During its four years of activities the VPHOP consortium has developed, validated and deployed a personalised multi-scale modelling technology able to predict in a much more accurate way the absolute risk of fracture in patients with low bone mass and assist clinicians in prognosis and treatment planning (both pharmacological and interventional). VPHOP has recently came to an end bringing some excellent results.
A short interview with the Coordinator, Prof Marco Viceconti - University of Sheffield, will talk us through the project.
VPHOP
Osteoporosis is a disease, which causes bone tissue to thin, making them fragile and prone to breaking after a minor bump or fall. This fragility is due to the reduction of the bone mineral density (BMD), the deterioration of the bone microarchitecture, and the alteration of the amount and variety of proteins in bone, conditions that are very common in women after menopause. Nearly four million osteoporotic bone fractures cost the European health system more than €30 billion per year. By 2050 this figure could have risen to 60 billion Euros.
Osteoporosis itself has no symptoms. Its main consequence is the increased risk of bone fractures, which currently can only be estimated empirically, i.e. based on observations of past cases. The VPHOP project has imagined an alternative pathway, which consists in developing a patient-specific computer model, called "hypermodel" that is capable of assessing the risk of fracture in a deterministic way, with much higher accuracy, taking in consideration multi-level factors.
The occurrence of an osteoporotic fracture is a multi-scale event, which is influenced by:
- the daily loading spectrum, including para-physiological overloading events at BODY LEVEL;
- the fracture event at ORGAN LEVEL;
- the bone elasticity at the TISSUE LEVEL;
- the composition and the morphology of the bone tissue that changes over time due to the metabolic activity, at the CELL activity LEVEL;
- the strength of the tissue due to the molecular composition of the bone matrix, at the CONSTITUENTS LEVEL.
The hypermodel takes in consideration the relevant phenomena taking place at each one of the many dimensional scale levels involved (run in separated sub-models), putting them in connection to one another and making possible to solve this incredibly complex problem.
This modelling technology can solve four clinically relevant problems:
- PROBLEM 1 -Screening: to supplement the conventional Dual X-ray Absorptiometry (DXA) screening of subjects at risk, in order to include in the diagnosis also the propensity to fall, and possibly 3D densitometric information.
- PROBLEM 2 -Diagnosis: in osteoporotic subjects, use 3D densitometry information to develop a personalised assessment of the risk of fracture at the hip and the lower spine that clinicians can use to better modulate the life-style recommendations and the treatment options.
- PROBLEM 3 -Prognosis: for patients at high risk of fracture, develop a predictive model based on tissue-level imaging that estimates the variation of such risk over time due to bone remodelling with, and without, pharmacological treatment.
- PROBLEM 4 -Interventional treatment planning: simulation- based pre-operative planning to decide which vertebral body is at higher risk, in which region of that vertebrae the augmentation would be more effective, and to estimate the reduction of the risk of fracture that the treatment would produce.
The VPHOP consortium included 20 partner institutions, so it would be impossible to provide an exhaustive answer. In general terms, some institutions worked on thedata collection technologies(i.e. EOS Imaging, Philips Medical Systems, Sylvia Lawry Centre, univ Upsala), some focused on themodelling methodsat the various scales (Rizzoli Institute, Charité Berlin, ETH Zurich, TU Eindhoven, KU Leuven, Univ Sheffield), some on thecore technologies(Ansys France, Univ Bern, ARTS Paris, SCS, Univ Berdfordshire, BrainLab) and some conducted theevaluation of the developed technologies(Univ Geneva, INSERM Lyon, Charité, Rizzoli, Icelandic Heart Association, Empirica).
From the outset. All four clinical partners agreed on a baseline clinical protocol, that was used to collect data for over 200 patients. Then each centre explored innovative information sources: Bologna focused on CT, Berlin on whole body modelling, Lyon on the EOS imaging system, and Geneva on the Philips XperCT system.
The full hypermodel currently require over 60,000 CPU hour for each case, so it is impractical for clinical use, for the time being (incidentally we project that in only 5 years time also that full approach will become cost-effective, due to the reduction of computational costs). The reduced hypermodel that we developed, which does not include the bone remodelling at the tissue scale, but only at the continuum scale, has been tested on three retrospective cohorts (Bologna, Reykjavik, and Sheffield) always showing a marked improvement in predictive accuracy over the current standard of care. We also examined about 40 prospective cases to confirm the usability and impact in real world in clinical conditions.
We can provide preclinical and clinical estimates of the predictive accuracy, showing that VPH individualise models are more accurate than epidemiology based standard clinical tools.
Some of our industrial partners are discussing about the possibility to provide VPHOP as a commercial service, in partnership with a company that already offer simulation services in other contexts. Some of the technologies we developed are already being commercialised. The VPHOP pipeline will also be exposed on the VPH-Share Infostructure, for not-for profit usage.
