical judgement to select the muscle/soft-tissue area and treatment technique from the range of soft-tissue mobilisation techniques. Therapeutic taping was included only following screening regarding safety and the exact number of taping applications (up to three) was determined by the therapist. The specific stretches selected from the stretch sets also individualised the HEP.

In the exercise arm, the baseline assessment was used with the graded list of exercises and Borg RPE scale to allow the physiotherapist to prescribe a HEP tailored to individual participant capacity. Specifically, this involved manipulating the exercise level selected within each set, the number of repetitions and number of sets, the difficulty level in the balance exercises and walking time. Physiotherapists also worked with participants to set up a personal exercise plan that included a goal relevant to that individual and planned how and when to carry out exercises. The way in which the three exercise elements came together was tailored to fit the person’s daily life. For instance, participants might alternate different programme elements across the week (e.g. in weeks 1 to 4, strength training exercise was completed on a Monday and Thursday and the walking programme and balance training were completed on a Wednesday and Saturday). Over the 12 weeks, as the intensity of the programme progressed, elements were likely to overlap to reach the overall goal of 45 minutes of moderate-intensity exercise on 3–5 days per week by 12 weeks.

Centre characteristics

Twenty-one NHS trusts participated in the trial (Table 3). The recruitment period at each trust varied from 8 to 36 months; this resulted in variation in the number of participants recruited at each site.

Screening

Screening and recruitment took place between September 2013 and September 2016. A total of 6841 patients were screened in rheumatology clinics, DEXA clinics and GP lists (Table 4). Of the screened patients, 1213 (17.7%) were eligible and, of these, 636 (52.4%) were willing to attend a research clinic appointment (Table 5); 616 patients (50.7%) actually attended an appointment. One patient was subsequently found to be ineligible, leaving 615 participants to proceed to the trial. A further withdrawal from the SSPT arm took place following randomisation and prior to assessment and one baseline assessment data set was not obtained from the site, resulting in 613 baseline data sets to be included in the final analysis. The variation in the conversion from screening to recruitment related to the strategy used: some centres screened all clinic attenders, whereas others screened only those participants who had already positively responded to trial invitation information.

A total of 5628 screened patients (82.3%) were not eligible; Table 6 provides the reasons for ineligibility when known. Of those eligible but not willing, reasons were specified in the majority of cases (57.2%), with the remaining 42.8% of patients giving no reason (Table 7).

Recruitment

A total of 615 participants were enrolled in the PROVE trial between September 2013 and September 2016, with 216 being randomly allocated to the exercise therapy arm, 203 to the manual therapy arm and 196 to the SSPT arm. Although the recruitment target of 600 was achieved in August 2016, a further 15 participants were recruited in September 2016 as appointments made in the later stages before the target was reached were all honoured. The final recruitment total was 615 participants over a 36-month period.

Of the 636 participants initially given appointments for baseline assessments, 20 did not attend or cancelled, resulting in 616 participants being randomised (see Figure 2), one of whom was subsequently found to be ineligible, leaving 615 participants to proceed into the trial. A further withdrawal from the SSPT arm took place following randomisation and prior to assessment and one baseline assessment data set was not obtained from the site, resulting in 613 baseline data sets being included in the final analysis.

Demographic and baseline data

Baseline and demographic data summaries for each of the three intervention arms and for all participants are given in Table 8. The three arms were well matched at baseline.

Withdrawals

In total, 63 participants (10.2%) withdrew from the trial during follow-up, with a slightly greater proportion of these in the exercise therapy arm. The timing and reasons for withdrawal are shown in Tables 9 and 10. Most withdrawals took place between the baseline and 4-month assessments, with just 13 taking place between the 4- and 12-month assessments.

In addition, 17 participants (2.7%) were unable to be contacted at the 4-month time point, rising to 23 participants (3.7%) at the 12-month time point.

Participants in the SSPT arm were more likely than those in the other two arms to withdraw because they were not happy with their allocated arm. There were more withdrawals due to ill health in the treatment arms. The withdrawal rate met the prespecified 10% allowed for in the sample size calculation.

Comparison of centres

Figure 5 shows the mean change in QUALEFFO-41 and TLS test scores at 12 months for each centre, combining all intervention arms. Plotted circles have radii proportional to the number of observations in each centre, so that larger circles correspond to centres with more observations. Half of the centres randomised < 15 participants and we have not attempted to show means for individual treatment groups as the numbers would be small. As would be expected, because the SE of the mean difference is proportional to the reciprocal of the square root of the sample size for each centre, results for smaller centres show greater variability. Other than this, there is little indication that there is any effect of centre size on either end point nor that there is correlation between centre effects on change in TLS test and change in QUALEFFO-41.

FIGURE 5.

Mean QUALEFFO-41 and TLS test score change at 12 months for each centre.

Sampling procedure

Purposive sampling was used to achieve a sample that included:

• Female and male participants. Because the majority of research with people with osteoporosis fracture has previously been undertaken with women,131 it was considered important to capture the views of male participants within the study.

• Participants with varying number(s) and location(s) of fracture. It was anticipated that this approach would also include people describing a wide range of levels of physical activity and include middle-aged and older people.

Sample size

We used interpretative phenomenological analysis (IPA), which recommends involving small numbers of participants to gain a rich and in-depth account. 132 We anticipated that approximately five participants from each arm (approximately 15 in total) would provide a rich insight into their experience of the intervention. Eighteen participants were recruited and recruitment halted when a wide range of views had been obtained and no new topic areas were being raised by participants.

Approaching and recruitment of participants

Potential participants to achieve the specified sample were identified by the trial co-ordinator and invitations to participate in the qualitative study were posted out to potential participants. Recruitment took place in two stages: the first stage was early in the trial, to explore the acceptability of the intervention at the outset of the PROVE trial. The PROVE trial was then gradually rolled out across multiple sites. Recruitment to the qualitative research was paused while more sites began recruitment and was then restarted when participants from a wider geographical area and from different treatment centres could be recruited.

Twenty people were approached, of whom 18 were willing to participate in the study. Interviews were held at the homes of participants at a time convenient for them.

Interviews

Participants engaged in in-depth semistructured interviews after they had received their allocated intervention. Written consent was obtained at the start of visits before interviews commenced. An interview guide was developed using data from the PROVE pilot; however, this was not followed rigidly: follow-up questions and probes were used to help the interview flow, ensuring that relevant areas were covered and allowing participants to introduce new areas of relevant discussion. 133 The same experienced physiotherapy researcher (CML) carried out all interviews. Participants were encouraged to discuss any areas they believed to be relevant to their experiences of participating in the PROVE trial. After the interviews, time was spent in everyday conversation with the participants to allow the interviewer to check the well-being of the interviewees. 134 Field notes and memos were recorded in a study diary.

Interviews were digitally audio-recorded and fully transcribed, then all transcripts were checked again against the recordings and transcribing errors were corrected.

Data analyses

Audio-recordings were listened to and transcripts were read until they became familiar. The data were coded in accordance with IPA. 132 Transcript data were broken down into discrete units, making concerted efforts to remain close to the data and continually explore meaning (by CML). The first three transcripts were independently coded by two researchers (CML and KLB). During discussion of these three transcripts, it was agreed that, because similar coding units had been identified by both researchers, it was unnecessary in this instance for all transcripts to be independently coded. All transcripts were coded by one researcher (CML), who then grouped together units found to be conceptually similar into categories. All categories, units and data for each unit were imported into NVivo 11 software (QSR International, Warrington, UK). The units, descriptors of the units and data for each unit were then checked (by Karen L Barker and Dr Francine Toye, who provided independent peer review during analyses) to ensure that the data fully supported the unit and category descriptors. The process of constantly comparing data, codes and categories took place throughout all analyses and the team met regularly to discuss the data. Five randomly chosen transcripts were recoded at least 6 months after first coding. Rigour was therefore promoted by the use of peer review, dependability and confirmability audits, a reflexivity diary and field notes and by using an experienced researcher. 135

Findings

The characteristics of people participating in the qualitative study are presented in Table 20.

A wide geographical and urban/rural spread of interviewees was obtained; interviews took place across central and south-eastern England. Interview visit durations ranged from 1 hour and 15 minutes to 3 hours; the average duration of digital recordings was 61 minutes (range 41–81 minutes).

The findings are presented in three main categories:

1. the experiences, perceptions and views of participants regarding the interventions received in the PROVE trial, their physiotherapists and any further treatment received by participants

2. the experiences, perceptions and views of participants regarding participating in the wider PROVE trial, treatment preferences and identified improvements

3. adherence to interventions and being physically active.

Manual therapy treatment arm

This intervention was well accepted and appreciated by the participants who were interviewed. It was perceived as a very enjoyable and effective treatment by participants, even if participants described a brief period of symptom reduction (this varied from hours to months):

I was fortunate enough to be in the group whereby there was a hands-on massage of my back as well, I could have laid there forever, that was just wonderful.

Interviewee 3

Participants had no concerns about receiving ‘hands-on’ care because they were confident in the expertise and professionalism of their physiotherapist: ‘I knew they were professionals that knew what they were doing’ (interviewee 7) and ‘I know how good physios [physiotherapists] are’ (interviewee 14).

One participant (interviewee 5) was initially apprehensive (‘a bit ooo to start with’) but then found it ‘very very good’.

A range of benefits from manual therapy were described by participants, including improved mobility, reduced stiffness, soothing, energising and the immediate results of ‘hands-on’ treatment, which were appreciated by participants in this arm:

I think I am more mobile and I am holding up better. I think I would have gone over.

Interviewee 2

Very good. It eased the muscles. It eased the joints. Made me more mobile. Got rid of the niggles in the back, etc., and it did improve, sort of, the quality of life, from the back point of view.

Interviewee 14

I think invigorated, you feel quite mmm yes.

Interviewee 2

However, not all participants’ posture and back pain improved: one person said ‘I think it’s the same’ (interviewee 6). Participants also valued the mental health benefits of the intervention:

Also the fact that I’m doing something towards it, I’m not just sitting there thinking ‘oh well, this is my lot’.

Interviewee 2

No side effects of manual therapy were described, although one participant (interviewee 3) thought that not everyone might like ‘hands-on’ massage.

The use of taping was less positively viewed. Several people spoke of the difficulties of keeping the tape in place and removing it at home or disliked it:

[Taping] was good, but it’s difficult on your own to cope with that tape, you know, it started to come off. So I had to get, I had to get my daughter-in-law to help me get it off.

Interviewee 7

[Tape] which was weird, that drove me mad a bit.

Interviewee 2

The spinal exercises were viewed positively and some participants continued to use them, although one participant valued the manual therapy but felt that they needed (and arranged to have) further treatment to improve their core strength and posture using an exercise programme:

[Manual therapy] did loosen up the muscles and it got rid of the knots and it made me more mobile and it got rid of the pains, but what it didn’t do, it didn’t correct any postural trap problems or the core muscles.

Interviewee 14

The amount/dose of exercises was acceptable, although some participants found it hard to do the exercises. Three participants were unable to exercise in the prone position because it was ‘just too uncomfortable’ (interviewee 3) or was like ‘pinched nerves’ (interviewee 2).

Exercise treatment arm

The exercise treatment intervention was accepted and viewed positively by most, but not all, of the interviewed participants in this arm.

Some participants spoke of the benefits of exercise, including reduced pain and stiffness, improved strength and function/mobility: ‘they got me mobile’ (interviewee 15), ‘loosened up’ (interviewee 10) and ‘my core muscles have really strengthened up’ (interviewee 4).

Participants described treatment being individualised and progressed by their physiotherapist. Some participants (interviewees 1 and 4) initially found some exercises painful/too difficult, which led to the exercise being modified by the physiotherapist or halted by the participant:

At first . . . some of the exercises she gave me were very painful. And when I went back to see her she told me to stop doing those and to do different ones, and she said I’ve got to do exercises that don’t hurt but that help, and then gradually I then, as I built up I went back to the exercises that hurt and found that I could do them. So, it was a very gradual thing, I didn’t notice any benefit at all for quite a few weeks but then, after I had been doing it for a while, I really did see the benefit.

Interviewee 4

You’d start at the beginning and you go through it . . . get a more difficult one.

Interviewee 10

Most participants found the regular contact with, and monitoring by, their physiotherapists a helpful element of their treatment [‘I feel very supported when I’m in a programme’ (interviewee 1)] and useful for keeping them motivated and adhering to the programme [because it ‘focused you on what you were doing’ (interviewee 16), was ‘on my conscience more’ (interviewee 13) and is ‘all supporting . . . the skeleton’ (interviewee 10)]. Some found recording their exercises in the diary a helpful element of their treatment: ‘it gives you a structure . . . you tick tick tick’ (interviewee 10), ‘I have benefited from . . . keeping the diaries . . . gets you in the routine and you don’t forget’ (interviewee 4).

