EXOGEN ultrasound bone healing system for long bone fractures with non-union or delayed healing

The sponsor identified 3 economic studies, all in UK settings. Taylor et al. (2009) carried out a cost-effectiveness analysis on non-union tibial fractures treated by EXOGEN or by surgery (intramedullary nailing). The model developed for this study was adapted for use in the sponsor's submission. Kanakaris et al. (2007) presented a non-comparative analysis of the cost of compression plate fixation and bone grafting to treat aseptic non-union long bone fractures, and Patil et al. (2006) reported a similar analysis for the Ilizarov surgical procedure to treat complex non-union tibia or femur fractures.

The sponsor submitted a de novo cost analysis for EXOGEN. Full details of all cost evidence and modelling considered by the Committee are available in the assessment report overview.

Two cost models were submitted by the sponsor – 1 for non-union and 1 for delayed healing (both adapted from the model by Taylor et al. 2009). Markov models with a 1‑year time horizon and monthly cycles were used to carry out each cost analysis. The patient population included patients with fractures of the tibia initially treated by surgical insertion of an intramedullary nail.

For non-union fractures, the cost model evaluated the costs and consequences associated with the use of the EXOGEN 4000+ at diagnosis of non-union, followed by further surgery if the fracture did not heal within 6 months, compared with surgery at diagnosis, followed by repeat surgery if the fracture did not heal within 6 months.

The non-union model had 4 health states: 'non-union fracture', 'healed fracture', 'infection' and 'post infection'. All patients began in the 'non-union fracture' health state. Patients in the EXOGEN arm had treatment with EXOGEN 4000+ from baseline, whereas patients in the control arm had surgery at baseline. In both arms, if healing had not occurred after 6 months in the non-union fracture health state, it was assumed that further surgery was needed. In the surgery arm, patients were at risk of infection as a complication of surgery from the time of diagnosis of non-union, and also if they had further revision surgery after 6 months in the non-union state. The model assumed that no infection would occur in the EXOGEN arm.

In the sponsor's base case for non-union fracture, the key assumptions were cited as follows:

• Healing rates and healing times are equivalent for both the EXOGEN 4000+ and surgery in the case of stable, well-aligned fractures.

• Average length of bed stay for surgery is 4.9 days (Hospital episode statistics online 2010/11).

• Average theatre time for non-union surgery is 3 hours.

• All initial non-union surgical management includes autologous iliac crest bone graft.

• In the EXOGEN 4000+ group, in most cases, only 1 additional operation will be offered over 1 year if the fracture has not healed.

• Non-procedure-related costs (for example, physiotherapy, X‑ray) are the same in both treatment arms.

• Infection rates in the EXOGEN 4000+ and control groups are assumed to be 0% and 1.4% per month (Health Protection Agency, 2011) respectively.

• Infection lasts for a maximum of 2 months, but all costs associated with its treatment are incurred in the first month.

• In the case of osteomyelitis, staged revision surgery is carried out.

• Patients with osteomyelitis are given intravenous antibiotics in hospital over a minimum of 3 weeks.

For delayed healing, the costs and consequences associated with the use of the EXOGEN Express at diagnosis of delayed healing followed by surgery if the fracture did not heal within 6 months (9 months after fracture), were compared with no intervention at diagnosis followed by surgery if the fracture did not heal within 6 months.

The delayed healing model had 5 health states: 'delayed union', 'healed fracture', 'non-union', 'infection' and 'post infection'. All patients begin in the delayed union state. It was assumed that surgical intervention (intramedullary nailing) had been carried out before delayed healing was diagnosed, shortly after the fracture occurred. The model for delayed union was run twice; once for the EXOGEN arm, when patients started using the EXOGEN Express device at the beginning of the modelling period; and once for the control arm, when patients were assumed to have no further treatment (observation only) until non-union was diagnosed. In subsequent cycles, patients could move to 'healed fracture' (an absorbing state), 'infection', or after 6 months in the model, to 'non-union'. After infection, a staged revision surgery process began, with the administration of intravenous antibiotics and removal of metalwork. It was considered that the infection would take 2 months to clear, at which point revision surgery would take place. Patients could become re-infected having previously moved into the post-infection state. After 6 months of delayed healing, and no infection occurring, the patient could progress to 'non-union fracture', when further surgery would take place. In subsequent cycles, non-union fractures may have healed or become infected.

In the sponsor's base case for delayed healing the key assumptions were as follows:

• For both arms in the model, patient treatment pathways start with a surgical intervention (insertion of an intramedullary nail) to treat a fresh fracture.

• On diagnosis of delayed union, the patient will either have treatment with the EXOGEN Express or will receive no further treatment (observation only) until either bony union is achieved or non-union is diagnosed.

• Healing rates for delayed healing at 6 months are a linear progression of those reported at 4 months in the Schofer et al. (2010) study in the absence of any comparative data on healing rate from other randomised controlled trials.

