Although the published economic evaluations identified in the economic literature review addressed the interventions of interest, they did not use Canadian costs, nor did the authors take a Canadian perspective. Owing to these limitations, we conducted a primary economic evaluation using Ontario-specific cost inputs and clinical care pathways.
If you want to learn more, please visit our website Wonderfu.
Finally, we conducted a scenario analysis where all implant surgeries were single-stage instead of two-stage. We estimated that compared to two-stage surgery, single-stage surgeries would have lower costs, as they would have reduced inpatient stays and would not require multiple operating room days to complete the procedure. As there is currently no published literature evaluating single-stage osseointegrated prosthetic implant surgery for lower-limb amputees, the model assumed the same clinical effectiveness, utility gains, and complication rates as two-stage surgery.
We excluded disutilities from the reference case as they used measurement methods (i.e., EuroQol [EQ-5D] and Assessment of Quality of life [AQoL] instruments) that differed from the rest of the core model utility values, which were derived from the SF-6D survey. In a scenario analysis, we used disutilities found in the literature for fractures and deep infections. In the model, when a complication occurred in a health state, the state was assigned a disutility value representative of the amount of time individuals would be affected by the complication. The resulting QALYs incurred were calculated as the utility value of the current health state, minus the utility value of the complication multiplied by the duration of the complication (representing the disutility). For example, to calculate the change in QALYs for an individual who is in the &#;osseointegrated prosthesis&#; health state and who has a fracture during one model cycle, 0.692 would be subtracted by the result of 0.120 &#; 0.5 (with 0.5 representing 6 of 12 months), which would result in a QALY of 0.632. 26 , 57 We also tested alternative utility values for the health states in additional scenario analyses.
We pooled event rates for the reference case, as the available studies were of equally low quality and reported varying event rates. In a scenario analysis, we used time-varying event rates from Branemark et al. 10 Expert opinion informed us that the risk for complications (e.g., superficial and deep infection), although always present, may be higher immediately following surgery (James Waddell, MD, communication, September ). We calculated time-varying rates for 0 to 12 months postsurgery and 12 to 24 months postsurgery. Rates at 0 to 12 months were assigned to the health states of &#;implant surgery and recovery&#; and &#;reimplant,&#; while rates at 12 to 24 months were assigned to the health state &#;osseointegrated prosthesis.&#;
The reference case parameter for stump revision rates came from a previously published cost&#; utility analysis that cited rates from hospital data. 45 Furthermore, we were unable to determine a measure of uncertainty around the estimate. Therefore, in a scenario analysis we used the soft tissue refashioning rate in the osseointegration cohort for the stump revision rate in the socket cohort.
The reference case included costs for femoral fractures but only for the osseointegration cohort, due to insufficient data on the rate of femoral fractures among socket prosthesis users. But that assumption may overestimate the costs for patients with osseointegrated prosthetic implants if the fracture rate in a socket cohort is not close to zero. Therefore, in a scenario analysis, we excluded fractures to consider the possibility that the two cohorts have similar risk of fracture and that the incremental cost difference for treatment is marginal.
As identified in the clinical evidence section of this report, a deep infection required surgical intervention in all cases in the literature. Our reference case assumed that patients with deep infection stayed in hospital an average of 9.9 days for postoperative care, but clinical experts indicated that IV antibiotics could be administered in an outpatient or home care setting (Nancy Dudek, MD; Wade Gofton, MD; communications, November ). Therefore, a scenario analysis considered this possibility, and based on the results of Wolter et al, 56 we assumed home care expenses would be half the cost of an inpatient stay.
We modelled differences in maintenance cost in the scenario analysis, such as the number of visits to a prosthetist, as outlined by Haggstrom et al, 52 by assuming 7 visits to a prosthetist per year for a socket user and 3 visits for an implant user. The costs were calculated by using the hourly fee for a prosthetist in Ontario ($183.29/hour) and assuming each adjustment session takes 1 hour for an implant user and 2 hours for a socket user (Nancy Dudek, MD, communication, August ).
As a conservative assumption, the model currently incorporates the costs of both inpatient and outpatient rehabilitation after implant surgery. However, due to the intensive inpatient rehabilitation process, it is possible that patients may not require additional rehabilitation services in the outpatient setting. Therefore, we explored excluding it in a scenario analysis.
As described in , we conducted several scenario analyses. Given that osseointegrated prosthetic implant surgery has yet to be performed in Ontario, these scenario analyses tested not only different input parameters, but also some of the assumptions required to estimate Ontario-specific costs. For each scenario, we recalculated the mean incremental costs and QALYs for each treatment, along with the ICER. All scenarios were performed probabilistically unless otherwise stated. Appendix 6 , and , provide a full list of input parameters.
