We used a Markov decision model to evaluate the clinical and economic consequences of MoM HRA compared to THA. The population studied was men and women aged 50 years or older undergoing MoM HRA or THA for advanced OA of the hip. A 30-year time horizon was used to evaluate the incremental clinical effectiveness (in terms of quality adjusted life-years (QALYs) gained) and cost-effectiveness (cost per QALYs gained) of MoM HRA and THA. The incremental cost-effectiveness of MoM HRA versus THA was examined from a healthcare system perspective (focusing on health care costs and patient quality of life) using hospital and professional reimbursement to estimate costs and quality adjusted life years to estimate effectiveness.
The decision tree (Fig. A–B) begins with the decision to choose either MoM HRA or THA for patients with advanced OA of the hip. Each alternative is represented as a Markov model with mutually exclusive states. Patients transition between states (or remain in a state) over time. The model used intervals (Markov cycles) of 1-year duration. While in each state during each yearly interval, patients experience a quality of life (QoL) and incur direct medical costs; in addition, transitions associated with revision surgery (conversion from HRA to THA, major total revision THA [revision of both the acetabular and femoral components], major partial revision THA [revision of the acetabular component only], or minor revision THA [exchange of the modular acetabular liner and femoral head only]) are associated with a short-term transitional decrement in QoL (or disutility) and an increase in direct medical costs associated with revision surgery. The probability of transition between states depends on the patients’ age, gender, and type of procedure (MoM HRA or THA). For MoM HRA, the health states are year of initial MoM HRA, post-HRA, post-conversion from HRA to THA, post-major total revision THA, post-major partial revision THA, post-minor revision THA, death due to any HRA or THA surgery, and death due to other causes. Thus, the MoM HRA cohort may experience an initial failure (requiring conversion from HRA to THA) or a subsequent failure requiring revision after THA. For the THA cohort, the disease states are year of initial THA, post-THA, post-major total revision THA, post-major partial revision THA, post-minor revision THA, post-second major total revision THA, post-second major partial revision THA, post-second minor revision THA, death due to any THA surgery, and death due to other causes. Decision analysis software (TreeAge Pro 2008, Williamstown, MA) was used to create a Markov decision model.
Fig. 1A–B (A) A Markov decision tree compares the clinical outcomes for MoM HRA and THA patients. MoM THA and primary THA are represented as Markov nodes (“M”). The branches are the Markov states. Conversion from HRA to THA is analogous to first (more ...)
Information on implant survivorship was sought from large national or multicenter registries with implant survival data of sufficient duration to estimate annual age, gender, and procedure specific probability of implant survival and implant failure. Additionally, because the probability of implant failure varies by year of followup, we sought data of sufficient duration (5 or more years after initial surgery). The Australian Orthopedic Association (AOA) National Joint Replacement Registry Hip and Knee Arthroplasty Annual Report [4
] provides gender and decade of age stratified cumulative percent revision for 5 to 7 years of followup for 9956 patients who received HRA for primary diagnosis of OA (excluding infection) and 109,972 patients who received primary conventional THA for a primary diagnosis of OA. The report also summarizes the type of revision (major total revision, major partial revision, and minor revision) and probability of subsequent revision for 2616 revision THA procedures. Annual probability of revision of MoM HRA and THA was estimated from the summary gender- and age-stratified data in the AOA National Joint Replacement Registry 2008 report by fitting a general failure time model (Weibull distribution), which allowed for time varying hazard [22
] of failure for each gender and age stratum for 5 years of followup (the longest followup interval for which data was available for all strata). As indicated in the AOA registry report, the probability of failure in each stratum was highest in the first year of followup and declined thereafter, resulting in cumulative revision curves that increased with time but at lower rates after the initial year. After year 5, the annual probability of failure was assumed to remain constant. The analyses of MoM HRA and THA are summarized separately for six gender and age strata to correspond with the available data on failure rates in the AOA registry report. For both genders, the analysis used a patient age of 50 years representing the age younger than 55 years stratum, a patient age of 60 years representing the ages 55 to 64 years stratum, and a patient age of 70 years representing the age 65 to 74 years stratum. The probabilities of perioperative mortality for MoM HRA, THA, and all revision THAs were derived from the literature [15
]. Annual gender- and age-specific all-cause mortality rates were based on United States life tables [3
The effectiveness of each surgical procedure was based on the quality-adjusted life-years associated with each procedure. This measure assigns a QoL weight to each year of followup. The QoL values range from 0 (death) to 1 (perfect health) and reflect the average QoL associated with that health state. The QoL weights for patients with advanced OA of the hip and patients with successful primary THA were obtained from the literature [13
]. The QoL values for patients with successful MoM HRA, conversion from HRA to THA, and revision THA were derived from literature comparing each of these health states to patients with primary THA [5
]. Perioperative morbidity and recovery were captured by applying a lower QoL for a defined period of time after each surgical intervention (longer for revision than primary procedures). QoL weights that are measured by methods that reflect patient preferences for a health state are described as utilities in the health economics literature.