Post treatment, people described being active and continuing doing the exercises they found helpful, ‘good’ (interviewees 1 and 10) and/or easy and were ‘just going to carry on doing them now’ (interviewee 4).

I do things that are easy but make a difference.

Interviewee 1

The participants were more likely to jettison exercises that they perceived provided no benefit or were ‘too hard work’ (interviewee 1), too difficult or were in starting positions (such as on the floor) that participants perceived as potentially unsafe (n = 1).

The dose/amount of exercise was accepted, but some participants (interviewees 13 and 17) admitted that they no longer/had not exercised as much as they should (owing to factors such as time, effort and fatigue):

I’m afraid I haven’t done them as thoroughly, am afraid since I stopped going. I do the stretching exercises now and again. I do lots of stretching ones and pushing up ones nearly every day. Sometimes I haven’t pursued very far.

. . . I’m so tired that I go to the [sic] sleep, I might sleep for an hour and a half in the afternoon and then it’s time to start doing things again but I try and get the walking done, that is good.

Interviewee 13

I have to admit that I didn’t pursue the exercise correctly as well as I should have done because I just didn’t have time and by the time the evening came I was too **** tired to start doing exercises.

Interviewee 17

The exercise intervention was experienced as less valuable/helpful, or of no benefit, by some participants who were already exercising regularly when they entered the trial (e.g. they were attending Pilates groups or were gym members):

Frankly . . . I know about exercise, I belong to a gym, I go walking, not as much as I ought to but who does, so it was a bit of a waste of time.

Interviewee 17

This participant ‘got low’ and fed up during treatment and would have preferred a ‘motivating’ group setting. Another person also stated that they would have preferred a group setting with longer treatment sessions and greater supervision of exercise: ’20-min [minute], 30-min sessions sometimes, that’s not really long enough’ (interviewee 15).

Single session of physiotherapy

The majority of people randomised to SSPT who were interviewed did not expect to receive any physiotherapy treatment and did not consider attending SSPT as having received any physiotherapy treatment in the trial:

Oh well I didn’t have physio [physiotherapy] treatment.

Interviewee 8

I thought there’d be nothing.

Interviewee 16

Some were ‘disappointed’ (interviewee 9) or ‘left out’ (interviewee 18), feeling that they had received lesser care than those in the other treatment arms, and one person acted on this by arranging non-trial treatment:

I’m in the control group, as you know, which in theory nothing is happening to, but I . . . thought I must do something, I am not just going to sit around and do nothing so I have been practising the Alexander Technique with a teacher . . .

Interviewee 9

Some participants (interviewees 9 and 11) felt that they had learnt nothing new, because they knew the information provided prior to entering the trial: ‘I knew because I had read stuff’ (interviewee 11). Others described learning new helpful information (interviewees 8 and 18) and continuing to carry out exercises (interviewees 8 and 12).

Benefits of physiotherapy and clinical physiotherapists

Some participants spoke of their perception of physiotherapy in general, for example the mental health benefits from feeling helped [‘mentally you feel that you are being helped, and I think, makes you feel better’ (interviewee 2)] and being listened to and encouraged [‘she listened to what I was saying about my back . . . and she was encouraging (interviewee 13)]. There were no negative views regarding the physiotherapy clinicians providing treatment in the PROVE trial. Physiotherapists were viewed as pleasant, knowledgeable and effective practitioners:

Whoever does it, needs to have the knowledge, understanding and training, which, I think, is probably, basically a physio [physiotherapist].

Interviewee 14

One participant valued the continuity of care and seeing the same physiotherapist each time:

Obviously they knew what they were doing, I mean it’s quite a busy little department there, isn’t it? And because I tended to see the same person.

Interviewee 16

Additional treatment

In addition to the participant allocated to the SSPT arm who was improving after arranging his own private treatment (interviewee 9), several participants had received further physiotherapy following their participation in the PROVE trial. One participant in the manual therapy arm was referred for further NHS physiotherapy after they noticed that their posture was deteriorating, and they were prescribed an exercise programme (interviewee 14). One participant had been involved in a car accident after the PROVE trial and was receiving physiotherapy as part of their recovery (interviewee 16). Another participant (interviewee 8) arranged for private physiotherapy.

Overall perceptions of participating in the PROVE trial

Some participants expressed how positive and good for them participation in the PROVE trial had been (interviewees 7, 10 and 18). One person appreciated how organised the trial was as a whole, with treatment being superior to usual clinical practice, and was glad that they had taken part (interviewee 14).

Although most participants viewed the trial positively, one participant felt that the trial adversely affected his mood:

I found myself and after I finished the PROVE course, I realised I’d got a bit not clinically depressed but a bit low because it kept bringing to mind the fact that I’d got osteoporosis and again, there wasn’t enough explanation of the purpose of the exercises and so on and question . . . I mean maybe it’s just my individual response but there was a whole psychological area that I thought could usefully be thought about.

Interviewee 17

This person described becoming low in mood because participation in the PROVE trial kept bringing his condition to mind, and he did not believe that he was in the right treatment arm or that it helped him. He believed that this was because he received insufficient explanation of his exercises and received one-to-one treatment rather than exercising with a supportive peer group. He compared this treatment with his previous cardiac rehabilitation care and believed that lifestyle advice and peer support should be included in osteoporosis care. He improved after the trial: ‘I got a lot happier because it was over’ (interviewee 17).

Reasons for participation

The majority of participants took part in the PROVE trial because they wanted to help other people [‘if it helps someone else’ (interviewee 2)] and/or wanted to improve knowledge and treatment for women and men with fractures due to osteoporosis; one participant also thought that it would help them understand their condition (interviewee 4):

Well I feel that this osteoporosis thing ought to be explored to its fullest. I think people ought to get much more information and background knowledge of it because it’s getting more and more relevant. Not only to women but I think perhaps men.

Interviewee 3

The impact of research was understood by participants and there was a confident expectation that the research would be useful and used: ‘people that take part in research really do help, it does count’ (interviewee 1). A few participants were grateful for the NHS treatment they had already received and wanted to support the NHS in return: ‘it’s literally wanting to do anything to support the people that have supported me’ (interviewee 1).

One person participated because they appreciated that not many people of his advanced age would be able to take part in the trial; he was also encouraged by his family (interviewee 12). Another participant (interviewee 16) took part because he was made aware that not many men were in the trial and, as a man, he thought that it would help the trial. One participant (interviewee 5) hoped that the trial would include hydrotherapy, which they had previously ‘loved’. One person (interviewee 15) participated in the trial because they were ‘desperate’ to access treatment and another (interviewee 18) joined because their treatment would be provided sooner in the PROVE trial than in usual clinical care.

Trial procedures and processes

Treatment arm preferences and optimal treatment

The majority of participants had a preference for a particular trial arm when they entered the PROVE trial. The most popular preference was for the manual therapy arm (seven participants preferred this): ‘Ooh I rather fancied the massage!’ (interviewee 1). This was either because participants had previously enjoyed massages and found them helpful (interviewee 1) or because participants had not yet experienced ‘hands-on’ physiotherapy and were keen to try it (interviewees 4, 7 and 17): ‘I was so pleased when it [manual therapy] came up. I couldn’t believe it!’ (interviewee 7). One participant believed that there would be an ‘immediate response’ (interviewee 14) to manual therapy whereas exercise would take longer to produce results.

A couple of other participants preferred either manual therapy or exercise therapy (they did not mind which) but did not want to be allocated to receive a SSPT (interviewees 4 and 9). According to the trial database, one participant who was randomised to SSPT did not attend for treatment (interviewee 18). One participant reported a preference for a non-PROVE treatment (hydrotherapy), because they had loved previous hydrotherapy treatment: ‘And I was so hoping I was going to be in there again this time actually’ (interviewee 5). Another participant did not want to be allocated to the exercise therapy arm, preferring manual therapy, because they were already active and knowledgeable about exercise; they were disappointed to not be allocated to manual therapy (interviewee 17). According to the trial database, this participant did not attend for treatment.

Some participants ‘did not really mind’ (interviewee 2) which treatment arm they were allocated to (interviewees 2, 10 and 11) and were happy to go with the flow (interviewee 13). One participant stated that receiving any trial treatment would be beneficial for health and/or mental health (interviewee 10). Several participants had reported experiencing fracture(s) during non-physiotherapy manipulations; one was slightly nervous about the thought of manual therapy but also believed physiotherapy to be different to chiropractic care (interviewee 16).

Optimal treatment

Some participants described what they believed constituted the optimal physiotherapy treatment and this varied widely. Some felt that going to an exercise group for/after their physiotherapy treatment would be enjoyable and beneficial [‘you realise you are not on your own’ (interviewee 17)] and would give them ‘motivation’ (interviewee 15) to continue exercising in the longer term:

In an ideal world it would be nice for somebody to have, to see a physiotherapist and do some exercises and then be reinforced by going into a group with other people.

Interviewee 1

Other participants preferred individualised treatment, valuing the opportunity to discuss their condition one to one: ‘the one on one’s best really, yes because then you can talk it through can’t you?’ (interviewee 7).

Other participants believed that early treatment post onset/diagnosis is important to help inform and treat patients (interviewees 7 and 10). Another participant felt that treatment might need to be sequential [i.e. manual therapy first then followed, at some stage, by an exercise programme and with ‘top-ups’ in the future as/if needed (interviewee 14)]. The amount of resources required, and the issues involved in this, were identified by this participant. Hydrotherapy was very positively viewed by people who had experienced this before participating in the PROVE trial (interviewees 7 and 15). One person was unsure whether or not further treatment would be too much for his back:

The amount of damage that is in my spine at the moment, I don’t know whether physio [physiotherapy] would be a good idea because I don’t know if it would cause more fractures.

Interviewee 11

Perceptions of trial personnel

The trial team were viewed positively by trial participants:

Everybody has been brilliant.

Interviewee 2

I couldn’t fault the people in [trial hospital] that carried out this stuff and all of the staff that I met there. They were brilliant. They were very encouraging and they were good fun.

Interviewee 10

Queries raised by participants were promptly addressed (interviewee 2). Staff were described as nice, kind and encouraging people. One participant believed that she received special attention from staff because she was in the trial:

You know, I felt a bit special because I was the one on the trial and all the other people in the waiting room sort of, he’d come out and see me, he was always very, very pleasant, really nice man.

Interviewee 18

Identified improvements

Participants were asked if there was any way that they thought the trial could have been improved. Participants were generally positive about and ‘very impressed’ (interviewee 3) by the trial and its organisation and many were unable to suggest ways in which the trial could be improved (interviewees 2, 6, 7, 9, 11, 12 and 18). The following suggestions were each made by one or two participants (there were no frequently suggested improvements):

• a further clinical follow-up visit (at 3 or 6 months) to promote adherence to/progress with the exercises (interviewees 1 and 13)

• include hydrotherapy (interviewee 5) as an intervention

• offer technologies such as scanning and e-mailing exercises to participants (interviewee 14)

• ‘a reminder [telephone call] of what I had to be doing, every other month perhaps’ (interviewee 13)

• involving a psychologist in the trial to consider the impact of participation on participants’ mental health [‘I think there is a point there’ (interviewee 17)].

Experience of physical activity and exercise: ‘as long as I’m active I’m much better’ (interviewee 4)

Exercise and/or physical activity was viewed as ‘vital’ (interviewee 12) and ‘necessary’ (interviewee 14) and was undertaken by the participants, both by those who had always enjoyed exercise and been active (interviewees 3, 4, 7, 10 and 17) and by those who had never been naturally athletic/sporty (owing to not enjoying it, perceived laziness, busyness of life or having a desk job) but who understood the importance of being/keeping active (interviewees 1 and 14):

No, the programmes have definitely, definitely helped, because I’m not an athletic person. I was never any good at games, I hated gym at school, I’m not naturally athletic at all, but I do know that exercise, exercise and keeping your weight down, which I find really really difficult.

Interviewee 1

I’ve always been active in sport, which I think, stands me in very good stead now, because if I think if I hadn’t, I think I would be in a very much worse way.