We conducted deterministic sensitivity analyses to assess how sensitive our reference case results were to specific parameters. In the one-way sensitivity analyses, we varied specific model variables (e.g., transitional probabilities, costs, utilities) and recorded and presented their impact on the results in a tornado diagram. Details of these analyses and the specific parameters varied are presented in Appendix 6 , .
For the reference case analysis, we performed a probabilistic analysis to determine the mean incremental cost and mean incremental QALYs, and we calculated the incremental cost-effectiveness ratio (ICER) for an osseointegrated prosthesis compared with an uncomfortable socket prosthesis. We performed a probabilistic sensitivity analysis by running 5,000 Monte Carlo simulations to capture parameter uncertainty. When possible, we specified distributions around each estimate, using the mean and standard deviation. Costs were characterized by gamma distributions, and probabilities and utilities were characterized by beta distributions.
describes costs for complications. We assumed all complications except superficial infections were treated in an inpatient setting. Femoral fractures can be subdivided into fractures requiring fixation (stable stem) and complex fractures requiring fixation and implant revision (unstable stem). As the rates of femoral fracture related to osseointegrated prosthetic implants reported in the literature did not differentiate between stable and unstable stems, we assumed the rate for implant revision would include fractures requiring both fixation and implant revision; therefore, femoral fractures were costed as requiring only fixation (i.e., for people with these fractures, an implant revision is unnecessary). For superficial infections, we assumed patients would visit their physiatrist and receive a prescription for antibiotics. Patients with a deep infection would be admitted to a hospital and undergo a debridement procedure.
contains the costs components of the remaining model health states (i.e., osseointegrated prosthesis, implant extraction, reimplant, uncomfortable socket prosthesis). In both the &#;osseointegrated prosthesis&#; and &#;uncomfortable socket prosthesis&#; health states, we assumed patients would have a yearly physiatrist check-up. However, when costing physiatrist visits for complications, we assumed the socket cohort had 4 annual visits, while the osseointegration cohort had their physiatrist visits built into the model whenever a complication occurred (Nancy Dudek, MD, communication, August ). Fixing the number of physiatrist visits for socket-related complications was necessary as few complication rates were available in the literature to inform the model, yet experts indicated this population would be frequently seen for active problems such as socket pain and cysts (Nancy Dudek, MD; James Waddell, MD; communications, August to September ). For the &#;reimplant&#; health state, we assumed that inpatient costs and prosthetic fitting would both cost 25% less than the initial implantation. Expert consultation indicated that rehabilitation following reimplantation would go through the same progressive weight-bearing following a repeat surgery, but there would be no need for additional gait training once full weight-bearing was achieved, as the patient would have already learned how to optimize their walking pattern (Nancy Dudek, MD, communication, August ).
presents itemized costs for the &#;implant surgery and recovery&#; health state, which includes patients who undergo a two-stage osseointegration surgery and rehabilitation. Costs include the implant system, diagnostic testing for screening eligibility, professional service fees during hospitalization, inpatient hospital costs (including rehabilitation), prosthetist services, and outpatient care. As this procedure has not been conducted in Ontario and there are no specific billing codes, we estimated the costs using proxies informed by expert opinion (Nancy Dudek, MD; Richard Jenkinson, MD; Dan Mead, CP(c); Amanda Mayo, MD; James Waddell, MD; Wade Gofton, MD; communications, August ). As shown in , the cost of the implant system was derived from a surgical group in Montreal who conducted the first osseointegrated prosthetic implant surgery in Canada (Natalie Habra, MD, communication, June ). After expert consultation, we assumed that postsurgical rehabilitation was provided in an inpatient setting, due to the anticipated need for nursing support with dressing changes and the intensive rehabilitation process (Nancy Dudek, MD, communication, August ). Furthermore, given the specialized nature of the surgery for a low annual volume of patients, we assumed the surgery would be conducted in a select number of specialty hospitals across the province, which would make outpatient rehabilitation inequitable for patients travelling long distances for the procedure. Finally, prosthetist fees could not be directly estimated using the Limb Prostheses (Conventional) Product Manual 53 from the province's Assistive Device Program (ADP), so we used hourly rates for clinical ($183.29/hour) and technical ($126.22/hour) prosthetic services in Ontario (as described in the product manual under &#;modifications not listed&#;). 54 This information, alongside the estimated time for services required by patients with osseointegrated prosthetic implants described in Frossard et al, 55 provided a cost estimate for prosthetic services. The model incorporated only 75% of the total cost of prosthetic services; the ADP requires patients to cover the remaining 25%, and, because the model took the perspective of the Ministry of Health and Long-Term Care, we excluded out-of-pocket expenses.
