Rheumatoid arthritis (RA) is a systemic autoimmune disorder that causes chronic inflammation of the joints as well as other organs in the body. Uncontrolled, RA is typically a progressive illness characterized by joint swelling, pain, and joint destruction, which are associated with functional disability and premature mortality.1-10 RA affects 0.5% to 1% of the US population, and the direct costs (eg, medication, office visits, surgeries) associated with the treatment of RA have been estimated at $8.4 billion annually in the United States.8

Treatment options for RA include analgesics, nonsteroidal anti-inflammatory drugs (NSAIDs), glucocorticoids, traditional disease-modifying antirheumatic drugs (DMARDs), and biologic DMARDs. The use of traditional DMARDs in RA, for example methotrexate (MTX), has been shown to ease pain, improve functioning, and help improve long-term outcomes such as reduced radiographic progression and disability for some treated patients.11-14 Early diagnosis and treatment are considered essential for effective disease management of RA.15-18

Recent advances in the understanding of the pathogenesis of RA and a recognition of the incomplete therapeutic outcomes associated with conventional therapy have led to a shift in treatment strategy, with clinicians preferentially using aggressive interventions early in the course of the disease. Rheumatologists are increasingly treating RA with biologic agents—protein-based therapies that target specific components of the disregulated immune response characteristic of RA. Despite their incremental efficacy, the higher acquisition cost of biologic agents has raised important health policy questions.19,20 Moreover, their efficacy, particularly tumor necrosis factor (TNF) inhibitors, has been well established in controlling signs and symptoms of disease, improving functional outcome, and preventing joint damage17,21; but gaps remain in the data regarding their role in reducing the long-term burden of illness associated with RA.17 Biologics are often prescribed when patients exhibit suboptimal response or intolerance to traditional DMARDs such as MTX, and are typically used in combination with MTX. The current study focuses on the first 3 TNF inhibitors introduced to the clinic: etanercept (Enbrel, approved by the FDA in 1998), infliximab (Remicade, approved by the FDA in 1999), and adalimumab (Humira, approved by the FDA in 2002). In addition to their efficacy, it has been suggested that these agents are associated with economic benefits such as fewer physician visits and less use of expensive procedures within 2 to 3 years of treatment initiation.21,22

In light of recent shifts in RA management strategy, the growing use of biologics for disease control, and the current environment of health policy reform, this study sought to analyze the change in the burden of RA by examining trends in the most prevalent procedures delivered to patients with a primary and secondary diagnosis of RA. The 2 most prevalent procedures that meet these criteria are total knee replacement (TKR) and total hip replacement (THR) surgery.

Several hundred thousand total joint replacements are performed worldwide annually.18 As the prevalence of arthritis, aging, and obesity increases, it is anticipated that there will be further growth in the demand for TKR and THR procedures, especially in the population of patients aged less than 65 years.23

Joint destruction due to RA is one of the leading causes for TKR and THR. The knee joint is affected in approximately 80% of patients with RA, with severity and length of disease impacting a patient’s functionality.24 TKR and THR can greatly improve mobility24 and quality of life25 in RA patients, but outcomes are often less successful than those in osteoarthritis patients.24,25 In a prospective population-based study, Schrama and colleagues26 found that the risk of revision of TKR was 1.6 times higher for RA than osteoarthritis.

Harris and Sledge27,28 describe both TKR and THR as major surgeries with inherent risks, including loosening of the implant, fractures of surrounding bones, and infection. THR should only be undertaken in patients for whom more conservative treatments have not proven effective in managing pain and functional status.27 TKR and THR are not as effective in patients who are younger, active, or obese.27,28

At one time, it was believed that approximately 25% of patients with RA would require a TKR or THR procedure over a 10-year time period.29 However, in a recent Swedish study using a structured review of medical records, Hekmat and colleagues30 demonstrated that total joint replacement in RA was on the decline, even as joint replacement surgeries in general were increasing. This may be in part because of improved treatment modalities, including the use of biologics to treat RA.

We analyzed the universe of patients (RA and non-RA) who received these procedures. For patients receiving TKR or THR, we studied the impact of biologics for the treatment of RA on the likelihood that RA was the primary reason for surgery.