Interviewee 3

One participant (interviewee 15) thought that getting a friend to go to an exercise group with them would help their motivation. Several participants (interviewees 10, 13, 14 and 17) also described how, although they did fewer PROVE exercises than they felt they should do, they were going for walks and keeping active: ‘nothing will stop me walking’ (interviewee 10). Several participants who disliked going to the gym enjoyed walking and keeping active that way (interviewees 14 and 15). For those participants who had always enjoyed exercise/being active, it could be depressing/limiting when certain activities become no longer possible/advisable, although one participant (interviewee 16) simply switched to another type of exercise and did not mind doing this. Exercise was seen as being beneficial for mental health and ‘morale’ (interviewee 15) as well as physical health: ‘mentally you feel that you are being helped, and I think, makes you feel better’ (interviewee 2). Exercise/activity was considered beneficial, especially for people with osteoporosis, for whom the need to keep muscles strong and the spine supported is recognised (interviewee 10). Inactivity is considered harmful, resulting in adverse effects of seizing up, deep-vein thrombosis and ever-deteriorating function (interviewees 10 and 12).

Daily or regular routine

All participants described how they had a regular exercise and/or physical activity routine and had built this into their daily/weekly life: by exercising at the same time of day, attending regular gym sessions/classes, going ‘religiously’ for daily walks/walking every other day [‘for the rest of my days’ (interviewee 10)]:

I have a little routine of exercises that I do every morning.

Interviewee 1

One participant described doing one activity a day even though it exhausted them (interviewee 8). Not all participants always managed a daily/regular pattern despite planning to do so; factors such as fatigue could prevent this or reduce motivation, but they did not cease exercising and returned to it. For a couple of participants, keeping an exercise log helped them to monitor/plan their exercises (interviewees 4 and 14).

Perceived difficulty

The extent to which participants perceived exercises to be difficult or easy influenced adherence:

I do things that are quite easy but make a difference.

Interviewee 1

She gave me these exercises and again they seemed very basic, and I said to her at first I didn’t see any benefit in doing them, but over a period of time, you really do feel the benefit, my core muscles have really strengthened up and I’m just going to carry on doing them now.

Interviewee 4

When advised by their physiotherapist to do an exercise that the participants found painful, difficult, tiring or uncomfortable, participants tried and then stopped doing that exercise: ‘she told me to do some things which I just can’t do so I don’t’ (interviewee 1). Exercises in lying positions, especially in the prone position, were perceived as more difficult by some participants (interviewees 1, 3, 5, 13 and 15):

And she wanted me to lay on my front and do something but there was no way I could do that. I just, wasn’t . . . Oh, so uncomfortable, so uncomfortable.

Interviewee 3

A couple of participants also avoided lying because it could be difficult to get up. For one participant, who lived alone and could not get up from the floor, this exercise was perceived as unsafe. Participants were more likely to do exercises they found easy, although one participant (interviewee 17) did not adhere because they felt that the exercises were not adding anything new [‘the rest of them I was already doing’ (interviewee 17)] and preferred going to the gym and doing Pilates.

Perceived or observed benefits promote adherence

The participants (interviewees 1, 2, 4, 5, 6, 7, 8, 14 and 15) adhered to their exercises because of the perceived benefits and observed results of doing them. Some participants (interviewees 3 and 17) who did not see any improvements attributable to exercise either stopped exercising/the exercise or changed to another form of exercise: ‘I don’t think it makes a great deal of difference’ (interviewee 3).

The factors improving or reducing adherence are summarised in Figure 6.

FIGURE 6.

Summary of the factors that reduced or increased adherence to the PROVE trial.

Impact of symptoms and comorbidity

Adherence was adversely affected for participants experiencing symptoms of osteoporosis and comorbid conditions. Some older participants (interviewees 8 and 12) described how tired they were or how tired their back made them (a younger participant, interviewee 14, descried being tired after work) and how they had slowed down with age:

I do try to take a walk every day and I find it exhausting . . . but my whole back it just hurts, aches all the time. I mean, really deeply aches, yeah.

Interviewee 8

Some participants in the exercise arm (interviewees 13 and 17) admitted that they no longer/had not exercised as much as they should (owing to factors such as time, effort and fatigue):

I’m afraid I haven’t done them as thoroughly, am afraid since I stopped going. I do the stretching exercises now and again. I do lots of stretching ones and pushing up ones nearly every day. Sometimes I haven’t pursued very far.

. . . I’m so tired that I go to the [sic] sleep, I might sleep for an hour and a half in the afternoon and then it’s time to start doing things again but I try and get the walking done, that is good.

Interviewee 13

I have to admit that I didn’t pursue the exercise correctly as well as I should have done because I just didn’t have time and by the time the evening came I was too **** tired to start doing exercises.

Interviewee 17

The factors that influenced adherence are described earlier in this section and are summarised in Figure 6.

Visual proof

The impact of participants seeing visual images of their spines from scans is a strong incentive to adhere to exercise and keeping active: ‘it makes me do my exercises more’ (interviewee 1). Strong terms were used to describe how powerful and upsetting seeing such visual evidence of damage is for people with vertebral fractures due to osteoporosis:

Well, I, when [researcher] measures my spine I am horrified when I see the graph on the paper, I say, I ask her ‘Good ***. Is that my spine?’, and she said ‘Well, yes’, it’s a bit horrifying to come to the realisation that, that is what it looks like.

Interviewee 3

She showed me on screen and explained everything and my T12 I have made one hell of a mess of . . . it does upset me, to think what damage I’ve done to my spine . . . it’s quite horrific.

Interviewee 5

The pictures were on the computer so I saw them and I thought, ‘Blimey, hey’. And it’s quite shocking really when you see it in black and white. And you can see that the bones were just broken and really it’s just the fact your muscles are holding it all together.

Interviewee 16

People relate to these images at a deep level: a picture ‘speaks a thousand words’ (interviewee 16).

Visual changes in posture also prompted adherence:

I can remember all the exercises, such that, when I start realising I’m stooping over again, I don’t have to go back to get more physio [physiotherapy] or anything else.

Interviewee 14

I’ve lost my height as well . . . It is upsetting, you don’t want to get wider and shorter and hunched over.

Interviewee 15

Family history

Some, although not all, participants reported a family history of osteoporosis. The effect of this knowledge on adherence varied. Some participants (interviewee 1) feared becoming like their relative and this spurred them on to fully adhere to all of their osteoporosis treatment:

I was livid but it was also kind of quite inevitable because of my mum and I thought ‘Oh **** here we go, I’m going to be just like her’. But I won’t be. One, because of kind of medical science, because I’ve been able to take the Fosamax and because I’ve been really really aware of my diet. And the exercise. So I’m hoping that I’ll be better off than her, I’m sure I will be.

Interviewee 1

This participant felt that, with advances in treatment, the future would/could be different and she hoped that she would be better off than her relative. Not all inherited traits were negative; one participant (interviewee 1) felt that she had inherited her relative’s determination alongside their osteoporosis. Some participants were realistic about their family history, but one was still surprised when they were diagnosed with osteoporosis. Several participants expressed sorrow/awareness that they had/could pass osteoporosis on to their children; they informed their children and encouraged them to be aware and proactively healthy, active and promptly seek health care/checks if/when needed:

Unfortunately my daughter’s now been, she’s 44 [years old], she’s been diagnosed . . . I was sorry really because I felt that, thinking about it, it’s in our family so probably there’s nothing I could do about it. I’m sad, I’m sad she’s got it.

Interviewee 2

I’ve got two sons and we have kept them fully informed haven’t we . . . we’re very conscious, you know, that these things can be passed down and they’re very much aware aren’t they?

Interviewee 4

Fighting osteoporosis

For some participants (interviewees 2, 4, 6, 7, 11 and 12), adherence to exercise and/or keeping active was a way of fighting the progression of their osteoporosis and not letting osteoporosis win or beat them. Participants spoke of exercise being something they could do [‘I have done everything I can’ (interviewee 4)] instead of giving up and thinking that there is nothing that can be done to prevent/stave off deterioration:

I think the exercise, and also the fact that I’m doing something towards it, I’m not just sitting there thinking ‘oh well, this is my lot’.

Interviewee 2

Adhering to treatment can help give a sense of control over the future progression of osteoporosis [‘I will not let it beat me if I can help it’ (interviewee 5)], although one person believed that they had little control over the risk of future fracture (interviewee 9). People want to know that they have done all they possibly can to help themselves and strictly adhering to treatment is seen as part of this approach. One participant tried to not think about the future but still adhered to treatment (interviewee 18).

Potential consequences of osteoporosis

The potential consequences and/or expectation of worsening osteoporosis motivated adherence to both exercise/activity and medical treatment [i.e. all osteoporosis treatment was seen as, in general, ‘preventing’ (interviewee 8) deterioration as much as possible]. Further events and changes in osteoporosis could be depressing and upsetting:

I must say when I hit the ground for the second time I thought ‘Oh ****. I’ve done it again’. That was immensely depressing because I knew I’d got another 6 months of, you know, not being able to [do] very much and having to start all over again.

Interviewee 1

People do not ‘want to get any worse’ (interviewee 5) than they need to be [‘I am hoping I shan’t go over too much’ (interviewee 2), ‘I was kind of worrying about me ending up in a wheelchair’ (interviewee 10)] and do not see a situation in which they will no longer need to exercise or take care of themselves [‘I think most people with that would recognise that you need to do something to help yourself’ (interviewee 9)]. The potential consequences were taken seriously by participants. One participant felt that previous experience of fracture had prepared them for further fractures. Another participant expected his osteoporosis to worsen but not become ‘the be all and end all of life’ (interviewee 14):

I know there is light at the end of the tunnel, you will, you know, as long as it gives up like it did before, yes I could get through it. I think the first time it happens it’s the unknown, and it is quite difficult.

Interviewee 2

Further fracture or events

Further events/fractures influenced the extent to which participants could adhere to the PROVE trial and continued with exercise/activity post the PROVE trial in the long term. One participant was low in mood after appearing to have sustained further fracture(s) during a car accident; they were in pain, off work and receiving further treatment, ‘It does really [get me down], yeah . . . almost 2 years it was sort of fine then now it’s not and you couldn’t make it up could you?’ (interviewee 16). These participants were keen to resume exercise as soon as symptoms permitted because ‘it’s got to help’ (interviewee 5). Fear of further fracture was a concern for participants (interviewee 1). One participant described a light bulb moment: after his first fractures, he was in denial and did not comply with medical advice at all, but after a subsequent fall and fractures, he became fully adherent.

Complying with health-care professionals

Some participants spoke about adherence to the PROVE trial/NHS treatment for their osteoporosis being affected by a motivation to do as they were told by health-care professionals. One participant felt that if the PROVE trial team bothered to prescribe exercises then she should do everything possible to adhere/comply: ‘if somebody is bothering to put me on this, that I should do everything to try’ (interviewee 2). A couple of people felt that you ‘should listen’ (interviewee 11) to health-care professionals and do as they advise [‘everything he told me I just did. I did exactly what he told me to do’ (interviewee 11)], although one participant recognised the role of the individual and that people could not be made to listen or comply unless they wanted to (interviewee 11). Several participants described partial compliance; doing fewer exercises than prescribed when stable, but returning to or increasing exercises when symptoms such as stiffness returned (interviewee 14 and 15).

Monitoring and support

Participants spoke of the support provided during the intervention and follow-up stages of their treatment (interviewees 1, 2, 15 and 16). Most of the participants who spoke about monitoring/supervision believed that this helped their adherence during the trial and that longer follow-up/further monitoring would motivate them to continue adhering post treatment [‘so it really focused you on what you were doing’ (interviewee 16)] and that people would be more likely to continue exercising:

It’s been wonderful to be back in a programme again, because I feel very supported when I’m in a programme . . ., if I were to know that in 5 years’ time I’d be on another programme, and then 5 years after that, I would really keep going . . . Because if you’re actually being seen in 6 months then you really would keep, would be more likely to keep going.

Interviewee 1

Royal Osteoporosis Society

Although not part of adhering to PROVE trial interventions, some participants (interviewees 2, 4, 9, 14, 16 and 17) spoke about the ROS and adhering/following helpful advice provided by the ROS:

I think the information I have read has helped me a lot to understand, mostly it came from the National Osteoporosis Society [former name of the ROS].

Interviewee 4

The ROS support and information was used by participants and included advice on exercise and activity that might have indirectly supported adherence to the PROVE trial.

Letters from PROVE trial participants

Ten PROVE trial participants, and one physiotherapist, wrote letters to the trial team about PROVE trial treatment. Although not part of the qualitative study, we have included excerpts here because they identify additional factors relating to adherence or provide additional valuable information about factors that were already identified. Two participants described how the recent death of their spouse had understandably affected their participation in the PROVE trial and their questionnaire responses (letters 3 and 10): ‘my mind and body are in turmoil . . . extremely difficult’ (letter 3). Bereavement is an important event to capture in studies, especially those involving older people.