Due to the comparative nature of this analysis, we used several costing assumptions to simplify the analysis. We did not include the cost of replacing prosthesis components over time, because we assumed patients in both cohorts used the same external components. Since all individuals receiving osseointegrated prosthetic implants were prior socket users, we excluded the cost of a socket prosthesis device, because this cost would be incurred prior to the implant procedure. Costs of screening to determine patients' eligibility for implants and training costs for surgeons were assumed to be out of scope and were not included. We did not evaluate bilateral osseointegrated prosthetic implants compared to bilateral socket prostheses due to a lack of utility data specific to bilateral amputees and the assumption that the results of a single implant would provide a valid approximation of the results for bilateral implants. We did not include minor procedures along the patient pathway (e.g., staple removal at 3 to 4 weeks postsurgery), due to a lack of costing data and to simplify the model. Personal support workers and home care between the two stages of implant surgery were not costed, because clinical experts indicated that home care would be necessary only in the event of rare complications with wound healing that would require long-term dressings, which would also likely result in implant failure (James Waddell, MD; Amanda Mayo, MD; John Murnaghan, MD; Wade Gofton, MD; communications, September to November ). Finally, we did not account for differences in maintenance costs in the reference case analysis, such as the number of visits to a prosthetist as outlined by Haggstrom et al, 52 due to both insufficient patient-level data and data on the average cost per prosthetist visit. However, we did conduct a scenario analysis to test this assumption.
All costs were reported in Canadian dollars. We obtained cost inputs from standard Ontario sources and published literature. The fees for professional visits, procedures and consultations were obtained from the Ontario Schedule of Benefits for Physician Services. Hospitalization costs were obtained from the Ontario Case Costing Initiative (OCCI). Diagnostic and laboratory fees were obtained from the Ontario Schedule of Benefits for Laboratory Services.
Utilities represent a person's preference for certain health outcomes, such as being able to walk. These are often measured on a scale of 0 (death) to 1 (full health). summarizes utility data specific to each health state. All studies evaluating individuals' quality of life before and after two-stage osseointegration used the 36-Item Short Form Health Survey (SF-36). To obtain utilities, the studies converted SF-36 values to the SF-6D, a six-domain version of the survey. In cases where only the mean SF-36 domain scores were available, we obtained the utility value by using equation 1 in an SF-36 to SF-6D map/crosswalk published by Ara et al. 51 Hagberg et al 26 was the only study reporting a direct utility value using the SF-6D, and we therefore chose it for the reference case analysis. Utilities from other sources that used a crosswalk were used in scenario analyses. In the reference case, we conducted a probabilistic analysis and used a beta distribution around the values of the mean and standard error.
We performed a targeted literature search in MEDLINE for utility values on June 11, , for studies published from inception to the search date. We based the search on the clinical search strategy with a methodological filter applied to limit retrieval to health state utility values. 50 A second utilities search was conducted on June 27, , to retrieve studies on leg prosthetics using the same filter. See Appendix 1 for our literature search strategies, including all search terms.
The model assumed equal mortality rates for the two cohorts. Although the broad cohort of conventional socket prosthesis users has a higher mortality rate, this is in part due to vascular comorbidities such as diabetes and heart disease, which would preclude an individual from meeting eligibility criteria for an osseointegrated prosthetic implant (see Appendix 6 , ). However, because most osseointegration procedures are conducted several years postamputation, and our model compared a hypothetical population of the same patients who either remained as socket prosthesis users or converted to osseointegrated prosthetic implants, we assumed the survival rate was comparable for both treatments. Given that the average age of the target population for osseointegrated prosthetic implants is 46 years old and the procedure has strict eligibility criteria, this created a subpopulation of relatively healthy individuals; therefore, we used age- and sex-specific mortality rates from the Ontario general population. 49 Clinical experts verified this assumption because inputs for the alternative method&#;mortality rates or hazard ratios specific to a population with nonvascular amputation&#;were not available in the literature (Richard Jenkinson, MD; Nancy Dudek, MD; Amanda Mayo, MD; John Murnaghan, MD; : communications, August to November ).
