Two million out of the UK’s 5 million routine diagnostic CT scans performed each year incorporate the thoracolumbar spine or pelvic region. Up to one-third reveal undiagnosed osteoporosis or vertebral fractures. We developed an intervention, Picking up Hidden Osteoporosis Effectively during Normal CT Imaging without additional X-rays (‘PHOENIX’), to facilitate early detection and management of osteoporosis in people attending hospitals for CT scans.
A multicentre, randomised, pragmatic feasibility study. From the general CT-attending population, women aged ≥65 years and men aged ≥75 years attending for CT scans are invited to participate, via a novel consent form incorporating Fracture Risk Assessment (FRAX) questions. Those at increased 10-year risk (within the amber or red zones of the UK FRAX graphical outputs for further action) are block randomised (1:1:1) to (1) PHOENIX intervention, (2) active control or (3) usual care. The PHOENIX intervention comprises (i) retrieving the CT scans using the NHS Image Exchange Portal, (ii) Mindways QCT Pro software analysis of CT hip and spine none density with CT vertebral fracture assessment, (iii) sending the participants’ general practitioner (GP) a clinical report including diagnosis, necessary investigations and recommended treatment. Baseline CT scans from groups 2 and 3 are assessed with the PHOENIX intervention only at study end. Assuming 25% attrition, the study is powered to find a predicted superior osteoporosis treatment rate with PHOENIX (20%) vs 16% among patients whose GPs were sent the FRAX questionnaire only (active control) and 5% in the usual care group. Five hospitals are participating to determine feasibility. The co-primary feasibility outcome measures are (a) ability to randomise 375 patients within 10 months and (b) retention of 75% of survivors, completing their 1-year bone health outcome questionnaire. Secondary 1-year outcomes include osteoporosis/vertebral fracture identification rates and osteoporosis treatment rates. Stakeholder acceptability and economic aspects are evaluated.
Approved by committee (National Research Ethics Service) East of England (EE) as REF/19/EE/0176. Dissemination will be through the Royal Osteoporosis Society (to patients and public) as well as to clinician peers via national and international bone/rheumatology scientific and clinical meetings.
First osteoporosis screening study involving patients attending radiology CT waiting areas.
Individuals randomised to either comprehensive screening for osteoporosis (by applying Fracture Risk Assessment questionnaire, CT bone densitometry and CT vertebral fracture detection) or usual care.
Allows patients the flexibility to self-consent, and tests both the willingness of CT attenders to consent, CT technologies and information flow between trial centres and Cambridge hub.
Limited by an amendment of the primary outcome measure because recruitment during COVID-19 pandemic predicted skew towards the more seriously ill (via cancellation of non-urgent and routine scans).
Modifications to both the retention outcome and longer recruitment window were made to take account of pandemic redeployment of medical research team.
Missing a secondary care direct intervention arm, which would relieve some of the burden from primary care to enact osteoporosis treatments.
Older patients attending hospital for CT scans may have undiagnosed spine fractures or low bone density and therefore stand to benefit from osteoporosis screening. Timely diagnosis and treatment of osteoporosis benefits many patient outcome domains (pain, quality of life, morbidity and in the case of zoledronate treatment, mortality). An estimated 43% of patients >60 years of age undergoing CT scans (for abdominal and pelvic problems unrelated to their bones) were found to have osteoporosis or vertebral fractures when scan images were examined using computer-analysis methods.
Systematic reviews indicate that overlooking osteoporosis or vertebral fractures can lead to serious, preventable pain, disability and health costs (>£2 billion per year from osteoporosis in the UK).
CT attenders are a high-risk population: >30% of older adults have undiagnosed vertebral fractures or osteoporosis that can be effectively treated once identified.
The PHOENIX intervention diagnoses osteoporosis on scans already undertaken, so there is no requirement for additional X-ray exposure or hospital attendance.