One letter spoke of ‘being prone to spontaneous falls’ (letter 1) and another participant felt prone to falling because the exercises drained them (letter 5). This participant also described a breathing and panic attack (letter 5). Another participant described how a new comorbidity (a heart problem) adversely affected their ability to exercise (letter 4).

Implications for clinical practice

Participants were keen to receive manual therapy and all participants who were interviewed and had received manual therapy were vocally positive about its effects. Participants perceived manual therapy to be safe, enjoyable and effective. The use of taping was generally well tolerated, but two participants found this difficult to remove or unpleasant. Exercises in lying positions, particularly in the prone position, were painful, difficult, avoided or not possible for some participants. PROVE trial treatments were designed to be individualised by physiotherapists, avoiding uncomfortable, difficult or unsafe exercises; therefore, there may have been a training issue or additional training may have been beneficial. Most participants who were allocated to the exercise arm, with one exception, accepted this intervention as an appropriate and effective treatment approach. Participants allocated to the SSPT arm did not perceive this intervention as an approach that was equal to manual or exercise therapy and some participants believed that they had received no physiotherapy treatment. A few participants found travelling to appointments difficult and expensive; the ability to attend for treatment and pay transport costs may influence a patient’s ability to adhere to a treatment package. The trial database indicated that two participants did not attend for treatment and this seemed to be linked to not receiving their treatment of choice.

Issues identified and recommendations

The findings of this study, as in previous studies, have identified and highlighted issues relevant to the design and implementation of both the PROVE trial and trials more generally.

Qualitative studies allow the findings of trials to be explored and explained, the development of specific recommendations about improving trial design(s) and the generation of questions that contextualise trial findings and help to increase the overall utility of the evidence generated by the trial. 136,137 Table 21 summarises the issues arising from this research and provides recommendations for consideration of further research or to address these issues in future research.

Strengths and limitations of the study

The sampling strategy successfully recruited a wide range of participants: both men and women with varying numbers of fractures and in a variety of locations. This has provided us with wide-ranging and rich data about participants’ experiences of the PROVE trial, its processes and interventions and adherence to exercise and advice. However, as we interviewed only 3% of all the participants in the PROVE trial, the data cannot be representative of the whole sample but rather provide a snapshot view.

It is difficult to estimate an appropriate sample size for qualitative studies when projects require the estimation of numbers of participants and resources in advance of funding being granted. 138 Previous experience in the team (of carrying out similar studies) suggested that 15 participants would provide an appropriate sample size given the detailed level of analyses involved in IPA. During the study, 18 participants were recruited before the team was satisfied that the range of experiences described by participants was sufficient to ensure that the aim of the study had been achieved. The need to adequately resource the qualitative research in combined qualitative and quantitative research has been previously raised. 136

A further strength of this research is that, rather than being peripheral or an ‘add-on’ to the study, both the qualitative and quantitative research were considered essential elements that provided a better opportunity to maximise the value of the qualitative research. 139 This chapter has detailed the qualitative research with participants post intervention. It builds on the pre-trial collection of qualitative and survey data from the development and piloting of the intervention from people with osteoporosis, clinicians and researchers and for this reason, the qualitative and quantitative strands of the research have intertwined throughout the PROVE trial.

Key points

• Exercise therapy resulted in more QALYs but higher costs than SSPT, whereas manual therapy was more costly and resulted in fewer QALYs than SSPT.

• The two physiotherapy interventions were not cost-effective relative to SSPT using the £20,000-per-QALY threshold.

• Despite the substantial number of missing data, the various sensitivity analyses conducted showed that SSPT remained the most cost-effective option.

• Overall, exercise therapy was more effective but more costly than usual care, whereas manual therapy was less effective and more costly than SSPT.

Introduction

Osteoporotic vertebral fractures have a significant economic burden in terms of health-care and non-health-care costs140 and a significant impact on patient QoL, ranging from pain to impaired mobility. Current non-invasive management of OVF in the UK involves starting patients on appropriate medication, education about osteoporosis and providing general advice about exercise. There is some evidence that physiotherapy interventions can be beneficial for people with at least one previous OVF and back pain. However, UK-based high-quality evidence on which physiotherapy interventions are more effective and cost-effective in this population is unclear.

This chapter summarises the methods and results of an economic evaluation that was based on evidence from the PROVE trial. Two physiotherapy approaches (exercise therapy and manual therapy) were compared with a single 1-hour session of physiotherapy (SSPT). The cost-effectiveness analysis reports the incremental costs per QALY gained from the alternative options. The target population comprised individuals with at least one previous OVF and back pain identified in a NHS hospital setting.

Existing research

We conducted a literature review of cost-effectiveness studies that were published between 2000 and January 2018 concerning interventions for individuals with vertebral fracture. The MEDLINE and EMBASE electronic databases were searched; the search terms used and results are reported in Appendix 23. We identified eight studies estimating the cost-effectiveness of various treatment approaches in this population but no studies examining physiotherapy interventions. Three studies looked at kyphoplasty compared with standard medical treatment. 140–142 Five studies37,143,144 compared vertebroplasty with standard medical treatment. Of the eight studies identified, two were based on randomised clinical trials,140,143 one was based on a retrospective observational study144 and five used decision models (four used Markov models37,141,142 and one used a treatment state model37).

The three papers comparing kyphoplasty with standard medical treatment studies reported disparate results: one study concluded that kyphoplasty was not cost-effective and the other two studies concluded that it was. The study by Fritzell et al. ,140 a RCT, reported an incremental cost per QALY of Swedish krona (SEK) 884,684, well above what Sweden is willing to pay for a QALY gained (SEK 600,000). In contrast, the two British studies (Svedbom et al. 142 and Ström et al. 141) reported balloon kyphoplasty to be cost-effective compared with standard treatment, with incremental costs per QALY gained of £2706 and £8800, respectively. Fritzell et al. 140 use a Swedish cohort from the Fracture Reduction Evaluation (FREE) trial, Ström et al. 141 use all FREE trial participants to inform the analysis and Svedbom et al. 142 use data from the FREE trial as well as the VERTOS II (Vertebroplasty vs. conservative treatment in acute osteoporotic vertebral compression fractures) trial.

The reasons behind the difference in results are likely to be multifactorial. One possible explanation is that Fritzell et al. 140 used a wider perspective in costs, including more categories of costs (e.g. informal care), which was not the case in the other two studies. Fritzell et al. 140 report the cost-effectiveness analysis over a 2-year period, whereas the British studies used Markov models to estimate the cost-effectiveness of the interventions using a lifetime horizon. Hence, Fritzell et al. 140 report the QALYs gained to be 0.085 for kyphoplasty, whereas Ström et al. 141 and Svedbom et al. 142 reported 0.17 and 0.50 QALYs for kyphoplasty relative to non-surgical management, respectively. The differences in the data inputs, methods and time horizons between the three studies mean that a direct comparison of their findings may not be fully appropriate.

Four out of five studies comparing vertebroplasty with standard medical treatment concluded that it was the most cost-effective option. Masala et al. 144 reported costs as mean costs per point increase in ADL, ambulation and visual analogue scale (VAS) pain, whereas the remaining studies reported incremental costs per QALY gained. Klazen et al. 143 conducted an open-label randomised trial (VERTOS II) comparing vertebroplasty with standard treatment, and found the incremental cost per QALY gained to be cost-effective in the Netherlands. Stevenson et al. 37 reported the results of UK-based cost-effectiveness analyses submitted to NICE by the authors (UK assessment group) and two different manufacturers [Johnson & Johnson (New Brunswick, NJ, USA) and Medtronic (Dublin, Ireland)]. The UK assessment group reported results that were similar to those in the Medtronic and Johnson & Johnson analyses [i.e. vertebroplasty was cost-effective relative to non-invasive treatment; incremental cost-effectiveness ratios (ICERs) were < £5000 per QALY gained].

Health economic data collection

All participants were followed up to 12 months. The economic data were collected via two self-reported participant diaries covering the time period from randomisation to 4 months and from 4 months to 12 months. The self-reported economic diaries asked the participants to record primary care contacts (with a GP or nurse, for surgery or home visits), physiotherapy contacts, newly prescribed medication and secondary care contacts (outpatient and inpatient episodes and lengths of stay). The diaries also captured equipment that the participants started using (e.g. walking sticks), over-the-counter (OTC) medication, private health-care visits as well as social care, informal care and impact on work (e.g. sick days). Furthermore, the EQ-5D-5L questionnaire was administered to participants at baseline, 4 months, 6 months (via post), 9 months (via post) and 12 months. This QoL instrument assesses health utilities from five domains: mobility, self-care, usual activity, pain/discomfort and anxiety/depression.

Quality of life

We estimated the QoL in the physiotherapy intervention arms and the SSPT arm using the EQ-5D-5L data collected within the trial. Following NICE recommendations,146,147 to derive utility weights from the five-level version, we applied a UK crosswalk from the EuroQoL group to convert the five-level responses to three-level responses. 148 We then used the EuroQol-5 Dimensions, three-level version (EQ-5D-3L), time trade-off (TTO) values set developed by Dolan149 on a UK sample to map utility values to the three-level responses.

We assumed changes in mean utility values between each time point to be straight-line transitions, and weighted the average change in utility between each time point for each patient by the within-trial survival time to estimate QALYs gained during the study period, equivalent to an area under the curve calculation. We truncated the analysis between the randomisation date and 12 months of follow-up. Differences in the mean numbers of QALYs between the three arms during the 12 months of the trial were adjusted for baseline differences in utilities.

Intervention costs

The interventions delivered did not require enhanced skill sets but were within the normal scope of musculoskeletal physiotherapists working in the NHS. Each trial centre recorded the number of physiotherapy sessions attended by the trial participants. The costs of exercise and manual therapy interventions were estimated by valuing the recorded physiotherapy sessions (£49) attended by the participants using NHS Reference Costs 2015–16. 150 Manual therapy was identified to be £1.50 more expensive per physiotherapy session owing to the additional consumable materials used. Each physiotherapy session amounted to 30 minutes for both interventions, with the exception of the initial assessment session, which lasted for 1 hour and was costed at £98. All participants in the three arms of the trial were offered the initial 1-hour assessment session, which involved clinical assessment of the participant by the therapist and the provision of advice. We also complemented the number of sessions recorded by each trial centre with data from the participant diaries. We assumed that the total number of therapy sessions attended in the first 4 months was the higher of the two independent sources (participant diaries covering the period up to 4 months post randomisation and trial centre data). For example, if a centre reported that a participant attended three physiotherapy sessions but the participant reported five sessions in their diary up to 4 months, we assumed that the total number of sessions in that time period was five. Similarly, if a centre reported five physiotherapy sessions attended but the participant recorded attending only three sessions, we assumed that a total of five sessions were attended. We then added any physiotherapy sessions reported in the participant diaries covering the period between 4 months and 12 months of follow-up. The cost of the intervention per participant was calculated by counting the total number of therapy sessions undertaken by all participants in each arm divided by the number of participants.

Primary and secondary costs

Costs per health-care contact in a primary care setting were calculated using unit costs from a national database,151 including qualification costs and direct care costs when applicable. Hospital inpatient and outpatient contacts were valued using NHS Reference Costs 2015–16. 150 Each hospital admission was assigned into a 2015/16 Healthcare Resource Group using standard software (HRG4+ 2015/16 Reference Costs Grouper; NHS Digital, Leeds, UK) based on what was recorded as the reason for admission (converted into International Classification of Diseases, tenth edition, code) and/or operational procedures [converted into OPCS Classification of Interventions and Procedures (OPCS-4)]. HRGs consist of standard groups of clinically similar treatments that consume a common set of health-care resources. All unit costs are reported in Appendix 23.

Medication costs

Any new medications, both OTC medication and prescribed medication, were recorded in the self-reported participant diaries. This medication use was valued using weighted-average medication unit costs per item prescribed [net ingredient cost (NIC)] from national prescribing volumes and costs in 2016. 152 Each individual medication reported in the participant diaries was reviewed and based on the diary-reported length of treatment, type of medication, indication for the medication and referencing with the British National Formulary; the total number of packs of medication expected over the 12-month period was determined and costed appropriately. For recurring medications, it was assumed that each NIC per item prescribed was incurred once a month for daily use medications. Intravenous medication, injections and medications with frequencies other than once a day were looked at individually to determine the expected number of items used over the period of time when the medication was taken. The cost per NIC was multiplied by the number of prescriptions expected over 12 months and a sum of the costs in each intervention arm was calculated and divided by the number of participants.

Out-of-pocket expenses

Any equipment purchased by participants was recorded in the self-reported participant diaries as well as use of private health-care services. Equipment was valued using prices listed in large UK retailers and a national database. 151 Use of private health-care services was valued using NHS Reference Costs 2015–2016150 and the NHS Choices website. 153 All unit costs are reported in Appendix 24.