contains the rates from multiple sources for complications in both cohorts. Due to a lack of data, we could not derive the risk of fractures for patients with an uncomfortable socket, but we included fractures in the model as they were expected to have a greater cost in the osseointegration cohort compared with the socket cohort. Given this potentially conservative estimate for osseointegrated prosthetic implants, we conducted a sensitivity analysis that excluded fractures from the model in the event the risk of fracture is similar between cohorts and the incremental cost difference for treatment is marginal. Treatment pathways for deep infections involved either IV antibiotics alone, IV antibiotics and debridement (with variations in whether the debridement was conducted as inpatient or day surgery). Unfortunately, the rates found in the literature did not differentiate between these pathways, so the model conservatively assumed that patients received IV antibiotics and debridement in an inpatient setting. We tested this assumption in scenario analyses, where we assumed that IV antibiotics were administered in a home-care setting. Given the lack of data on complications for users of conventional socket prostheses, specifically in a population with nonvascular amputations, only stump revisions were included as a complication for the socket cohort.
summarizes transitions between health states in the model, derived from pooled rates. All transitions not explicitly mentioned in the table but found in (i.e., osseointegrated prosthesis and reimplant) are calculated as the complement of the sum of the other branch probability, which is calculated by subtracting the probability found in from the value 1.00.
In our economic evaluation, we considered the impact of costs and quality of life associated with both treatments. We included adverse events that are severe, expensive to treat, or have a large impact on patients' health-related quality of life (e.g., infection, soft-tissue refashioning, implant extraction, fracture, stump revision). We excluded adverse events that have a negligible impact on health effects or resources (e.g., skin rash, blisters, cysts). Other complications considered for conventional socket prostheses included pressure ulcers, neuromas, fungal infections, and mechanical limb pain, but we could not find incidence rates for these complications. Nevertheless, we assumed these chronic conditions were represented by the published utility value for users of an uncomfortable socket prosthesis. Despite the lack of incidence rates, the complication rate may be similar between the socket and osseointegration cohorts, as it has been stated that in osseointegration patients, &#;infection and irritation of the soft tissue in the skin penetration area are common during the first 2 years.&#; 27 Regardless, if the previously published stump revision rates are not inclusive to pressure ulcers and neuromas, the economic model may be underestimating the total cost of the socket prosthesis cohort.
Expert consultants advised that the &#;implant extraction&#; health state could be treated similarly to joint arthroplasty infections: the infected prosthesis is removed during an initial debridement, a temporary prosthesis is fitted, and patients are treated with intravenous (IV) antibiotics and then considered for reimplantation (Richard Jenkinson, MD; Amanda Mayo, MD; John Murnaghan, MD; Nancy Dudek, MD; Wade Gofton, MD; communications, August to November ). This process, although specific to an extraction due to infection, was estimated to take 6 to 12 months. Although other causes have been identified for reimplantation, our model assumed this process would take 12 months (i.e., the model's cycle length). This assumption provided a conservative estimate of the QALY gain in the osseointegration cohort, penalizing implant extractions.
The osseointegration cohort began in the &#;implant surgery and recovery&#; health state, where they underwent surgery and rehabilitation. Patients could then transition to either &#;osseointegrated prosthesis&#; or &#;implant extraction.&#; In the &#;osseointegrated prosthesis&#; health state, patients had full use of their prosthesis but were still at risk of transitioning to &#;implant extraction.&#; Anytime they were in the &#;implant surgery and recovery&#; and &#;osseointegrated prosthesis&#; health states, patients were at risk for complications specific to osseointegration, which included soft-tissue problems, fractures, superficial infections, and deep infections. These complications had distinct costs assigned to their occurrence. We assumed that patients in the &#;implant extraction&#; health state would not be susceptible to implant-related complications, given that the implant was extracted (no complications were identified in the literature specific to implant extraction). In the &#;implant extraction&#; health state, the implant failed and was removed. Some patients then transitioned to the &#;reimplant&#; health state to have the implant reinserted, but a proportion who were no longer suitable for osseointegration permanently returned to their original socket prosthesis and remained in the &#;uncomfortable socket prosthesis&#; health state. In that state, as previously mentioned, patients were at risk for stump revisions, a complication specific to socket prosthesis users.
We modified a previously published Markov decision-analytic model structure from Hansson et al 45 to estimate the long-term clinical and economic outcomes of osseointegrated prosthetic implants for lower-limb amputees. The cycle length was 1 year, because patients are usually monitored annually by a physiatrist (Nancy Dudek, MD; Amanda Mayo, MD; John Murnaghan, MD; communications, August to November ). We applied a half-cycle correction on all health-state transitions. The model was built using TreeAge Pro 48 and repeated cycling until the time horizon was reached.