The PHOENIX intervention can be applied to avoid the necessity of extra dual energy X-ray absorptiometry (DXA) visits and DXA reporting in multiple clinical scenarios, such as in the setting of radiology reporting quality improvement programmes that increase CT diagnosis of vertebral fractures, or to formally diagnose patients with osteoporosis after large-scale artificial intelligence (AI) vertebral fracture identification methods (eg, Optasia or ZEBRA AI1 tools).
Screening is effective at reducing hip fractures. A recent trial of primary screening based on clinical risk factors and targeted bone density scanning reduced the hip fracture incidence by 28% over 5 years, with a number needed to screen (NNS) of 111 to prevent one hip fracture.
Systematic reviews of screening for osteoporosis suggest a NNS of 43 people (aged 75 years or over) to prevent one vertebral fracture.
Fracture risk is currently assessed opportunistically by general practitioners. Current advances in IT software could augment this practice to make more efficient and effective use of healthcare resources, as recommended by the National Institute for Health and Care Excellence, Clinical Guideline 146.
A 2013 study found that radiologists failed to report 84% of clinically important vertebral fractures from routine CT scans.
There is a fourfold risk of another vertebral fracture occurring after the first one
Generalisability: 1 million CT scans were performed in the UK in 1996, increasing to 5 million in 2016,
This feasibility study aims to test our PHOENIX intervention, which begins by offering adults attending hospitals for CT (for any reason) a comprehensive bone health screen, followed by ‘opportunistic’ reuse of their CT scans for bone densitometry. The intervention ends with their GPs receiving detailed written advice on managing the participants’ bone health. Patients are offered an invitation pack comprising a novel informed consent form/data capture tool that the patient can fill out with or without assistance, and a participant information sheet. Specifically, we want to determine whether it is feasible for our team to undertake a large trial of the PHOENIX intervention to prevent fractures in a larger population of CT attenders. Hence, our co-primary outcome measure is our ability to randomise 375 ‘high-risk’ participants within 10 months. For the larger fracture prevention trial to be feasible, the other co-primary outcome measure must be achieved, which is retention at 12 months; our ability to record bone active treatment data (yes/no) in 75% of surviving patients. Power calculations for any future trial will be informed by the secondary outcome measure; the percentage of participants in each arm who have (i) received a treatment recommendation (due to osteoporosis, vertebral fracture or high FRAX risk) and (ii) received and commenced treatment 12 months after their CT scan.
Patients attending CT waiting rooms are offered a PHOENIX invitation pack comprising a novel informed consent form/data capture tool to ascertain Fracture Risk Assessment (FRAX) score (
The intervention (tested in group 1) comprises (i) retrieving the CT scans of those ‘amber’ or ‘red’ risk individuals with moderate to high 10-year fracture risk by FRAX, (ii) hip and spine bone mineral density analysis with vertebral fracture assessment and (iii) sending the general practitioner (GP) guidance from a clinical decision tree.
The intervention take place ‘behind the scenes’ without impacting on patient flow or the radiography/radiology workflow.
Data (imaging, clinical risk factor, Patient Identifiable Information (PID) including sensitive clinical information) flow via secure encrypted and approved methods including the NHS Image Exchange Portal, Picture Archiving and Communications System (PACS) and nhs.net secure clinical email.
Technical staff can ‘pull’ clinical CT images for assessment at any time.
Feasibility of recruitment: defined as ability to randomise 375 ‘at-risk’ participants over continuous 10-month period.
Retention: defined as retaining 75% of surviving participants to 12 months postrandomisation, proven by receipt of follow-up postal questionnaire (specifically an answer to the bone active medication received; yes/no question).
New vertebral fracture identification across three arms assessed at 12 months.
Osteoporosis identification across three arms assessed at 12 months.
Bone active medication initiation across the three arms assessed at 12 months.
Survival at 12 months across three arms assessed at 12 months.
Attending for a routine clinical CT scan.
Able to provide informed consent.