Social care costs

The use of social care services (e.g. home care and home nurse) reported in the participant diaries was valued using national databases150,151 by multiplying the unit costs (see Appendix 24) by the number of hours per week and the number of weeks during which Personal Social Services were used.

Informal care and lost earnings attributable to morbidity

Unpaid care provided by friends and relatives of participants was recorded in the participant diaries and valued using the 2016 median weekly gross pay if the carer was employed (£539) or the 2016 minimum hourly wage (£7.20) if the carer was unemployed. These unit costs were multiplied by the number of hours and number of weeks of informal care received (see Appendix 24).

The participants also recorded the number of days taken off work during the trial follow-up in the self-reported diaries. The occupation of each participant was coded using the Office for National Statistics Standard Occupational Classification coding tool154 and matched against the national 2016 full-time and part-time weekly gross pay. 155 The weekly gross pay was converted into daily pay and multiplied by the number of days taken off work to estimate the lost earnings for each participant.

Discounting

As the time horizon of the analysis was 1 year, costs and QALYs were not discounted.

Statistical analysis

We reported descriptive statistics (means, SDs and SEs) for resource use based on the participant diaries at 4 and 12 months. We reported resource use and costs within 12 months of randomisation per diary completed (4 or 12 months) and for participants who had completed both diaries.

We reported descriptive statistics (means, SDs and SEs) for EQ-5D-5L utilities at baseline and 4, 6, 9 and 12 months. We compared EQ-5D-5L utilities between the intervention arms and SSPT using multilevel linear mixed models with no imputation of missing data. We used the scheduled times of EQ-5D-5L questionnaires for this analysis (baseline and 4, 6, 9 and 12 months) and not the dates when the questionnaires were completed (which may have varied from the scheduled times in some cases). The models were run in Stata^® version 14.2 (StataCorp LP, College Station, TX, USA) using the mixed command with robust SEs and assuming normal error distributions. These models included participant-level random effects. Fixed effects included baseline EQ-5D-5L weights, treatment assignment and time since baseline. Additional fixed effects representing an interaction between treatment and time since baseline, age at randomisation and sex were also conducted.

In addition to the analyses comparing health utilities described previously, we also estimated QALYs by estimating the area under the curve resulting from the linear interpolation of health utilities between the time points: baseline and 4, 6, 9 and 12 months. We used the dates of completion of the EQ-5D-5L questionnaires to inform the estimation of the area under the curve. We also truncated the analysis between the randomisation date and 12 months of follow-up. Two complete-case analyses (CCAs) were conducted: (1) CCA1, in which all five EQ-5D-5L questionnaires had to be complete, and (2) CCA2, in which only the baseline, 4-month and 12-month EQ-5D-5L questionnaires had to be complete for each participant. Differences in mean QALYs between the three arms during the 12 months of the trial were adjusted for baseline differences in utilities.

We then conducted a within-trial economic analysis, with total health-care costs and QALYs gained per participant calculated for the 12 months of the trial in the exercise and manual therapy arms and SSPT arm. Our base-case analysis used intent-to-treat principles, wherein medical resource use, costs and EQ-5D-based health utilities were analysed in accordance with randomised treatment assignment.

We calculated the ICER by dividing the mean cost difference between the intervention and SSPT arms by the mean QALY difference. Three methods of analysis were used to estimate the ICERs: (1) CCA1, as defined above, and both health economics diaries completed; (2) CCA2, as defined above, and both health economics diaries completed, and (3) MI (see below). The difference in costs and QALYs was estimated with regression analysis, using a system of seemingly unrelated regressions. The mean difference in QALYs was adjusted for baseline imbalances in the utility scores. We report the probability that the intervention is cost-effective at a threshold of £20,000 per QALY gained. The probability of the intervention being cost-effective at the threshold of £20,000 per QALY was determined using the net benefit framework and Fieller’s theorem. 156

To deal with missing QoL and cost data, MIs were conducted using a chained model with 70 iterations, regressed on the baseline complete covariates age at randomisation, sex, recruitment centre, number of falls in previous year, walking distance, number of spine and non-spinal fractures and EQ-5D-5L utility score. Mean imputation was used for missing baseline EQ-5D utility scores. Imputed variables consisted of EQ-5D-5L utility scores at 4 and 12 months and total NHS and social care costs up to 4 months and between 4 and 12 months. The patterns of missing data were evaluated to assess whether or not the missing data were non-monotonic and we examined the association between data missingness and baseline variables and observed EQ-5D-5L scores and costs. 157 Given the non-normality of EQ-5D-5L and cost data, we used predictive mean matching, using the five nearest neighbours, to ensure that only plausible values of the missing variable were imputed.

In sensitivity analysis, we explored the following: (1) using NHS costs only, (2) adjusting the difference in costs and QALYs for baseline imbalances in the utility scores and (3) using different MI models (varying the number of baseline covariates). We also conducted subgroup analysis by estimating the incremental costs, QALYs and ICERs by sex and for participants aged older and younger than 70 years.

Participant sample and health economic data completion

At the end of the 12 months, 90% of individuals were still in the trial (see Appendix 23). The proportion of participants with all EQ-5D-5L questionnaires completed was highest in the SSPT arm (61.73%), followed by the exercise therapy arm (52.78%) and the manual therapy arm (51.23%). The proportion of participants with EQ-5D-5L questionnaires completed at least at baseline, 4 months and 12 months was highest in the SSPT arm (77.55%), followed by the manual therapy arm (74.88%) and the exercise therapy arm (68.06%) (see Appendix 24). In terms of resource use and cost data, 69.39% of participants in the SSPT arm returned both participant diaries, of which 92.65% were returned completed (see Appendix 24); participants in the manual therapy arm had a lower return rate (65.02%) and completion rate (89.39%). Of the three trial arms, participants in the exercise therapy arm reported the lowest return (55.09%) and completion (88.24%) rates for the participant diaries. Combining the health economic data reduced the proportion of participants who were eligible for the health economic analysis to 44% in the exercise therapy arm, 52% in the manual therapy arm and 57% in the SSPT arm. Table 22 reports the baseline demographics and characteristics for participants in the trial and for those with complete and incomplete health economic data. We defined complete data as having EQ-5D-5L data at least at baseline, 4 months and 12 months together with completed self-reported diaries 1 and 2 with NHS and social care resource use. Participant characteristics were found to be similar across participants with complete and incomplete health economic data. In the exercise therapy arm, participants with complete health economic data had a lower body mass index (BMI) (mean difference: –1.82 kg/m², p = 0.007) and lower rates of back pain (odds ratio 0.48, p = 0.018) than those without complete data. No other significant differences were identified between those with complete and incomplete health economic data.

Utilities

Appendix 23 reports the completion rate of EQ-5D-5L questionnaires at baseline and 4, 6, 9 and 12 months. We also report the timing of the completion of the questionnaires relative to the expected time points. Appendix 23 reports the dimensions and levels of the EQ-5D-5L questionnaires by intervention arm of the trial and time point.

In the mobility and self-care dimensions, there were small changes between baseline and 12 months for all three intervention arms. The pain and discomfort dimension captured improvements across the three interventions. The largest change in the pain and discomfort dimension was seen in the SSPT arm, with a 15.04% increase in the no/slight levels, compared with an 11.16% increase in the exercise arm and a 6.55% increase in the manual therapy arm (see Appendix 24). There was also an improvement in the usual activity dimension for all three intervention arms. The largest change in the usual activity dimension was also seen in the SSPT arm, with an 8.88% increase in no/slight levels, compared with small changes in the no/slight levels in the manual and exercise therapy arms (see Appendix 24). In the anxiety and depression dimension, there was a decrease of 5.50% in the no/slight anxiety levels in the exercise therapy arm (more anxiety and depression reported at 12 months) in contrast to an increase of 4.74% in the SSPT arm (less anxiety and depression reported at 12 months). However, these changes were not statistically significant (see Appendix 24).

Table 23 and Figure 7 report the mean utilities at each time point by intervention arm and their change relative to the baseline value (see also Appendix 24). There was an imbalance at baseline in the reported utilities, with participants in the SSPT arm reporting the highest average utility, followed by those in the manual therapy arm and the exercise therapy arm, albeit these differences were not statistically significant (p = 0.167 and p = 0.159, respectively). The mean changes at each time period and across the three arms were not statistically significant (95% CIs contain 0). Furthermore, the mean changes in EQ-5D-5L index scores in the exercise therapy arm were, overall, positive relative to baseline, with the magnitude of change at 12 months (0.017) being similar to that reported in the SSPT arm (0.016). These changes were more marked at 4 months (0.040 in the exercise arm and 0.021 in the SSPT arm). In the manual therapy arm, the EQ-5D-5L index score was worse at all time points relative to baseline.

FIGURE 7.

The EQ-5D-5L changes from baseline (95% CI): ITT analysis, observed data.

The SSPT and exercise therapy arms reported the highest mean utilities at 4 months, with SSPT reporting a lower mean utility at 6 months (below the baseline value) and increasing again at 12 months (above the baseline value). In contrast, the average utility in the exercise therapy arm was consistently above the mean baseline value at all time points. The variation in utility values in the SSPT arm may be explained by the changes in the pain and discomfort and anxiety and depression dimensions. In the SSPT arm, a greater percentage of participants moved to the more severe levels (moderate/severe/extreme pain and discomfort) between months 4 and 6 relative to the exercise therapy arm. However, in subsequent months, these increases levelled out and moved closer to the baseline values. The manual therapy arm appeared to follow a relatively stable downward trajectory throughout, which was consistently below the mean baseline value.

Table 24 reports the results from the multilevel mixed-effects linear models comparing the utilities and adjusting for several covariates, such as baseline EQ-5D-5L score and number of months since randomisation. There was no significant difference in utilities across the three intervention arms.

Following MI of the utility scores at 4 and 12 months (Table 25), results similar to those in the CCA were found in terms of direction and magnitude in the SSPT and manual therapy arms. In the exercise therapy arm, the change in utility scores at 4 months became significant (0.031, 95% CI 0.010 to 0.052). The utility score at 12 months was considerably lower than what was observed in the CCA and, conditional on the imputation model used, it could be lower than the mean baseline utility, albeit never statistically significant.

Table 26 reports the QALYs for participants with the five EQ-5D-5L questionnaires completed, three questionnaires completed (baseline, 4 months and 12 months) and MI. We found no significant difference in QALYs across the three arms of the trial after adjusting for baseline EQ-5D-5L imbalances in any of the analyses. However, the numbers of QALYs were higher in the exercise therapy arm in the three analyses, followed by the SSPT arm and the manual therapy arm.

Resource use and costs

Participants in the manual therapy arm reported the highest number of physiotherapy sessions attended in the first 4 months [mean 5.03 sessions (SD 2.57 sessions)], followed by participants in the exercise therapy arm [mean 4.33 sessions (SD 2.77 sessions)] and the SSPT arm [mean 1.69 sessions (SD 0.59 sessions)] (see Appendix 24).

Table 27 reports the resource utilisation in the three intervention arms during the 12 months of follow-up for participants with complete economic data (n = 349). The most physiotherapy sessions were reported in the manual therapy arm (mean 5.93 sessions), followed by the exercise therapy arm (mean 5.03 sessions) and the SSPT arm (1.69 sessions). The SSPT arm reported the highest number of contacts with a GP or nurse or outpatient visits (7.12 contacts/visits) and this health-care category was the most used across all three intervention arms.

Converting the resource utilisation into costs, the highest cost category reported was inpatient hospital stays, varying from a mean of £499 (SD £1361) per participant in the SSPT arm to a mean of £767 (SD £2932) in the exercise therapy arm (Table 28). This was followed by GP/nurse/outpatient costs, which varied from a mean of £402 (SD £458) per participant in the exercise therapy arm to a mean of £435 (SD £422) in the SSPT arm. The total NHS costs were found to be higher in the manual therapy arm (mean of £1638 per participant), followed by the exercise therapy arm (mean £1627) and SSPT (mean £1410). The difference in NHS costs was not statistically significant between exercise therapy (mean £217, 95% CI –£446 to £880) or manual therapy (mean difference £228, 95% CI –£340 to £796) and the SSPT arm.

Combining NHS and social care costs, the exercise therapy arm had the highest mean costs per participant (£1733), followed by manual therapy (£1676) and SSPT (£1430). Although the difference in mean costs between exercise therapy (£304, 95% CI –£388 to £995), manual therapy (£246, 95% CI –£328 to £819) and SSPT increased, relative to using NHS costs only, they remained clinically not significant. Adopting a broader perspective and including participant expenses (e.g. equipment purchased and private care) resulted in the mean costs per participant in the exercise therapy arm and manual therapy arm being £277 (95% CI –£423 to £976) and £211 (95% CI –£368 to £789), respectively, higher than in the SSPT arm. Table 29 reports the costs per individual receiving social care, informal care or reporting lost earnings in the three trial arms.