We used a lifelong time horizon in the reference case analysis, given the chronic nature of the health condition and intervention, thus capturing all health effects and costs relevant to the decision problem. We also conducted sensitivity analyses around the time horizon, including a 10-year time horizon, which approximates the longest average follow-up recorded in an observational study on osseointegrated prosthetic implants. 30 Additionally, we used time horizons from other cost&#;utility analyses of 6 years and 20 years in sensitivity analyses. 44 , 45
To further refine the scope of this analysis, we also excluded certain variations of osseointegrated prosthetic implants. For example, we did not cost implants combined with total hip replacement and total knee replacement, due to a lack of available data. Furthermore, this analysis excluded customized implants not commercially available, also due to a lack of available data. summarizes the interventions evaluated in the economic model.
For this economic evaluation, we evaluated the cost-effectiveness of both two-stage and single-stage osseointegration surgery. However, given the data available, we compared the two-stage procedure to conventional socket prostheses in the reference case, and evaluated the single-stage procedure in a scenario analysis. Despite the absence of published literature on single-stage surgery, it appears to be a technological innovation, as the first osseointegration surgery conducted in Canada was a single-stage surgery performed in Montreal in March . In modeling single-stage surgery in a scenario analysis, given the absence of published evidence, we assumed complication rates and effectiveness equal to the two-stage procedure; therefore, we pooled two-stage data and used them as estimates for the single-stage model inputs.
The company is the world’s best lower limb prosthesis for sale supplier. We are your one-stop shop for all needs. Our staff are highly-specialized and will help you find the product you need.
As described in the Background section of this report, unlike a conventional leg prosthesis that uses a specially fitted cup-like shell (socket) that fits over the remaining portion of the residual leg, osseointegrated prosthetic implants are inserted into an amputee's remaining bone to connect the bone to an external prosthetic limb. The implant systems include an intramedullary component that integrates with the bone, and a bridging connector (also called an abutment) that connects from one side to the intramedullary implant and from the other side to the prosthetic limb. Traditionally, the treatment requires two operations: the first inserts the intramedullary implant, and the second, performed 6 to 9 months later, creates a percutaneous skin opening allowing for abutment attachment and prosthesis fitting. 7 However, the duration between surgeries has been reduced in more recent publications, with the overall time between surgeries being approximately 6 to 8 weeks. 32 Recently, a group in Australia has published a protocol describing a single-stage surgery for osseointegration that they have been conducting since April . 7 Despite this development, no published papers are available on the effectiveness of single-stage osseointegration, outside of a protocol from the OGAAP-2 (Osseointegration Group of Australia Accelerated Protocol-2) cohort study. 7
The model's population characteristics were based on sample-size weighted averages from observational studies used to inform the clinical and state-transition parameters of the model ( Appendix 6 , ). 24 The population was 46 years old on average at the time of receiving their osseointegrated prosthetic implant and consisted of 70% males and 30% females. Clinical experts validated these characteristics as representative of the Canadian population (Nancy Dudek, MD; Richard Jenkinson, MD; Dan Mead, CP(c); Amanda Mayo, MD; John Murnaghan, MD; communications, August to November ). We also assumed everyone in the model had a unilateral above-the-knee amputation, which represented the majority of osseointegrated prosthetic implants and allowed us to standardize and simplify costing.
This model evaluated a population of individuals with a lower-limb amputation, not due to diabetes or severe vascular disease, with a medical history of issues related to the use of a conventional socket prosthesis that resulted in an uncomfortable prosthesis fit and difficulty walking. Additionally, patients had to meet distinct eligibility criteria specific to receiving the osseointegrated prosthetic implant, such as having reached full skeletal maturity, not currently undergoing chemotherapy, and not taking corticosteroid or immunosuppressant drugs. For the full list of clinical requirements, see Appendix 6 , .
We conducted a cost&#;utility analysis to determine the costs and health outcomes (i.e., quality-adjusted life-years [QALYs]) associated with each treatment strategy. We chose this type of analysis because utility inputs are available and a generic outcome measure such as QALYs allows decision-makers to make comparisons across different conditions and interventions. The outcomes reported are total costs and total QALYs for each treatment, and incremental cost per QALY gained compared to the next most effective strategy. For this analysis, incremental costs and QALYs are key outcomes considered by decision-makers, while total costs and QALYs of treatment options are informative measures for decision-makers.