Costs incorporated into the model included both hospital and professional fees for primary THA, revision THA, MoM HRA, and conversion from HRA to THA. Hospital costs were based on average Medicare payments for diagnosis-related groups 544 (primary lower extremity arthroplasty procedures) and 545 (revision lower extremity arthroplasty procedures) for fiscal year 2008. Similar to previously published cost-effectiveness analyses [40
], Medicare reimbursement was chosen (even though the patient population being studied included men and women older than 50 years) since it more closely reflects the actual costs [24
] associated with HRA and THA procedures, as opposed to private payer reimbursement, which is based on a negotiated rate, which often exceeds the true costs of the procedure. Revision THA procedure costs were further delineated by procedure complexity, such as isolated femoral component revision, acetabular component revision, both component revision, or femoral head and liner exchange only, based on previously published data [9
]. Device costs for primary THA, revision THA, and HRA were obtained from published sources [38
]. Costs associated with ambulatory visits and radiographs were also included in the analysis, based on average professional fees for evaluation and management services and both professional and technical fees for hip radiographs [12
The clinical course for patients who receive MoM HRA was compared to the clinical course for patients who receive THA by comparing the cumulative discounted total quality-adjusted years of life (QALYs) and cumulative costs of MoM HRA with the cumulative discounted total QALYs and cumulative costs of THA. The measures used in the comparison were the incremental QALYs (a measure of effectiveness), the incremental costs, and the incremental cost-effectiveness ratio (ICER), which is the ratio of the incremental costs to the increment effectiveness. In accordance with the recommendations of the Panel on Cost-Effectiveness [21
], we discounted all costs and utilities and report reference case estimates for a discount rate of 5%. The base case estimates for the probabilities, utilities, and costs were derived from the literature (Table ).
Variables used in cost-effectiveness analysis and ranges for sensitivity analyses
One-way sensitivity analyses were performed for each of the independent variables (Table ). In these analyses, each variable was varied from 50% to 200% of the point estimate (Table ), per decision analysis modeling convention, and the impact of each variable on the ICER was calculated. One-way sensitivity analyses for selected variables (discount rate, difference in utility [QoL] after conversion from HRA to THA compared to primary THA, and incremental cost of HRA compared to THA) were calculated for each gender and age stratum. One-way sensitivity analyses were used to identify thresholds for selected independent variables where MoM HRA would be cost-saving compared to THA and thresholds where MoM HRA would be considered cost-effective based on an ICER of $50,000 per QALY. Two-way sensitivity analyses were performed to identify ranges for the incremental cost of HRA compared to THA and difference in utility (QoL) after conversion from HRA to THA compared to primary THA where MoM HRA or THA was optimal based on net monetary benefits [49
], using a willingness to pay threshold of $50,000 per QALY gained.
A probabilistic sensitivity analysis (Monte-Carlo sensitivity analysis) was performed to evaluate the combined impact of the individual independent variables jointly on the incremental costs, incremental QALYs gained, and the ICERs. In this analysis, each variable was represented as a probability distribution (Table ) and a random sample for each variable was drawn from its probability distribution and entered into the model. The incremental costs, incremental QALYs gained, and the ICERs and their 95% confidence intervals were calculated from a Monte Carlo simulation using 10,000 samples for each gender and age stratum. An acceptability curve showing the proportion of samples for each gender and age stratum that were below a given willingness to pay threshold was calculated and graphed for a willingness to pay range of $0 to $100,000.
Variables and distributions for probabilistic sensitivity analysis