The cost models were from an NHS cost perspective. The cost analyses included costs associated with surgery (including surgical intervention, theatre time, drugs, bed stay) and costs associated with GP visits, outpatient visits, treating infection (including surgery and medication), X‑rays, wheelchair, crutches and physiotherapy.

In the sponsor's base case for non-union fractures, the average cost per patient for the EXOGEN 4000+ device was £4,647 and the average cost per patient for surgery was £6,957. The EXOGEN 4000+ was therefore associated with a cost saving of £2,310 compared with surgery. The sponsor carried out a deterministic sensitivity analysis to vary the rates of healing and infection. The analysis showed that the model is not sensitive to changes in rates of healing and infection, and the EXOGEN 4000+ remained cost saving for non-union fractures in all scenarios tested.

For delayed healing, the sponsor's base case presented an average cost per patient of £4,290 for the EXOGEN Express and £4,974 for current management (observation followed by surgery at non-union if needed). The EXOGEN Express was therefore associated with a cost saving of £684 per patient on early use compared with current management. The sponsor varied the rates of healing and infection in a sensitivity analysis and showed that the model is sensitive to changes in these parameters.

For the non-union model, the External Assessment Centre considered that a number of the assumptions were not justified and it made several changes to the sponsor's base-case model. These were as follows:

• applying the healing rate for EXOGEN from Mayr et al. (2000)

• allowing for infection in the EXOGEN arm following surgery at 6 months

• applying a one-off rate of infection following any surgery

• adjusting the post-surgical infection rate

• correcting minor errors in the model.

For the non-union model, the External Assessment Centre's additional analysis showed average costs per patient for the EXOGEN 4000+ of £5,688 and for surgery of £6,852. The EXOGEN 4000+ was therefore associated with a cost saving of £1,164 compared with immediate surgery for non-union. Sensitivity analysis showed that the model is relatively insensitive to changes in assumptions about the relative effectiveness of surgery compared with EXOGEN. The External Assessment Centre considered that the EXOGEN 4000+ is significantly cheaper than surgery.

In a 2‑way sensitivity analysis, varying the baseline healing rate with EXOGEN, and the relative risk of healing with surgery compared with EXOGEN, showed stable results. Only if the healing rate with EXOGEN was reduced to its lower limit and the relative risk of healing with surgery increased to its upper limit did EXOGEN become more expensive than surgery. The External Assessment Centre also carried out sensitivity analyses to apply no delay to the onset of healing, add VAT on devices and consumables, and use healthcare resource group costs for infection and surgery. The EXOGEN 4000+ remained cost saving for all scenarios tested.

For the delayed healing model, the External Assessment Centre considered that several of the sponsor's assumptions were not justified and made a number of changes to the base-case model. These changes included:

• allowing for infection in the EXOGEN arm following further surgery for patients who have not healed after 6 months (9 months after fracture)

• changing costs to apply to delayed healing resource use (as per model) at baseline, not fresh fracture

• adjusting the infection rate and associated costs as for non-union.

After applying these changes, the External Assessment Centre estimated results for 8 scenarios that reflected different sources of healing rates (Mayr et al. 2000 for the EXOGEN Express arm and relative risk from Schofer et al. 2010 compared with Schofer et al. 2010 alone), different assumptions about the minimum time to healing following surgery and EXOGEN (no delay compared with 2‑month delay before healing is observed), and the persistence of relative benefits of EXOGEN (persistence of enhanced healing rate compared with no persistence between the end of EXOGEN treatment at 4 months and further surgery if needed at 6 months).

For delayed healing the External Assessment Centre's preferred scenario from among the 8 in its report applied the following key assumptions:

• The best estimate of the healing rate with EXOGEN is from the register data reported by Mayr et al. (2000).

• The best estimate of relative healing rates with the EXOGEN Express compared with no further treatment until non-union is provided by Schofer et al. (2010).

• It is reasonable to assume that healing following either surgery or the start of treatment with the EXOGEN Express will not usually be observed within 2 months (expert opinion).

• It is conservative to assume that EXOGEN does not continue to enhance the background healing rate once ultrasound treatment has finished after 4 months (the duration of follow-up in Schofer et al. 2010).

The External Assessment Centre's preferred scenario for the treatment of long bone fractures with delayed healing showed a total cost for the EXOGEN Express of £3,033 and a total cost for current management of £2,529. The EXOGEN Express was therefore associated with a cost increase of £504 per patient compared with observation followed by surgery at non-union if necessary.

The External Assessment Centre carried out a 2‑way sensitivity analysis in which the baseline healing rate was varied with the EXOGEN Express and the relative risk of healing compared with control using the preferred scenario. They found that the results were not sensitive to varying these estimates: the EXOGEN Express remained more costly than waiting to see if fractures healed without further intervention. The External Assessment Centre carried out further sensitivity analyses to vary the risk of infection, applying VAT on devices and consumables, using healthcare resource group costs for surgery and for the treatment of infection. The EXOGEN Express remained more expensive than the comparator for delayed healing under all of the scenarios tested.