We conducted a reference case analysis and sensitivity analyses. Our reference case analysis adhered to the Canadian Agency for Drugs and Technologies in Health (CADTH) guidelines when appropriate and represents the analysis with the most likely set of input parameters and model assumptions. 47 Our sensitivity analyses explored how the results are affected by varying input parameters and model assumptions.
presents the results of all scenario analyses, previously described (see ). They show a wide range of ICER values based on variations in model inputs and model assumptions. At a willingness-to-pay value of $100,000 per QALY, the probability of osseointegrated prosthetic implants being cost-effective in the reasonable best scenario was 92.2% and in the reasonable worst scenario was 35.1%.
presents the results of the one-way sensitivity analyses through a tornado diagram. The ICER was highly sensitive to five variables: the utilities for osseointegrated prosthesis and uncomfortable socket prosthesis, time horizon, implant extraction rate, and stump revision rate. When the utility input for the &#;osseointegrated prosthesis&#; health state (reference case = 0.692) was set at 0.726 (high estimate), the ICER dropped to $51,620 per QALY gained, but rose to $653,555 per QALY gained if the utility input was 0.658 (low estimate). When the model time horizon dropped below a lifetime horizon to only 5 years, the ICER rose to a peak of $265,581 per QALY gained. As the probability of implant extraction rose to 0., the ICER increased to $154,899 per QALY gained, but if the implant extraction rate dropped to 0. the resulting ICER was $78,062 per QALY gained. Finally, if the stump revision rate in socket prosthesis users increased to 0.186 (a rate similar to the soft-tissue refashioning in the osseointegration cohort) the ICER dropped to $57,076 per QALY gained.
The cost-effectiveness acceptability curve presented in shows the probability of both interventions (osseointegrated prosthetic implants and uncomfortable socket prostheses) being cost-effective across a range of willingness-to-pay values. At a willingness-to-pay value of $100,000 per QALY, the probability of osseointegrated prosthetic implants being cost-effective was 54.17%. As the willingness-to-pay value crossed $40,000 per QALY and continued to rise, the probability of implants (the more costly strategy) being cost-effective also rose. Appendix 6 , , presents an incremental cost-effectiveness scatterplot for the reference case results.
presents results from the reference case analysis. Over a lifetime horizon, the osseointegration cohort had an average total cost of $101,166 and 19.12 QALYs per person. Compared with an uncomfortable socket prosthesis, an osseointegrated prosthetic implant has an incremental cost of $84,559 and an incremental effect of 0.890 QALYs. The reference case ICER for an osseointegrated prosthetic implant compared with an uncomfortable socket prosthesis is $94,987 per QALY gained.
Discussion
The results of the analysis indicated that, in eligible patients, an osseointegrated prosthetic implant is more costly and more effective than continuing to use an uncomfortable socket prosthesis. The reference case analysis showed osseointegrated prosthetic implants, when compared to an uncomfortable socket prosthesis, had an ICER of $94,987 per QALY gained, which corresponds to a 54.2% probability of being cost-effective at a willingness-to-pay value of $100,000 per QALY. As willingness-to-pay values increased, the probability of implants being cost-effective increased to 74.2% at $150,000 per QALY.
However, sensitivity analyses indicated these results were largely influenced by both parameter uncertainty and key assumptions. One input to note is the probability of stump revisions in the uncomfortable socket cohort. This was the only complication evaluated for the socket cohort in our model, and given a lack of evidence in the published literature, we used a probability estimate taken from hospital data from a previous cost&#;utility analysis for our reference case.45 This probability estimate was the lowest complication in the model (0.026) and was slightly less common than infrequent complications in the osseointegration cohort such as fractures (0.027) and deep infections (0.033). The reference case estimate may be conservative, given that stump revisions often treat a wide variety of conditions related to an uncomfortable socket prosthesis, such as bone spikes, soft tissue&#;socket interface problems, neuromas, and infections.58 A one-way sensitivity tested this by increasing the probability to the same value as osseointegration's soft tissue refashioning (0.18), and the ICER decreased to $45,519 per QALY gained.