Appendix 24 reports resource utilisation and costs by type of health-care category and intervention for participants who completed diary 1 (first 4 months) or diary 2 (4 months to 12 months). Combining the available data from diaries 1 and 2 resulted in lower NHS and social care costs per patient than using only participants with both diaries completed. For example, the total NHS and social care cost at 12 months in the exercise therapy arm was £1472 when adding the available data in the two time periods, compared with £1733 obtained using complete cases only.

Multiple imputation

The pattern of missing data across the three interventions was visually assessed and found to be non-monotonic (e.g. participants completed EQ-5D-5L questionnaires at 12 months but not at 4 months). Lower EQ-5D-5L scores at baseline and allocation to the exercise therapy arm were significantly associated with missing costs, adjusting for sex and age at randomisation. Lower EQ-5D-5L scores at baseline were found to be significantly associated with missing EQ-5D-5L data at 12 months, adjusting for treatment allocation, sex and age at randomisation. However, observed values in the preceding time period were significantly associated with missing EQ-5D-5L data, adjusting for all baseline variables. As a result, the data were assumed to be missing at random. Eight MI models were defined, with the most complex model used as the base case for MI. We checked the fit of the imputation model by comparing the imputed values with the observed ones and found the distributions to be similar for costs and utilities.

Table 30 reports the EQ-5D-5L data and costs in each intervention arm in two CCAs (CCA1, in which all five EQ-5D-5L questionnaires had to be complete, and CCA2, in which only the baseline, 4-month and 12-month EQ-5D-5L questionnaires had to be complete) and the MI data. Overall, participants with incomplete data reported lower mean utility scores than those included in the CCA. Across the three intervention arms, when comparing observed with imputed data in participants with incomplete data, MI resulted in similar EQ-5D-5L scores at baseline, higher EQ-5D-5L scores at 4 months and lower EQ-5D scores at 12 months. There were also considerable differences between CCA1 and CCA2 in terms of costs and EQ-5D-5L scores.

Cost-effectiveness analysis

Table 31 reports the cost-effectiveness results. Overall, exercise therapy was more effective but more costly than SSPT, whereas manual therapy was less effective and more costly than SSPT. However, the mean incremental costs and QALYs varied across CCA1 and CCA2 and MI. For example, relative to SSPT, the incremental costs of exercise therapy were £536 (95% CI –£183 to £1255) for CCA1, £347 (95% CI –£376 to £1071) for CCA2 and £206 (95% CI –£228 to £641) for MI. The difference in QALYs, adjusted for baseline EQ-5D score, was 0.017 (95% CI –0.013 to 0.048) for CCA1, 0.014 (95% CI –0.013 to 0.041) for CCA2 and 0.002 (95% CI –0.020 to 0.025) for MI. However, using a threshold of £20,000 per QALY gained, the physiotherapy interventions were found to not be cost-effective relative to SSPT across the three methods of analysis. Figure 8 shows the cost-effectiveness acceptability curves for exercise therapy relative to SSPT with the three methods of analysis. CCA2 was the only analysis in which exercise therapy had a probability of > 50% of being the most cost-effective option under a threshold of £30,000 per QALY gained.

FIGURE 8.

Cost-effectiveness acceptability curve for exercise therapy relative to SSPT, using three methods of analysis: CCA1, CCA2 and MI.

Sensitivity analysis

Table 32 reports the incremental costs, QALYs and ICERs for difference sensitivity analysis scenarios explored using the MI method. Although there was considerable variation in the ICERs across the different MI models used (from £58,828/QALY to £1,490,543/QALY for exercise therapy relative to SSPT), the two physiotherapy interventions were consistently found to not be cost-effective. In all analyses, the manual therapy intervention was dominated by SSPT (i.e. it was more costly and less effective) with a probability of being the most cost-effective option of 5–7% at the £20,000-per-QALY threshold. This is in keeping with our findings in the base-case analysis.

Subgroup analysis

Table 33 reports the cost-effectiveness results by participant subgroup using the MI method. There were no statistically significant differences in incremental costs or incremental QALYs across the three arms of the trial. Exercise therapy was more effective than SSPT for females but not for males, albeit these differences were not statistically significant. SSPT was the most cost-effective option only for females and for participants aged > 70 years. Manual therapy was cost-effective only for males, but cheaper and less effective than SSPT. For participants aged ≤ 70 years, exercise therapy was suggested to be the most cost-effective option at the £20,000-per-QALY threshold.

QUALEFFO-41

We chose to use the QUALEFFO-41 as our primary outcome, as it is the recognised disease-specific QoL measure for this population and is targeted towards fracture assessment and deterioration. 73,171 We also chose to use the total QUALEFFO-41 score (sum of all 41 items) as our primary outcome measure rather than specifying one of the five constituent domains: pain, physical function, social function, general health perception and mental performance. There is no published MCID for the QUALEFFO-41, so interpreting the significance of the size of any change would require further research.

At 12 months, a small improvement in the mean QUALEFFO-41 total score was seen in the exercise therapy and SSPT arms and a minimal improvement was seen in the manual therapy arm. Differences between arms were not significant and outcomes in the manual therapy arm compared with SSPT worsened. The QUALEFFO-41 pain subscale score changes at 12 months showed an improvement at a clinically important level: the exercise therapy arm had the largest improvement, with a mean score of –9.07 points (SD 22.88 points) points, and improvements in the manual therapy and SSPT arms were of a similar magnitude, with mean scores of –6.75 and –6.25 points, respectively. Liu-Ambrose et al. 116 conducted a three-arm trial in women with osteoporosis and compared active interventions of resistance training or agility training with a general stretching intervention. As in our trial, with three interventions, they found no statistically significant changes in total QUALEFFO-41 scores between arms but relevant change within the exercise arm for the pain (of a similar magnitude to the PROVE trial) and social activity subscales (in the resistance arm) and for the physical function subscale in the agility arm. They concluded that resistance or agility training provided important QoL benefits to women with osteoporosis.

Several physiotherapy and exercise RCTs that have used the QUALEFFO-41 total score as their outcome measure of QoL have failed to show an improvement. 63,64,69,116,172 Some studies have failed to demonstrate improvements in total QUALEFFO-41 score but show change in QUALEFFO-41 subscale scores. 65,69 Arguably, using subscales might be more sensitive than the total QUALEFFO-41 to detect changes expected from physical interventions. In other trials in which no change has been observed in QUALEFFO-41 score, significant changes have been detected in other QoL measures: Short Form questionnaire-36 items (SF-36) subscales and General Health Questionnaire – 20 items total score and clinical measures that assess similar domains (e.g. pain and physical function at the same time points). 64,65,69 Other disease-specific QoL instruments have been used to measure the effects of physiotherapy interventions. Madureira et al. ,73 Hongo et al. 70 and Chien et al. 101 found improvements in QoL using the Osteoporosis Assessment Questionnaire and its Japanese derivation (Japanese Osteoporosis Quality of Life Questionnaire). These findings need to be treated with caution; Chien et al. 101 did not conduct an ITT analysis, and assessors were not blinded to arm allocation or outcome assessment in the study by Hongo et al. 70 Papaioannou et al. 75 demonstrated QoL benefits on domains of the Osteoporosis Quality of Life Questionnaire (OQLQ) immediately post treatment, but these did not persist at the 12-month follow-up. A comparison of the two measures found the OQLQ to be superior to the QUALEFFO-41, particularly when used with women with more than one OVF. 173 Future studies should consider whether or not the QUALEFFO-41 is the most appropriate measure for assessing the impact of physical interventions on QoL in this population; alternatively, the QUALEFFO-41 could be used in conjunction with other generic measures of QoL, such as the SF-36.

In contrast, the studies by Evstigneeva et al. ,174 Bergland et al. 65 and Angın et al. 107 do report significant between-group changes using the QUALEFFO-41. In their study, Evstigneeva et al. 174 demonstrated an improvement of 5.8 points in the mean QUALEFFO-41 score immediately after their intervention but the comparison with the control group was influenced by the fact that the control group deteriorated by 3.1 points, emphasising the treatment effect of their intervention. Bergland et al. 65 found that changes in QUALEFFO-41 total scores post treatment at 3 months were not significant, but between-group changes at 12 months were significant.

At 4 months, we observed a mean reduction in the total QUALEFFO-41 score in all three arms, the magnitude of the improvement being greatest for exercise therapy, with a mean reduction of 3.41 points, compared with reductions of 1.95 points for manual therapy and 1.21 points for SSPT. The reduction achieved in this trial at 4 months was similar to the 3.3-point reduction demonstrated by Bergland et al. 65 at 3 months, and our 1.31-point reduction at 12 months is comparable to the 2.8-point change they achieved despite our more active comparator intervention. 65 Thus, at 4 months, exercise therapy had a 2.20-point improvement in QoL compared with SSPT. At 4 months, QUALEFFO-41 subscale scores for pain and social function compared with SSPT also showed improvement, with a trend towards statistical significance (p = 0.068 for pain and p = 0.053 for social function). The unadjusted CACE analysis suggested a weak dose–response effect for QoL, but these differences were not significant or evident in the multiplicity-adjusted CACE analysis.

Timed loaded standing

At present, only a small number of studies involving relatively few participants have used the TLS test, resulting in few data to assist clinicians interpreting test times or estimating what a clinically meaningful change is. Average TLS test times in older populations are recorded in the study by Shipp et al. ,44 with a mean of 110.6 seconds (SD 54.4 seconds) in healthy, older women, and in a RCT involving 99 older adults with hyperkyphosis (16% of whom had vertebral fractures) by Katzman et al. ,175 which recorded a mean of 115.7 seconds (SD 60.5 seconds) and 123.0 seconds (SD 53.1 seconds) in the intervention and control groups, respectively. The average TLS test times at baseline in our trial of 49.7 seconds (SD 59.8 seconds), 47.9 seconds (SD 51.5 seconds) and 55.9 seconds (SD 73.7 seconds) for the exercise therapy, manual therapy and SSPT arms, respectively, are considerably shorter and indicate that our population generally had poor back muscle endurance and greater limitations at baseline. Our TLS test times are most comparable to those in (1) the study by Bennell et al. ,64 in which nine adults with at least one painful OVF in the intervention group had a mean TLS test time of 50 seconds (SD 39 seconds), and (2) the study by Shipp et al. ,44 who studied 121 women with osteoporosis and OVF, 45% of whom had back pain, in which average TLS test times were a mean of 67.6 seconds (SD 56.7 seconds).

At 12 months, increases in back muscle endurance were seen in both active intervention arms but not in the SSPT arm. Mean TLS test times in the exercise and manual therapy arms increased from baseline to 4 months. In contrast to the 4-month time point, improvements in TLS test duration were also found in the SSPT arm at 12 months. These changes were smaller and are less likely to be clinically relevant than those seen in either intervention arm (at a mean of 3.8 seconds or 7% relative to baseline), but the gains meant that differences between both the exercise therapy arm and the manual therapy arm compared with SSPT were non-significant at 1 year. Previous studies have also demonstrated improvements in back strength in the control group in RCTs of exercise in people with osteoporosis, attributing this to participants’ increased awareness, confidence or motivation to undertake physical activity. 70 The educational intervention in the SSPT arm in the PROVE trial may have amplified these effects. Certainly, participants in the qualitative study commented on the value of information and link to the ROS (which promotes exercise) highlighted by the educational session. We also know from the qualitative study and participant diaries that 21% of participants allocated to SSPT sought and received additional physiotherapy. This was more likely to happen after 4 months and might help to explain the improvements in TLS test endurance in the SSPT arm between 4 and 12 months. In addition, the lack of supervision and support for the HEP in the active intervention arms from 4 to 12 months may have contributed to the lack of difference, as gains in TLS test times in the active intervention arms were smaller during this period compared with those in the initial 4-month period.

At 4 months, increases in TLS test endurance were seen in both of the active intervention arms but not in the SSPT arm. The mean increase of 10.56 seconds above SSPT in the manual therapy arm was statistically significant in the unadjusted analysis and showed a strong trend on the multiplicity-adjusted analysis (p = 0.055) (see Appendix 24). It was greater than the mean increase of 7.46 seconds above SSPT in the exercise therapy arm, which was not significant. When compliance with treatment was considered greater, increases in holding time were seen in both arms in those who fully complied compared with those who partially complied.