Aged ≥65–90 years (women) or ≥75–90 years (men) attending for CT for any clinical reason, where the spine and/or hips are visible in scan images.
Metalwork in hips (due to streak artefact on the contralateral hip within CT).
Known to be receiving prescription treatment for osteoporosis other than calcium/vitamin D (ie, bisphosphonate drug, strontium ranelate, denosumab, raloxifene or teriparatide since it would be futile to identify already known and treated osteoporosis).
Prone CT scan (which cannot currently be processed by the software).
Actively participated in the study design in focus groups within the project team.
Assisted in the design of data collection tools, including the informed consent form and information leaflet.
Provided guidance on the acceptability of approaching participants in CT scan areas and insight into the anxiety of attending hospital among some more elderly members of society. Adding country of origin and ethnicity questions to consent form to facilitate FRAX.
Impact of COVID-19 pandemic
Recruitment had to be paused due to the COVID-19 pandemic, then restarted in just one centre without face-to-face contact with patients. Therefore, recruitment was either by clinical approach/discussion versus self-consent based on just handing out recruitment packs after COVID-19 restrictions were brought in.
The PHOENIX study is a parallel three-group, non-blinded randomised controlled study of volunteers aged ≥65–90 years (women) or ≥75–90 years (men) attending for CT for any clinical reason, where the spine and/or hips are visible in scan images. Specifically, CT-attending patient volunteers who have an elevated 10-year fracture risk (within the amber or red zones of the UK FRAX graphical outputs for further action) are randomised to one of three groups. The three arms allow comparison between (1) the PHOENIX intervention, (2) an alternative active control ‘middle’ group (to assess whether GPs will investigate or treat patients based solely on receipt of a populated FRAX questionnaire) and (3) a control (‘usual care’) group where no action is taken. Potential participants are identified ahead of attendance by the clinical or research team reviewing CT lists. Those eligible by age, sex and region of the body being scanned are handed the PHOENIX pack while attending the CT department. The specially designed informed consent form (
On the last page of the consent form, the eight validated bone health-relevant questions from the FRAX questionnaire are completed, along with patient address, telephone number, GP name and GP address. Completed paperwork is scanned and sent by email attachment within the national, clinically secure NHS.net email account infrastructure to our add-tr.p*******@nhs.net address (except for local patients where the paperwork is collected directly from the CT reception). FRAX answers are processed by the research team at Cambridge who use the results to calculate the 10-year fracture risk score
Study flow chart for the Picking up Hidden Osteoporosis Effectively during Normal CT Imaging without additional X-rays (PHOENIX) study. BMD, bone mineral density; GP, general practitioner; FRAX, Fracture Risk Assessment; VFA, vertebral fracture assessment; MSK, Musculoskeletal.
The PHOENIX intervention. In brief, bone health for the red/amber risk participant is assessed using the CT images already captured, with computer-aided diagnosis of vertebral fractures performed by one of two trained NHS and accredited bone density technicians using Mindways QCT Pro (Mindways, Austin, Texas, USA) in Cambridge. The output is a consultant clinician-verified report (
Written clinical guidance and report produced at the end of a Picking up Hidden Osteoporosis Effectively during Normal CT Imaging without additional X-rays (PHOENIX) intervention in a typical patient from group 1. These comprise the standard output pdfs of Mindways QCT Pro which are sent to the participants’ general practitioner (GP). In this case, there are both hip and spine results from a typical patient with a previously undiagnosed grade 3 vertebral fracture of T12 (found and graded using SlicePick). The femoral neck bone mineral density (BMD) T-score is −3.03 and spinal osteoporosis is shown with L1-L3 volumetric BMD of 63.4 mg/cm3. Their Fracture Risk Assessment (FRAX) 10-year risk of fracture is 25% and investigation/treatment is recommended by NOGG. The written clinical guidance is derived from the PHOENIX clinical decision tree (
CT scan retrieved via PACS into Mindways via its DICOM transfer tool (Cambridge) or in the case of the spoke sites, via the UK NHS Image Exchange Portal to Cambridge, followed by importing to Mindways QCT Pro.