Osseointegrated prosthetic implants are a novel technology in Ontario, and estimating costs required that we use proxy values, validated by clinical experts. Owing to the uncertainty this creates, we designed the model's reference case to conservatively estimate the cost-effectiveness of osseointegration. For example, the reference case did not include the cost of prosthetist visits for prosthesis maintenance and adjustment, even though these costs appear to be a key driver of cost-effectiveness in osseointegrated prosthetic implants. This decision was due to a lack of data to inform the cost differences between cohorts. Ontario has not developed fee codes for prosthetist care for amputees using an osseointegrated prosthetic implant; instead, we calculated these fees using an hourly rate for clinical and technical prosthetist services, which is typically used for cases where there are modifications not listed; for billing, these services require approval from the ministry's Assistive Devices Program. Although a previous publication found that implant users visit a prosthetist less frequently than conventional socket users (7 vs. 3 visits per year), the cost of each visit was difficult to estimate as it is specific to the purpose of the visit.52 Based on expert feedback, we expected the average cost of a prosthetist appointment would be lower for implant users because the soft tissue&#;socket interface is eliminated (Dan Mead, CP(c), communication, August ). To estimate costs, we assumed each adjustment would take 1 hour for an implant user and 2 hours for a socket user (Nancy Dudek, MD, communication, August ). Despite published data on the number of maintenance visits, we excluded prosthetic maintenance fees from the reference case due to the uncertainty around the true cost difference between cohorts.
We used scenario analyses to test various model assumptions. Creating a &#;reasonable best scenario&#; and &#;reasonable worst scenario&#; helps to better understand how the model assumptions and variability of the inputs impact cost-effectiveness when combined. Both best and worst reasonable scenarios excluded changes to the time horizon, discount rate, and utilities, and instead focused on other &#;reasonable&#; inputs. These analyses excluded alternative utility values from Branemark et al10 and Hagberg et al13 because those studies mapped health-related quality of life scores to derive utilities, specifically from mean SF-36 domain scores to SF-6D utility values. This technique, although useful if there are data limitations, is less precise and accurate than utility values directly elicited from a SF-6D questionnaire.
The ICER obtained from the reasonable best scenario was heavily influenced by the inclusion of prosthetist maintenance costs and time-varying complication rates. Unlike in the reference case, the reasonable best scenario used time-varying rates from a single study, instead of an average weighted rate from multiple studies. In that single study, no cases of soft tissue refashioning occurred, and there were no cases of deep infection after the first year. When modelling the time-varying complication rates, we used the event rate in the second year (zero) for all subsequent years. As there would be no cases of deep infection after the first year, compared to the reference case this approach would underestimate the costs of deep infections over a lifetime horizon. Based on these limitations, we did not use time-varying complication rates in the reference case. However, the single study informing these rates does suggest that rates of superficial and deep infection are lower in the second year after osseointegration surgery.10 If complication rates do decrease after the first year, our reference case model with its constant rates may overestimate the occurrence and costs of complications over a patient's lifetime. Other inputs used in the reasonable best scenario include a reduction in the device cost for osseointegrated prosthetic implants, as the cost in the reference case was based on a cost reported by a surgeon performing a single surgery. As previously discussed, another input, single-stage vs. two-stage osseointegration surgery, is a newer development that could reduce operating room costs but as yet has no published evidence to inform its effectiveness and complication rates.
The reasonable worst scenario included two key inputs that were excluded from the reference case due to insufficient evidence to properly inform their estimates. The first, the inclusion of mechanical part replacement, was derived from studies that did not specify the implant parts that required replacement, so we conservatively assumed that all external parts were replaced. The second, the inclusion of disutility values for complications (i.e., fractures and deep infections), was not specific to lower-limb amputees and was based on values from instruments other than the model's SF-6D values (i.e., EuroQol 5-Dimension [EQ-5D] and Assessment of Quality of Life [AQoL]). Further research is needed to refine these important model inputs, which influence the cost-effectiveness of osseointegrated prosthetic implants for lower-limb amputees.
Model Considerations
In our economic evaluation, we assessed osseointegrated prosthetic implants as a therapeutic class to inform recommendations about public funding the intervention. We did not differentiate between manufacturers, primarily due to a lack of manufacturer-specific data that could inform all model inputs. However, it should be noted that the majority of publications used to inform this analysis were from the OPRA and ILP systems. There is uncertainty around whether the longevity of the implant differs among the systems currently available and around whether implant design leads to different challenges in the event of surgical removal and reimplantation.
The model assumed that rehabilitation of patients undergoing osseointegration surgery took place in an inpatient setting. Despite costing more than care in an outpatient setting, an inpatient rehabilitation stay likely best represents both the clinical and Ontario health care context. A lengthy, intensive rehabilitation process is required as patients progress through a gradual, stepped approach to weight-bearing on the prosthetic limb.29 Clinical experts indicated that, if the program is implemented, it should be restricted to a select number of specialty health centres with interdisciplinary teams of surgeons, physiatrists, nurses, and physiotherapists. This approach would promote optimal surgical performance by ensuring surgeons receive an adequate annual volume of osseointegration surgeries (Wade Gofton, MD, communication, August ). Given this advice, we assumed that patients from across the province would travel to these centres to receive care and that a lengthy outpatient rehabilitation would be a burden to all but those living nearby. Inpatient rehabilitation was also anticipated so that wound care and medications could be actively managed by nursing staff after surgery and into early rehabilitation (Nancy Dudek, MD, communication, August ).