When the low levels of back muscle endurance in the PROVE trial population at baseline are considered together with the fact that deterioration or no improvement in back muscle endurance would be expected in this group, the increases in TLS test times were arguably clinically significant. 64,175 Relative to baseline, TLS test times increased by a mean of 15% and 22% compared with SSPT in the exercise and manual therapy arms, respectively; within these arms, 34% and 42% increases were seen in those participants who fully complied. A longitudinal cohort study has linked increased back muscle strength with lower incidence of OVF at 10 years. Although strength gains persisted at 1 year, the rates of new OVFs were not formally evaluated radiologically and low rates were reported, with no differences between arms emerging at 1 year. 115

Back fatigue and difficulty carrying out functional tasks in standing because of fatigue are commonly reported by people with osteoporosis and chronic back pain and negatively affect QoL. 44,176 Whether or not the improvements in back muscle endurance resulted in improvements in daily life for PROVE trial participants is uncertain. Interestingly, a range of immediate benefits were attributed to manual therapy by participants in the qualitative study (including improved QoL). However, in this group, the EQ-5D-5L index was worse at all time points and minimal change in QUALEFFO-41 scores was seen at 4 months despite significant increases in endurance. Whether or not the impact of increased endurance on daily life could be captured by the measures we used in the PROVE trial needs consideration. Although increased endurance would be expected to improve performance on some of the items in the QUALEFFO-41 (e.g. meal preparation), the weight of the questionnaire is directed towards other areas (e.g. pain, mobility and mental function). In addition, none of the other outcome measures we used (PASE, SPPB and 6MWT) assessed this dimension, instead focusing on lower-limb function, mobility and exercise capacity.

Improvements in back extensor muscle function were likely to underpin the significant, clinically relevant, increases in standing balance at 4 months that were observed in the manual therapy arm, in which participants underwent no specific balance training. These improvements in back extensor muscle strength are also likely to have contributed to improvements in the exercise therapy arm. The FRT requires the participant to lift the extended arm to 90 degrees of shoulder flexion and reach as far forward as possible without taking a step. This prompts activation of the thoracic erector spinae muscles prior to and during the task, creating an extension moment that counteracts and controls movement of the centre of mass to maintain equilibrium in the anterior–posterior direction. 177,178 This is comparable to muscle activity demanded during the TLS test, so it is not surprising to see both of these measures improving in parallel. 46

Overall, the improvements in back muscle endurance at 4 months for those who received manual therapy suggest that these techniques provide at least short-term benefits on back muscle function. Few trials have investigated the effect of manual therapy techniques on muscle strength or endurance in relevant populations, but our findings are consistent with those that have. In a RCT involving 46 younger women with hyperkyphosis, Kamali et al. 86 reported significant improvements in back muscle strength immediately following a manual therapy intervention. This consisted of 15 treatments over 5 weeks, comprising thoracic posterior–anterior spinal mobilisations, stretches and soft-tissue mobilisation techniques. Bennell et al. 64 also found significant improvements in TLS test duration immediately after a multimodal physiotherapy intervention that incorporated manual therapy. This resulted in a mean change of 45.6 seconds (95% CI 16.1 to 77.3 seconds) or a 65% improvement in holding time compared with baseline. The small, pilot nature of this trial and lack of correction for multiple statistical comparisons may partly explain the extent of its changes. In addition, compared with the PROVE trial, participants in this study, on average, were younger and had fewer OVFs. The intervention combined manual techniques with strengthening exercises, was more intensive, involving 10 sessions of physiotherapy over 10 weeks, and had high levels of compliance.

The mechanisms by which manual techniques might improve endurance in the absence of any resistance exercises, endurance exercises or exercises intended to load muscle require consideration. Reduced pain might result in greater endurance. However, although the QUALEFFO-41 pain subscale for the manual therapy arm shows a mean reduction of –2.54 points (95% CI –6.56 to 1.48 points) at 4 months compared with SSPT, these changes are small and not significant. The application of posterior–anterior joint mobilisations promoted extension of local spinal segments; stretches and soft-tissue mobilisation techniques may have also increased tissue flexibility, allowing for increased active thoracic extension. Small reductions were seen in thoracic kyphosis at 4 months in the manual therapy arm, although these were not significant when compared with SSPT. Improvements in scapulothoracic muscle strength have been observed in people with neck pain and in healthy SSPT participants following spinal mobilisations and stretches without strengthening exercises. 83,179 Joint mobilisations are thought to reduce central inhibition due to pain and stiffness and allow for increased recruitment and activation of spinal muscles, allowing for gains in strength. 83,179 The postural retraining element of the manual therapy programme may have been of benefit. Participants were asked to be aware of their posture and were taught to correct their posture and to practise achieving optimal alignment in everyday tasks, which may have provided sufficient load to prompt increased endurance.

The improvement in TLS test duration in the exercise therapy arm at 4 months suggests that the strength training programme was less effective at increasing back muscle endurance. These findings agree with those of Malmros et al. ,74 who found no significant change in back extensor strength after a combined 10-week intervention of exercise classes followed by a HEP. Similarly, Katzman et al. 175 found that three 60-minute classes per week for 6 months of postural education, back strengthening and stretching exercises in participants with hyperkyphosis (some with OVFs) resulted in no improvement in back extensor strength, spinal muscle density or TLS test time. In this study, change in TLS test time in the intervention arm after treatment was minimal [mean 0.3 seconds (95% CI –11 to 11.6 seconds)] and much lower than we observed in the exercise arm in the PROVE trial.

Subgroup analyses highlighted that participants aged ≤ 70 years in both the exercise therapy arm and the manual therapy arm were most likely to have increased TLS test muscle endurance at 4 months, whereas those aged > 70 years were less likely to benefit. Previous studies that report improved back strength64,70,101 were conducted in younger populations with lower disease severity (less pain and fewer OVFs) than the PROVE trial population. Younger participants may find it easier to participate in home interventions, which provide less clinical support and supervision, and may have been more able to undertake exercises such as back extension exercises in the prone position, which have previously been linked to significantly improved back extensor strength. 70,101,115 Increased age is associated with increased osteoporosis disease severity, number of comorbidities and frailty. 25 These factors may have meant that older participants were less able to tolerate treatments, had less capacity to improve, found it more difficult to attend for treatment, might improve at a slower pace or might need greater levels of supervision.

Ageing also results in neuromuscular changes linked to weaker, less fatigue-resistant, muscles,180 and this may combine with deficits due to spinal pathology to exacerbate back muscle weakness and postural deformity in people with osteoporosis. 180,181 Whether or not exercises were sufficiently graded and whether or not physiotherapists found it difficult to individualise the programme to allow those with the lowest capacity to participate is not known. Some participants in the qualitative study described finding initial stretches and exercise prescriptions too difficult or painful at times, and some participants found treatments in the prone lying position difficult or not possible. This may have particularly affected those with the greatest mobility problems, those with more severe pain at the start of treatment and those with fixed or more severe kyphosis, many of whom would have been older participants. A range of starting positions was available to trial physiotherapists, and guidance was given to omit exercises that were painful, unsafe or not possible. Conversely, the qualitative study also noted positive experiences of exercise prescriptions being adjusted and progressed, allowing for a gradual improvement. Further research into the most appropriate exercise prescription for participants with the greatest limitations is needed. A RCT by Yang et al. 76 found that medication review including analgesia prescription combined with a 4-week twice-daily programme of low-intensity, functional exercises that focused on ensuring that exercises were pain free improved mobility and independence in patients with painful OVF. This was a small study with methodological flaws that included both inpatients and outpatients, but a combined intervention of medication review and exercise prescription could be usefully investigated in an outpatient setting.

The CACE analysis suggested that attending more sessions was associated with greater endurance and increasingly significant outcomes. This effect was stronger in the manual therapy arm, in which, on average, participants attended more sessions than those in the exercise therapy arm. These findings could suggest that additional sessions might have resulted in further gains but also highlight the difficulty of achieving full compliance with treatment. This was particularly true of those in the exercise therapy arm. Overall, exercise therapy participants attended a mean of 4.33 sessions (SD 2.7 sessions); although this would be a common number of sessions in the context of a NHS outpatient physiotherapy department, this is a very limited amount of treatment in comparison with other exercise interventions designed for this patient group. Achieving sufficient treatment intensity is a prerequisite for achieving gains in muscle strength or endurance, including in people with OVF. Further studies of intervention dose and alternative ways of delivering physical interventions are needed in people with OVF to determine how this affects compliance.

Functional reach

The FRT is a measure of dynamic standing balance and stability. 49 Performance on the FRT correlates well with performance on a range of other measures of function, dynamic balance and mobility, such as the timed up and go (TUG) test, step climbing, walking speed, tandem walk and single-leg stand. 182–184 A reach distance of 27.2 cm (SD 0.9 cm) is considered normal for older adults,185 and poor performance on the FRT, of ≤ 18.5 cm, is associated with falls and frailty. 186 Baseline measures of the FRT in all arms in the PROVE trial suggested that most participants had mild to moderate levels of balance impairment.

At 12 months, mean changes in FRT distance had declined compared with the 4-month point in both active intervention arms but continued to be greater than baseline. Mean change in the SSPT arm remained lower than baseline at all time points but improved slightly between 4 and 12 months. Consequently, at 12 months, there were no statistically significant differences between arms, although the mean difference between the exercise therapy arm and the SSPT arm of 1.02 cm (CI –0.96 to 3.01 cm) could be considered clinically valuable.

At 4 months, improved balance was seen in both of the active treatments arms, in contrast to a deterioration in balance in the SSPT arm. Changes in both the exercise therapy arm and the manual therapy arm compared with SSPT were statistically and clinically significant at this point. A 1-cm improvement on the FRT is associated with clinical benefits and lowers the risk of having major mobility problems going forward. 182 Mean changes of > 1 cm were seen in both the exercise therapy arm and the manual therapy arm, and, when compared with the SSPT arm, reach distance was a mean of 2.43 cm (95% CI 0.71 to 4.16 cm) longer in the exercise therapy arm and a mean of 2.11 cm (95% CI 0.57 to 3.64 cm) longer in the manual therapy arm. Greater changes in the exercise therapy arm may have been due to the additive effects of balance and strength training.

Similar improvements in balance after physiotherapy treatment have been reported by several RCTs involving women with OVF, all of which involved higher dose and longer duration interventions. Evstigneeva et al. 174 found that standing balance and tandem walk increased after an intensive intervention of two 40-minute classes per week for 1 year (104 sessions). A 12-month (40 sessions) balance training programme by Madureira et al. 72 significantly improved balance measured by the Berg Balance Scale, the TUG test and Clinical Test of Sensory Interaction for Balance, and fewer falls happened in the intervention arm. Long-term follow-up was not conducted in either of these studies, so whether or not benefits persist once interventions cease is unknown. In studies like ours, which include long-term follow-up, there is evidence that treatment effects decline without supervised practice. Bergland et al. 65 saw significant improvements in balance measured by TUG and FRT distance immediately after a 12-week class intervention (32 sessions). Change in FRT distance was of a similar magnitude to that observed in the PROVE trial and, like in our trial, balance declined when active intervention ceased, so at 12 months differences in functional reach were no longer significant.

Short Physical Performance Battery and 6-minute walk test

Findings from the SPPB and 6MWT will be considered together; both evaluate aspects of lower-limb physical function and mobility, and similar patterns of change were seen in these measures in the trial, indicating change in this underlying domain. The SPPB combines the results of tests of walking speed, balance and sit-to-stand function and assesses lower-limb function and mobility in older adults. 48 Performance on the SPPB is predictive of falls, subsequent disability and difficulty carrying out ADL, risk of hospitalisation and death. 187 Summary scores range from 0 to 12 points, with higher scores representing better function. A change of 0.5 points on the SPPB is thought to represent a small clinically meaningful change and a change of 1 point is thought to represent a substantial meaningful change. 188 The SPPB scores across arms at baseline suggested that, on average, participants had mild to moderately impaired lower-limb function at the start of the trial. 187 At 12 months only the exercise therapy arm showed a small clinically meaningful change in SPPB score; at 4 months, this was larger and statistically significant.