The Mindways QCT Pro SlicePick vertebral fracture assessment tool is used to identify any fractures in the imaged CT volume during the bone density assessment, by applying 6-point morphometry criteria to the synthesised lateral image (upper panels of
Mindways QCT Pro CTXA module is used to measure bone mineral density (BMD) in the left femoral neck (
Each individual’s Fracture Risk Assessment (FRAX) fracture risk estimate is then recalculated online, using clinical Risk Factors (RFs) from FRAX and the femoral neck BMD value. Importantly, the ‘previous fracture’ category is switched to ‘yes’ in FRAX if spine imaging revealed any vertebral fracture.
The NOGG guidance tool button within FRAX is then used and the pop-up red/green risk charts followed to determine whether treatment is recommended, provided no contraindications exist. In the case of no femoral neck BMD possible (spinal BMD only), FRAX values are reported (without BMD input) and the PHOENIX clinical decision tree for spine BMD is followed to produce the report that is sent to general practitioners (GPs).
The clinician-verified report phrases are derived from the PHOENIX clinical decision tree (
The final Hip and Spine pdf is signed off electronically by the consultant physician (KESP) via nhs.net email (secure clinical, encrypted) and posted to GPs with a cover letter.
Any vertebral fractures overlooked by radiologists on the Radiology Information System (RIS) report are entered in the local radiology discrepancy review process and the original CT report addended accordingly.
Responsibility for subsequent investigations including referrals for dual energy X-ray absorptiometry and/or treatment, remain with the GP.
A small, cylindrical Mindways model 4 (asynchronous CT calibration phantom) must be scanned at intervals on each CT scanner, including all the various kilovoltage peak energies used clinically, according to a Standard Operating Procedure. This phantom can be posted or taken to radiology departments for quality assurance scanning.
The
Active control. Only the completed FRAX questionnaire results are sent to the GP, who can follow NOGG guidance themselves by entering responses onto the online FRAX website calculator. However, the patient’s data are put through the PHOENIX intervention at study exit (
Usual care. The GP receives a letter informing them of their patient’s involvement in the study, but no further reports or information at this point. However, the patient’s data are put through the PHOENIX intervention at study exit (
Baseline and 1-year follow-up postal questionnaires are sent to all randomised participants to collect data on quality of life and resource use relevant to bone health for health economic analysis. For non-responders, a telephone interview is attempted aiming to collect the questionnaire data. Patient and clinician/healthcare worker/administrator perspectives of study involvement are explored through semi-structured interviews by telephone. All study data are held securely processed in accordance with the Data Protection Act 2018 and not disclosed to third parties. Manual records will be held securely (eg, in locked filing cabinets). Electronic records will be held on a secure network requiring user ID and password access. Individuals are not identifiable from the REDCAP and JMP (V.15, SAS Institute) spreadsheet results of the trial. The Project Management Team members (KW, KESP, DC, JMB) are responsible for the day-to-day running of the study to discuss practical aspects and ensure it is progressing. The Trial Management Group (TMG) comprises the protocol authors and one patient and public involvement (PPI) member. They are responsible for the overall conduct of the study and ensuring the National Institute for Health and Care Research (NIHR) milestones are achieved. Prior to COVID-19, separate PPI meetings were held which enabled those attending to influence the study design.
Our primary intention is to evaluate the technical procedures needed for a definitive trial with fracture incidence (rather than recruitment/retention) as the primary outcome. To help set the target recruitment numbers for the feasibility study, we therefore used the surrogate outcome of osteoporosis treatment rates at 12 months follow-up to permit sufficient power to assess and conduct cross-group comparisons across trial arms within the period of recruitment to the feasibility study. Scenarios for scaling up to a definitive trial are based on previous data.