METHODS
This study utilized the Nationwide Inpatient Sample (NIS) of the Healthcare Cost and Utilization Project (HCUP) administered by the Agency for Healthcare Research and Quality. NIS is a nationally representative data set of US hospital inpatient discharges and contains approximately 8 million observations of hospital stays each year, comprising an estimated 90% of all hospital discharges in the United States. It also provides detailed information on patient admission source, principal and secondary diagnoses codes, International Classification of Diseases, Ninth Edition, Clinical Modification (ICD-9-CM) procedure codes, patient demographic information (eg, age, sex, race, and health insurance status), as well as characteristics of the admitting hospital such as hospital location and ownership type. In the HCUP inpatient databases, the first listed diagnosis is the principal diagnosis31 defined as “the conditions established after study to be chiefly responsible for the patient’s admission to the hospital.”32

Patients were required to be at least 18 years of age and to have received THR (primary ICD-9-CM procedure code 81.51) or TKR (primary ICD-9-CM procedure code 81.54) between 1993 and 2008 in 1 of 17 different US states (Arizona, California, Colorado, Connecticut, Florida, Iowa, Illinois, Kansas, Massachusetts, Maryland, New Jersey, New York, Oregon, Pennsylvania, South Carolina, Washington, and Wisconsin).

The main methodology for this project was the difference-in-differences (DID) technique, frequently used in economics and health services research, to measure the change induced by a particular treatment or event, in this case, the advent of biologics for the treatment of RA. The basic premise of this DID analysis was to examine the effect of biologics by comparing patients with RA before and after the introduction of each biologic (etanercept, infliximab, and adalimumab) with non-RA patients who had similar demographics and comorbidity profiles (the control group). DID utilizes such a control group to subtract out other changes (identical to the treatment and control groups) that took place across time. In contrast to a within-subjects estimate of the treatment effect, which measures the difference in an outcome before and after treatment, or a between-subjects estimate of the treatment effect, which measures the difference in an outcome between the treatment and control groups, the DID estimator represents the within-subjects difference between the treatment and control groups, pretreatment and posttreatment.

The study used the following regression equation:



Where Y is an indicator variable that equals 1 when RA is the primary diagnosis and 0 otherwise. PE, PI, and PA are time dummies, equaling 1 in the periods post etanercept, post infliximab, and post adalimumab, respectively (eg, post infliximab, PI, equals 0 up until November 1999 and equals 1 from December 1999 on).

RA is a treatment dummy indicating patients with RA as a comorbid condition (ICD-9-CM code 714.00, either as primary or secondary). The interaction of the time dummies and RA appears in the 3 bracketed terms. These are dummy variables equal to 1 only when a patient has an RA diagnosis in the post period.

The DID coefficients are β5, β6, and β7. Their signs, magnitudes, and statistical significance are the basis for inference under this study design. For the likelihood of RA as a primary diagnosis, all coefficients were expected to be negative.

X is a vector of patient, hospitalization, hospital, and local characteristics. Patient and hospitalization controls included patient age, sex, race, discharge outlet, mortality, inpatient length of stay, total charges, insurance type, source of admission, number of diagnoses, and procedures. Hospital characteristics included number of beds, teaching status, and ownership type. Additional variables included year, month, and state indicators, as well as local characteristics such as median income in the patient’s zip code.

Alternative model specifications produced similar results, including one with a fully interacted model (year RA) or stratified by procedure (see eAppendix A, available at www.ajmc.com). All regression models used sampling weights provided in NIS to reflect a nationally representative sample of US hospital inpatient discharges. All statistical analyses were performed using Stata 10 (StataCorp LP, College Station, Texas).

RESULTS
Table 1 summarizes the trends in key variables. There were more than 1.7 million THR and TKR procedures during the time period studied. The number of annual THR and TKR procedures almost tripled from 64,123 in 1993 to 182,359 in 2008. Over this period of time, the inpatient average length of stay (LOS) fell from 7.6 days to 3.5 days. The average patient age fell from 68.8 to 65.8 years, and the proportion of patients with private insurance increased while the proportion of patients with Medicare coverage declined.

Patients were partitioned into 3 mutually exclusive groups: 21,926 patients with RA as their primary diagnosis (first diagnosis code listed), 36,110 patients with RA as their secondary diagnosis (RA appeared between the 2nd and 15th diagnosis codes listed), and 1,684,899 patients with no indication of RA. The number of annual THR procedures more than doubled between 1993 and 2008 (from 25,987 procedures in 1993 to 56,478 in 2008), and the number of annual TKR procedures more than tripled over this period (from 38,136 procedures in 1993 to 125,881 in 2008). Patients with a primary diagnosis of RA were younger than patients with RA as a secondary diagnosis and non-RA patients (61.4 years vs 65.8 years and 67.2 years, respectively), more likely to be female (79% vs 76% and 61%, respectively), less likely to be white (78% vs 83% and 87%, respectively), and less likely to be admitted from the emergency department (Table 2).

