Cost-utility analysis is a mature health economics methodology that is one part of a greater universe of comparative effectiveness research. Concerns over the rising cost of medical care and the increasing sense of societal limits on those costs have caused a dramatic rise in the number of studies such as this investigation by Schousboe and Brown. The authors use cost-utility analysis and complex modeling to examine one of the enduring controversies in total joint arthroplasty: postoperative prophylaxis against venous thromboembolic events. They chose to compare aspirin (ASA or acetylsalicylic acid) and low-molecular-weight heparin (LMWH); this choice is acutely more relevant at a time of a somewhat restrained convergence of recent guideline recommendations from the American Academy of Orthopaedic Surgeons (AAOS)1 and the American College of Chest Physicians (ACCP)2, agreeing that either agent might be appropriate. The decisive measure that is estimated through their modeling is the cost of an added quality-adjusted life-year (QALY) relative to an estimated societally acceptable cost-effectiveness threshold of $100,000.
Ideally, a cost-utility analysis measures the marginal gains and/or losses of QALYs through real-world patient-reported outcomes using validated measures of general health. Patient-reported outcome measurements would be difficult for this topic, given the fortunate relative rarity of the important adverse outcomes being analyzed, such as death, pulmonary embolism (PE), hemorrhage, and postphlebitic syndrome. The authors resorted to creating a model using multiple external reports of the relative risk and gain or loss of utility for the various health states relative to the use of aspirin compared with LMWH. Their online supplement provides a reasonable justification for each step; the literature chosen provides a slight advantage to LMWH for efficacy and to aspirin for less bleeding. Creating a model of this complexity requires some a priori leaning in one direction or another to make it a better approximation of real-world observations.
The results of the modeling might surprise some readers. It would be easy to assume that aspirin would dominate in a cost-utility analysis, given the overall low incidence of clinically important venous thromboembolic (VTE) events as well as the magnitude of difference in cost between aspirin and LMWH. The study readily reached that conclusion for total hip surgery, with the cost per QALY gained using LMWH being substantially higher than the given cost-effective threshold of $100,000. Given the initial parameters, however, this was not true in knee arthroplasty for patients under the age of eighty years; the marginal cost per QALY gained with LMWH fell below the threshold, making the choice of agent less clear.
It should be kept in mind that cost-utility analysis is a health policy tool usually utilized at the level of national determinations regarding health-care spending, but not used as a guide in clinical decision making. The levels of the cost-effectiveness thresholds vary from one country to another as well as within; in the United Kingdom, the value ranges from £20,000 to £30,000 ($30,600 to $45,930 in 2013 U.S. dollars)3, lower than the given $100,000 in this study. The use of such analyses to determine policy is not without controversy. One example in this paper is age bias; the older patient’s shorter life expectancy causes results from the model that discount complications or death in a way that would not be so readily discounted by the individual patient or his or her family.
In addition to the “big picture” perspective of policy makers, there are the “small picture” perspectives of the individual patients, providers, and hospitals. Total joint arthroplasty has been demonstrated to be very cost-effective from a societal perspective; some values given include $10,402 per QALY for total hip arthroplasty and a range of $11,548 to $21,787 per QALY for total knee arthroplasty4. These are elective procedures costing between $20,000 and $30,000 that have associated high expectations of uneventful outcomes from patients and payers. Although the health utility of the patient might be eventually restored after a VTE event, the emotional impact of a thrombotic event is less well measured. The odds ratio of the estimated increased risk of death in patients with a VTE event compared with those without a VTE event has been reported to range from 2.95 for total hip arthroplasty to 5.0 for total knee arthroplasty5; the sense of this risk hangs over the patient and the provider alike. Because of the high utilization rates of total joint surgery, the procedures have come under increasing pressure; complications now lead to adverse public reporting of outcomes, withholding of payments for readmissions or certain complications, and potential reduction of value-based purchasing payments due to below average outcomes. The advent of bundled payments also creates a new calculus of risk.
Another way of assessing the cost-effectiveness of LMWH compared with ASA is the cost of preventing each additional VTE event, which has been reported to range from $1300 to $72006. Given the overall cost of a total joint admission, the newly increased cost-risk for complications and/or readmissions, and the impact of complications in such elective procedures, it is possible that the given added marginal cost in this paper of $730 per case incurred from using LMWH is more cost-effective at a microeconomic level than it might appear at the macroeconomic level of a cost-utility analysis assessment.
It is difficult to make such an assertion with complete confidence at this time, especially regarding symptomatic deep-vein thrombosis (DVT), PE, or death. The majority of high-level studies in this field used ascending venography to demonstrate DVT, symptomatic or not, as the so-called gold-standard end point. Over the last decade, there has been coalescing agreement that DVT per se is not an effective surrogate for risk for PE, symptomatic DVT, and/or death. Because of the low incidence of such events as death or PE, it has been estimated that an appropriately powered, randomized controlled trial would require 36,000 patients to show a difference between ASA and LMWH, a prohibitively large number7. In addition, the existing Level-I evidence in this literature consists of patients who have passed through the filter of substantial exclusion criteria; in a more standard population of arthroplasty patients, the constellation of potential risks for VTE or complications is more complex.
Registries might be the key to more definitive recommendations in the near future, especially with the decreasing marginal cost of increasingly sophisticated data management. In two recent papers, Jameson et al. linked data from the National Joint Registry for England and Wales with that of the English National Health Service to compare outcomes with LMWH and those with ASA; there was a weak advantage for LMWH in terms of mortality after total hip arthroplasty, no significant difference after total knee arthroplasty, and a surprising higher rate of return to the operating room for patients who had ASA after total knee arthroplasty7,8. As with any national payer data system, questions arise as to the veracity of the data that are uploaded. On a much smaller scale, but with great attention to the accuracy of the data, the Intermountain Joint Replacement Center reported on the risk-adjusted use of ASA compared with LMWH; they found that, in standard-risk patients, the rate of PE was 4.6% in the ASA group compared with 0.7% in the LMWH group, and the rate of DVT was 7.9% and 1.2%, respectively9. There are some caveats to the report, such as the participation of only one surgeon in the ASA arm of the study, the early cessation of the study, and the nonrandomization of the patients. Nonetheless, their results in a real-world setting are another indication of the need for further study.
Over the next several years, it is realistic to expect data-rich registry results from an increasing number of large health-care systems at the same time as national payers, such as Medicare, develop better collection and risk adjustment tools. The choice of VTE prophylaxis and meaningful outcomes are expected to become a standard part of such registry and administrative data. At that time, decisions regarding overall policy, practice management, and, most importantly, best care of our patients might be better discerned.