Fracture Risk Assessment (FRAX), sample size and power calculations for Picking up Hidden Osteoporosis Effectively during Normal CT Imaging without additional X-rays (PHOENIX). (A) FRAX UK age-specific fracture risk assessment and intervention thresholds. (B) Age-specific prevalence of red/amber FRAX risk category from preliminary local application of PHOENIX intervention in Cambridge. (C) Calculation of the number of invites needed to randomise 375 red/amber risk participants based on different responses (rates of agreeing to participate). (D) Number of PHOENIX invitation packs that must be given out in order to randomise 375 red/amber risk participants based on different prevalence of red/amber FRAX among participants. (E) Our feasibility study aiming to randomise 375 of 938 (40%) consenting participants (ie, 30% of 3125 total invited) would estimate the trial’s response rate with a precision of 1.6% (ie, half-width of the 95% CI,
Hypothesising that the recruitment (ie, consent, FRAX completion, eligibility and CT imaging, ie, ‘response’) would be similar to that observed in the SCOOP primary prevention trial (p0=30%); a feasibility study aiming to randomise 375 of 938 (40%) consenting participants (ie, 30% of 3125 total invited) would estimate the trial’s response rate with a precision of 1.6% (ie, half-width of the 95% CI,
Assuming that osteoporosis treatment percentages at 1-year follow-up in the usual care versus active control FRAX-GP versus PHOENIX intervention would be 4.5% vs 15.5% vs 20.3%, respectively (ie, as observed in SCOOP trial vs with expected improvement in PHOENIX), a total of 375 participants recruited and randomised in the ratio 1:1:1 will have 96% power (
Knowing that 375 participants were needed, we were advised to undertake block randomisation stratified by centre, sex and age group (female: <75 vs ≥75 years, male: <80 vs ≥80 years). Randomisation uses the web application Research Electronic Data Capture (REDCap
Daily screening logs will be kept determining: (i) the total number of days (n) where patients were approached in each centre and (ii) the daily totals of PHOENIX invitation packs distributed. This log will also record if a numbered pack was not given out on a day (eg, if the patient did not arrive for scanning, or the receptionists omitted to give out the pack). The total number of consented individuals (x) as well as the number of consented individuals per day (total x/n overall, and per centre) will be calculated. The number of consented but subsequently excluded individuals (y) will be kept (eg, for subsequently discovered metalwork in the scan, technical issues with scan). The number of randomised participants will be calculated (z=x−y), and the rate of randomisation (z/n overall, and per centre). These data will be captured from screening logs and entered manually onto a database. The REDCap database will be used to record details of all consenting patients (z). The database will be locked on completion of data entry and when missing data or any discrepancies are resolved. The data will be exported for statistical analysis to the study statisticians with assistance from the chief investigator.
Response rates will be calculated as the proportion of participants invited who (a) consent to take part, (b) meet eligibility criteria and (c) are randomised, with 95% CIs to assess the extent to which the realised response deviates from the expected as detailed in the power calculations above. Cumulative participant accrual over time will be plotted overall and by recruitment centre to assess any major differences in accrual and assess progress towards target. Logistic regression will be used to compare randomised groups with respect to the 12-month outcomes of treatment rates and participant retention. Baseline stratification covariates (ie, centre, sex and age) will be adjusted for in the models. Analyses will be on an intention-to-treat basis.
Both the baseline and 1-year follow-up questionnaires being sent to all patients include space for participants to feedback their experience of the study. First explaining how this will help, this section includes example open questions such as
At the ‘idea’ stage of the research, our small team presented our idea at the large workshop
Completion rates on the quality-of-life measure (EQ-5D-5L) and resource use items included within the postal survey will inform whether an economic evaluation is feasible using these data collection approaches in the definitive trial. Patterns of response and non-response will be explored to see if they suggest ways the postal survey can be refined to improve data completeness.