The upper and lower panels of the Figure plot the trends for patients receiving THR surgery and TKR surgery, respectively. The y axis on the right-hand side pertains to the third category, which is far larger in magnitude, and the y axis on the left-hand side pertains to RA patients. Of all patients receiving THR surgeries, 6612 had RA as their primary diagnosis and 12,111 had RA as a secondary diagnosis. Of all patients receiving TKR surgeries, 15,314 had RA as a primary diagnosis and 23,999 had RA as a secondary diagnosis. The trends for patients with RA as a secondary diagnosis and patients with no indication of RA were similar for both THR and TKR procedures, and rose during the study period. However, the number of procedures for patients with a primary diagnosis of RA decreased moderately. Furthermore, the mean number of diagnoses per patient grew from 3.84 in 1993 to 6.65 in 2008 (Table 1).

Osteoarthrosis was the most common primary diagnosis for both THR and TKR surgeries during the study period. For RA patients undergoing THR, 35% had RA as the primary diagnosis, 45% had osteoarthrosis as the primary diagnosis, 7% had aseptic necrosis of bone, 5% had fracture of the neck of the femur, and 8% had a different primary diagnosis. For RA patients with TKR, 39% had RA as the primary diagnosis, 58% had osteoarthrosis as the primary diagnosis, and 3% had a different primary diagnosis (see eAppendix B, available at www.ajmc.com).

Table 3 presents results from the regression equation on the full sample of patients. The results are presented for 6 different specifications in which additional covariates are added sequentially. In all models, the results for the 3 interaction terms were negative and statistically significant. Summing the coefficient estimates for the most saturated model suggests approximately a 30% reduction in the likelihood that RA was the primary reason for surgery among patients with RA listed as a diagnosis in any position, and could therefore be attributable to the timing of introduction of the 3 biologics. When looking at a subset of knee and hip replacement surgeries (see eAppendix B), the introduction of etanercept, infliximab, and adalimumab was jointly associated with a 32% and 24% reduction in RA being the primary reason for TKR and THR, respectively.

Table 4 presents results from the regression equation on a subsample of patients who were 65 years or older and covered under Medicare; this group represents a growing portion of the US population because of aging. The results were consistent with those of the full sample, with approximately a 28% reduction in the likelihood that RA was the primary reason for surgery. All regression models used robust standard errors.

DISCUSSION
Study results showed that during the study period the number of annual THR and TKR procedures almost tripled, but for patients with a primary diagnosis of RA, the rates fell. The introduction of biologics for the treatment of RA was associated with a reduction in the proportion of patients with a primary diagnosis of RA among all RA patients receiving TKR and THR surgeries. This finding supports earlier research by Wolfe and colleagues,29 who found that 25% of patients with RA would require TKR or THR over a 10-year period of time. However, like this study, a recent study from Sweden demonstrated a decline in total joint replacement for patients with RA, even as joint replacement surgeries in general were increasing. 30 Results from our study and the Swedish study suggest that the availability of biologic agents, as well as other changes in the therapeutic approach to RA, may confer long-term benefits to both RA patients and healthcare systems.

Although the annual cost of TNF inhibitors ranges from $15,000 and $20,000,33 evidence of these drugs’ ability to offset medical expenditures for patients with RA is an important factor when assessing their cost-effectiveness. Direct medical costs of RA in the United States have been estimated to be $8.4 billion annually, and when adding in costs to employers and caregivers, and quality-of-life-related costs, this figure rises to almost $40 billion annually.8 Evidence presented in this study points to the possibility that TNF inhibitors may be able to reduce some of the high medical cost of RA. Additionally, delaying TKR and THR may improve quality of life for patients with RA as well as improve productivity, thereby reducing overall societal costs of RA. Future research on the ability of TNF inhibitors to reduce costs related to RA is important when trying to weigh the cost-benefit ratio of these drugs.

LIMITATIONS
This study has several limitations. In the multivariable models, it was not possible to control for reimbursement changes over time that may have influenced coding practices. In addition, it is not always possible to know fully all the factors affecting the decision to code RA as the primary or secondary reason for surgery. When first introduced, TNF inhibitors tended to be mostly used in RA patients with the most severe disease. Many of these patients would have had RA for a number of years and suffered joint destruction related to the burden of inflammation over time. With highly effective treatment, the signs and symptoms of active disease would have come under control, and the patients would then have their RA coded as a secondary diagnosis when they underwent THR or TKR.

CONCLUSION
This study found that since the introduction of etanercept, infliximab, and adalimumab, there has been a reduction in the likelihood of RA being the primary diagnosis for hospital admission for total hip and knee replacement surgery. The results are similar in the overall patient sample and the subset of patients aged 65 years and over. This suggests that highly effective therapies can mitigate some of the longer-term health and economic burdens of this serious disease.

Acknowledgments
Ann Neuer of Medical deScriptions provided editorial support for this manuscript.