The 6MWT is an assessment of submaximal exercise capacity and walking endurance. Proposed values for the 6MWT in healthy adults range from 400 m to 700 m; distances walked reduce with age and 400 m has been proposed as the cut-off point indicative of mobility limitation and sarcopenia. 186,189 In a RCT by Angın et al. 107 involving 41 women with vertebral osteoporosis, the mean 6MWT distance was 407.82 m (SD 95.38 m) and 402.95 m (SD 90 m) at baseline for the intervention and control arms, respectively. In our trial, baseline values of 6MWT distance were considerably lower in all three arms, highlighting our participants’ low exercise capacities and poor walking endurance. 190 Our values are more comparable to groups with pulmonary disease and probably reflect the association between vertebral fractures and impaired lung function. 29,191 For the 6MWT, a change of 20 m has been estimated to be a meaningful change in response to treatment. 188

In both the SPPB and 6MWT, significant increases were seen from baseline to 4 months in the exercise therapy arm compared with the SSPT arm. In the exercise therapy arm, distance walked on the 6MWT increased by 26.09 m (95% CI 6.58 to 45.6 m), suggesting that change in these measures was both statistically and clinically significant. Angın et al. 107 also found a significant improvement in 6MWT distance after a 24-week (73-session) course of clinical Pilates supervised by a physiotherapist. The younger mean age and lower disease severity of their population, unblinded assessment and the more intensive nature of this intervention may explain the larger increase of 46.0 m (SD 22.75 m) between arms. In the manual therapy arm, changes were not statistically significant, and improvements in 6MWT distance were smaller and unlikely to be clinically significant.

The inclusion of simple functional exercises (sit to stand and step-ups) in the strength training programme, combined with balance exercises and a guided walking programme with a pedometer, resulted in short-term gains in functional mobility and exercise capacity. The qualitative study suggested that participants particularly enjoyed the walking component of the programme and were more motivated to undertake this aspect of the exercise therapy rather than other elements, although, for others, fatigue could be a barrier to engagement. Improvements in mobility may have had positive effects on QoL for participants. Outcomes on the 6MWT have been found to correlate well with changes in the SF-36 physical function subscale. 190 A trend for a significant improvement in the QUALEFFO-41 social function (p = 0.053) subscale compared with SSPT was observed in the exercise therapy arm at 4 months. Leisure and social activities, such as gardening or visiting a theatre or friends, asked about in this area of the questionnaire rely on lower-limb function and walking capacity, and positive changes in these areas may underpin results.

By 12 months, treatment effects had declined. Decline in improvement in physical function after supervised practice ceases is familiar in this population. Iwamoto et al. 122 investigated a pedometer-guided walking programme plus strengthening exercises in women with osteoporosis aged 55–77 years, but outcomes focused on lumbar BMD. At 1 year, BMD improved, but, after a subsequent year of detraining, BMD reverted to levels not significantly different from those in the control arm. Maximum walking speed increased significantly following a 12-week (25-session) circuit class in a RCT by Bergland et al. ;65 although speed declined after treatment stopped, it remained significantly better at 1 year than in the control arm, which had no intervention. Overall, exercise capacity and walking ability in people with osteoporosis and OVF is underinvestigated. Few studies to date have designed interventions that address the limited aerobic capacity and walking endurance in this population despite the relevance of these factors to everyday function and QoL. Measures that capture these domains, such as the SPPB and 6MWT, but also those that consider associated fatigue, have not been widely used or formally studied with people with osteoporosis. The PROVE trial suggests that further research in these areas is warranted.

Thoracic kyphosis

Thoracic kyphosis is expected to progress slowly with age, so that reducing kyphosis by even a small amount may be clinically significant, particularly if treatment effects are maintained over time. 175 Each new vertebral fracture in people with osteoporosis is associated with a 3.8 ° increase in thoracic kyphosis. 192 Even mildly increased kyphosis is associated with an elevated risk of further vertebral fractures due to increased spinal loads and is linked to pain, poor physical and respiratory function, impaired balance and falls. 22 Kyphosis of > 40 ° is recognised as hyperkyphosis and thoracic kyphosis of > 50 ° is associated with increased risk of falls and fractures. 175,192 At baseline, most participants in the PROVE trial were hyperkyphotic. Those in the exercise therapy arm had the highest mean kyphosis, of 51.2 ° (SD 34.8 °), and reported more falls in the previous year than those in the manual therapy and SSPT arms.

Improvements in kyphosis were seen within all three arms at 4 months, with no significant difference between arms. Between 4 and 12 months, kyphosis worsened in the SSPT arm, whereas further improvements in kyphosis were seen in both the exercise therapy arm and the manual therapy arm. There were no statistically significant differences between arms at any point; however, at 12 months the mean reduction in kyphosis of –3.02 ° (95% CI –8.14 ° to 2.10 °) in the manual therapy arm compared with SSPT was a clinically relevant difference and of a comparable size to other trials. Bautmans et al. 63 showed significant improvement in thoracic kyphosis of 3.4 ° following 3 months of manual physiotherapy that included up to 18 sessions of spinal mobilisations and taping and a HEP. They also demonstrated a dose–response relationship to treatment and problems with compliance; a per-protocol analysis in this trial showed that those who complied (at least nine sessions) had significantly more benefit, with a mean improvement (reduction in kyphosis) of –7.1 ° compared with the control.

Katzman et al. 175 completed a trial of targeted strengthening exercise and posture training aiming to reduce hyperkyphosis in older adults in the Study of Hyperkyphosis, Exercise and Function (SHEAF). Not all participants had a diagnosis of osteoporosis and vertebral fracture, with 12 out of 51 participants in the intervention arm and 4 out of 48 participants in the control arm having a confirmed vertebral fracture. Final assessment was at 6 months immediately after completing exercise classes. Katzman et al. 175 reported a mean reduction in thoracic kyphosis of 3 ° in their intervention group compared with their control group but no differences in other outcomes of physical function or QoL. They conclude that a targeted spine-strengthening and posture-training intervention over 6 months may be an effective treatment option for older adults with hyperkyphosis. The primary outcome was thoracic kyphosis measured as a Cobb angle from a standing lateral spine radiograph. They also measured kyphosis using the Debrunner kyphometer; using this method, they reported an identical 3 ° reduction in kyphosis angle. In our trial, we measured kyphosis angle using the from flexicurve as a simple and cost-effective method readily adopted in clinical practice, which did not involve any radiation. We found a mean reduction of 6.8 ° in the exercise therapy arm, 4.63 ° in the manual therapy arm and 2.73 ° in the SSPT arm. Although Katzman et al. 175 reported a between-group difference of 3 ° as being statistically significant, the greater difference of 4.15 ° that we found was not statistically significant; however, it may have been clinically important. At trial entry, the exercise therapy arm had the greatest mean kyphosis at 51.2 °, which was the only arm above the 50 ° threshold associated with increased fall risk and which reported the most falls by participant in the previous year. The within-group change brought the mean kyphosis below this threshold; at 12 months, no significant difference in the incidence of falls was seen between arms. This may be evidence of the stability, or of a ‘static success’, of the intervention, helping to maintain and avoid progression of a condition and associated secondary complications.

Our lower-intensity intervention was associated with clinically important changes in kyphosis in both the manual therapy arm and the exercise therapy arm, which increased in size over the 12-month period, in contrast to the SSPT arm, in which effects declined. Most participants did not receive the complete treatment as planned, and this is likely to have resulted in smaller treatment effects and, along with our more active comparator intervention, explains our lack of statistically significant findings. A range of benefits was attributed to manual therapy in the qualitative study, including decreased stiffness, improved mobility and improved well-being. In the qualitative study, the use of taping was highlighted as being difficult to tolerate and a reason for non-compliance for some participants. The qualitative research highlighted that feedback about spinal shape was very powerful and prompted engagement with treatment but orthotic interventions that are burdensome or uncomfortable can also affect well-being. 193 Bautmans et al. 63 asked physiotherapists and patients about their experiences of treatment. Specific reasons for non-compliance varied, as might be expected, but themes emerged that were also raised in the PROVE qualitative study, including barriers due to mobility or health problems and costs related to treatment (e.g. transport and time). Feedback from physiotherapists in this study included that patients needed lots of guidance with home exercises and that motivation for home exercise was low.

Pain

Chronic back pain is a characteristic feature of vertebral fracture. Over 95% of all participants in the PROVE trial reported experiencing back pain in the fortnight before entering the trial and ≥ 65% in each arm reported pain at the baseline assessment. Using a simple VAS (0 to 10), most participants rated their pain at a moderate intensity at this point. We considered the effects of our interventions on pain using the QUALEFFO-41 pain subscale and a VAS (pain today). In a population with chronic musculoskeletal pain, a change of 1 point or a 15% reduction in pain on a 0–10 VAS is thought to represent a MCID. 194

All arms reported that pain improved at 4 months on the QUALEFFO-41 pain subscale and within each arm there was a further reduction in pain intensity on the QUALEFFO-41 at 12 months. The largest improvement was in the exercise therapy arm, and improvements in the manual therapy and SSPT arms were of the same magnitude. Placebo effects or positive effects of attention provided by clinical and research staff are suggested to affect measures of pain more than physical outcome and may have contributed to the similarity of change in these arms. Alternatively, the educational intervention may also have provided support and prompted behaviour change.

At 4 months, we observed a trend towards significance in the QUALEFFO-41 pain subscale in the exercise therapy arm compared with SSPT arm (p = 0.68). However, the size of the improvement, at a mean score of 3.57 points, compared with SSPT was smaller than that seen in the trials by Bergland et al. ,65 Evstigneeva et al. 174 and Bennell et al. 64 This is likely to reflect the low intensity of our intervention and the fact that our control arm received an intervention. Other studies have reported no change in pain on the QUALEFFO-41 pain subscale but significantly reduced pain using measures such as the SF-36;69 the responsiveness of the QUALEFFO-41 pain subscale may be too low to demonstrate early changes after a short intervention.

At 4 months, small improvements on the VAS (pain today) were seen in each arm, none of which could be considered statistically or clinically significant. Relative to baseline levels, the largest changes were following manual therapy (9.1% reduction), followed by exercise therapy (7.5% reduction) then SSPT (3.2% reduction). In contrast to the direction of change on the QUALEFFO-41 pain subscale, these effects had declined at 12 months. Gold et al. 68 also found no significant change in pain in a trial of 185 women with osteoporosis, despite an intensive class intervention (120 sessions) that included education about coping strategies and provided high levels of psychological support. In their trial, participants had lower levels of pain at baseline and the comparator arm received an educational intervention. Other studies have reported changes in VAS pain intensity after exercise interventions. A small RCT by Karakasidou et al. 195 investigated the effectiveness of 12 weeks (36 sessions) of motor control exercise (including graded lumbar and trunk stabilisation exercises) in a group of 20 women with OVFs. They demonstrated a reduction in pain intensity of 33% following treatment; however, outcome assessment was unblinded and conducted by the physiotherapist who delivered the treatment, exposing the trial to significant bias. Angın et al. 107 found that VAS pain and QUALEFFO-41 pain scales significantly improved after a 24-week clinical Pilates intervention, which also involves specific graded lumbar and trunk stabilisation exercises; again, outcome assessment was unblinded. These trials are interesting, as the exercise therapy arm in the PROVE trial included similar lumbar and trunk stabilisation exercises and showed a positive trend towards significance.

Site principal investigators and research clinicians

Charlotte Heywood and Phillip Clarke (Addenbrookes Hospital), Heather Shilliday (Countess of Chester Hospital), Carol McCrum and Kate Weatherly (Eastbourne General Hospital, Conquest Hospital), Nicky Wilson (King’s College Hospital), Ira Pande, Marie-Josephe Pradere and Rachel Nicole (Queen’s Medical Centre), Tamsin Hughes (Nuffield Orthopaedic Centre), Kim Brown, Anna Thornhill and Emma McLoughlin (Queen Alexandra Hospital), Sarah Moore, Liz Jacobs and Stephanie Travers-Griffin (Heavitree Hospital), David Rogers and Gareth Stephens (Royal Orthopaedic Hospital), Rebecca Hull, Joanne Richardson and Michael Dawson (Royal Surrey County Hospital), Julie Russell and Sophie Park (Royal National Hospital for Rheumatic Diseases), Jessica Shipley and Jacqueline Harrison (Northern General Hospital, Royal Hallamshire Hospital), Steven Young Min (Queen Alexandra Hospital, St Mary’s Hospital), Linda Knott, Nicola Donlin and John Clare (Torbay District General Hospital), Hubert van Griensven, Billy Fashanu, Steven Young and Joanne Calver (Southend Hospital), Zoe Paskins and Lucy Huckfield (Haywood Hospital), David Cumming, Sajeed Kaleel, Maggie Dawson, Sheeba Suresh and Mary Knott (Ipswich Hospital), Jonathan Marks and Emily Watson (Royal Bournemouth Hospital, Christchurch Hospital), Sophia Mavrommatis (University College Hospital), Nicola Clague and Ross Dixey (Leicester General Hospital, Leicester Royal Infirmary) and Chris Mercer and Isobel Amey (Worthing Hospital).