To simplify economic costing, this model assumed external components were similar for socket and osseointegrated prostheses. Therefore, incremental costs and utility gains attributed to these components were excluded from the analysis. If future data become available, this assumption should be tested further, because another analysis by Haggstrom et al52 suggested that osseointegrated prostheses may be more costly due to the use of more advanced prosthetic components (e.g., microprocessor knees).
We did not use a societal perspective in this analysis due to insufficient data on the target population. However, osseointegrated prosthetic implants may provide unique benefits, such as allowing a person to go back to work and no longer require disability support because their mobility has improved, which may lead to financial savings from a societal perspective.
Comparison With Other Studies
Our study had distinct differences from two previous cost-effectiveness analyses. The first, by Frossard et al,44 found an ICER of $16,632 per QALY gained, but the study only included prosthetic costs from a dataset at a single prosthetic facility. Specifically, it did not include costs of surgery, rehabilitation, and complications. The authors used utility values similar to those in our reference case, but the overall QALY calculation was simplified as they used neither a Markov model nor other forms of decision-modelling. As the study pulled costs from their real-world dataset, the time horizon was only 6 years to reflect the length of follow-up in the dataset.
Hansson et al45 found an ICER of &#;83,374 per QALY gained, and based their complications and health state transitions primarily on a single study by Branemark et al.10 This ICER was higher than what was reported in our reference case, but this difference is likely due to the 20-year time horizon, as our scenario with that time horizon found a similar ICER of $123,112 per QALY gained. The authors stated that they chose this time horizon for their reference case as the technology is relatively new and they are uncertain how long patients will benefit from it. However, they do report being aware of 8 patients from previous studies who have passed 15 years of follow-up.45 The cost-effectiveness analysis by Hansson et al45 was similar to our work in that their model was sensitive to changes in the interventions' utility values, and we both used similar utility values and clinical complications. Of note, both Frossard et al44 and Hansson et al45 differed from our analysis in that they did not use discounting, with one stating that costs were not discounted because the majority of costs occurred in the first and third years, when prosthetic knees and feet were supplied.44
Strengths and Limitations
Our primary economic evaluation has several strengths. It is the first analysis to estimate the economic value in Canada of osseointegrated prosthetic implants for eligible lower-limb amputees. The study used Ontario-specific costs wherever possible, and Canadian costs for the implant system. Our analysis is also the first to include disutilities for complications related to the osseointegration surgery and implant (deep infections and fractures), as well as other analyses such as time-varying rates to inform event transitions. Another strength is the use of pooled rates for health-state transitions and complications to capture parameter variability across the published literature. Additionally, cost estimates were informed by a multidisciplinary team of medical professionals involved in the care pathway, including surgeons, physiatrists, physiotherapists, and prosthetists.
Our analysis also has limitations. As the osseointegration procedure is not currently being conducted in Ontario, the Schedule of Benefits has no physician fee codes to inform precise estimates of surgical costs and inpatient stays. In the absence of costing data, we used proxies validated by clinical experts. We also used simplifying assumptions in the absence of established clinical practices in Ontario for patients with osseointegrated prosthetic implants. There are potential areas for additional costs that we did not evaluate, such as additional prosthetist adjustments after femoral fractures in implant users or prosthetic adjustments to alter socket fit following stump revisions in conventional prosthesis users. As described earlier, the data on users of uncomfortable socket prostheses are limited. As a result, we could not cost additional issues identified by experts: scar revisions, wound care, superficial infections, material and suspension system costs for prosthetist socket adjustment and replacement, and (if we included a societal perspective) out-of-pocket costs for antibiotics, antifungals, and dressing supplies (Nancy Dudek, MD; Wade Gofton, MD; communications, November ). Another potential limitation is the applicability of the literature to our comparator: patients with an uncomfortable socket prosthesis. Previously published literature did not always specify the comparator in this way, so these studies may include patients with a comfortable socket fit and a preference to switch to an osseointegrated prosthetic implant. Therefore, given the pre-post study design commonly used in this literature, the rates of complications in our uncomfortable socket cohort may have been higher if we were able to strictly enforce this population in our model inputs.
Comments
Please Join Us to post.
0