Approved by committee (National Research Ethics Service) East of England (EE) as REF/19/EE/0176. Dissemination will be through the Royal Osteoporosis Society Bone Academy (to patients and public) as well as to clinician peers via national and international bone/rheumatology scientific and clinical meetings.
The PHOENIX study is dedicated to the memory of J. Keenan Brown, inventor of QCT Pro (9 February 1960 to 11 December 2021). Without Keenan, this study would not have been possible. The authors would like to acknowledge Karen Blesic for providing a valuable nursing perspective, for helping develop the idea and for her major contributions towards the NHS Research Innovation Fund pilot to PHOENIX (the clinical service called CORTEX). The authors acknowledge the contributions of Polly Barnes and Charlotte Tyson to CORTEX and also their contributions towards the practicalities of imaging data flow. The authors acknowledge Judith Brown’s contributions towards the practicalities of patient flow and analysis through the pathways and imaging services. The authors acknowledge Jeremy Dearling as our primary PPIE representative who gave valuable contributions to the design of the overall study, provided detailed feedback on the materials and protocol development. The authors also acknowledge the assistance of the Eastern region NIHR Research Design Service team, specifically Jonathan Scales, Andrew Sharpe, Analisa Casarin (improving and developing the research idea and application) and Doreen Tembo (PPI lead for the same). KESP acknowledges the Cambridge NIHR Biomedical Research Centre who support his work, and support DC. KESP also thanks the NIHR for funding his attendance at the Pragmatic Clinical Trials Course at Queen Mary University of London which helped him develop the idea. Finally, the authors acknowledge the major contribution of Dr Alan Lamont, chair of East of England (EE), Research Ethic Committee whose helpful suggestions and understanding of the relevant GCP issues and legislation rekindled our interest in adopting pragmatic patient consent without a researcher present.
KESP: chief investigator for the study and is responsible for the concept and design of the overall study and manuscript final approval. KESP and EC provided clinical expertise in the area of bone health which forms the basis for the study. They developed the concept and idea for PHOENIX, gained funding and wrote the clinical decision tree with assistance from DC. Also methods design, writing and providing feedback of manuscript, manuscript final approval. KESP is responsible for the medical aspects of opportunistic diagnosis of osteoporosis, recommending investigations and treatments to participants. SK (with assistance from the relevant trials work of LS, and pilot work of KESP/DC) developed the statistical analysis plan including calculation of sample size and definition of primary and secondary outcomes and power calculations. APW designed the health economics work, and developed suitable outcomes for the trial. DC provided information on the background and rationale for using the various software. He has helped clarify pathways for scan analysis in the study and assisted in the development of the Clinical Decision Tree. KW wrote the protocol and drafted the manuscript, clarified pathways and study flow and liaised with the team to finalise the protocol. JF designed the process evaluation substudy, with all qualitative interview work, and developed suitable outcomes for the trial. TT provided information on the background and rationale for diagnosing vertebral fractures. He was responsible for the medical aspects of opportunistic diagnosis of vertebral fractures. All authors contributed to and approved the final version of the manuscript.
This project is funded by the National Institute for Health Research (NIHR) under its Research for Patient Benefit (RfPB) Programme (Grant Reference Number PB-PG-0816-20027) and by the Cambridge NIHR Biomedical Research Centre (BRC-1215- 20014). Funding is in place to 31 March 2022.
The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.
None declared.
Patients and/or the public were involved in the design, or conduct, or reporting, or dissemination plans of this research. Refer to the 'Methods and analysis' section for further details.
Not commissioned; externally peer reviewed.
All authors agreed to be accountable for all aspects of the work to enable questions relating to the accuracy or integrity of any part of the work to be appropriately investigated and resolved.
This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.
Data are available on reasonable request. Researchers involved in the trial will have access to the full dataset. Applications for access to the final trial dataset will be through the Trial Management Group (TMG). All members of the TMG will have full access to the dataset, and a controlled access model (openly available to all applicants) will be followed as set out in the MRC and NIHR guidance (
Not applicable.