Jianwei Leng, MS

: An effectiveness algorithm that was validated with the Veterans Affairs Rheumatoid Arthritis Registry was used to compare effectiveness of different biologic disease-modifying antirheumatic drugs (DMARDs) and cost per effectively treated patient in the Veterans Health Administration (VHA). 

: Effectiveness was estimated with the algorithm using VHA claims data for biologics (abatacept, adalimumab, etanercept, infliximab, or rituximab) newly initiated from 2008 to 2011. 

: Patients were considered effectively treated if they did not fail any of 6 algorithm criteria during 1 year of follow-up: 1) low treatment adherence; 2) biologic dose escalation; 3) biologic switch; 4) addition of nonbiologic DMARD; 5) new or increased oral glucocorticoid; and 6) more than 1 intra-articular glucocorticoid injection. Index biologic costs included medication and administration costs. Subgroup analyses were conducted by gender, smoking status, serologic status, and body mass index.

: The analysis included 4696 patients (mean age = 61 years; 87% male). The proportion of patients categorized as effectively treated was higher for subcutaneous biologics (adalimumab, 33.4%; etanercept, 32.3%) than for intravenous biologics (rituximab, 27%; infliximab, 26.4%; abatacept, 24.8%). Index biologic cost per effectively treated patient was lower for etanercept ($39,432) and adalimumab ($41,519) than for infliximab ($54,524), abatacept ($55,776), and rituximab ($58,260). Men had lower cost per effectively treated patient than did women ($40,745 vs $57,450). Outcomes were comparable within other subgroups.

: In veterans with rheumatoid arthritis, the subcutaneous biologics adalimumab and etanercept had the highest proportions of patients categorized as effectively treated and the lowest costs per effectively treated patient.

This study applied a validated algorithm to estimate effectiveness and costs of the index biologic used by 4696 US veterans with rheumatoid arthritis.

Compared with intravenous biologics (abatacept, infliximab, and rituximab), subcutaneous biologics (etanercept and adalimumab) had lower 1-year biologic costs, higher rates of effectiveness according to the algorithm, and lower costs per effectively treated patient.

Men had lower costs per effectively treated patient than women; outcomes were similar within other subgroups.

Previous studies in managed care settings used similar methods, but this was the first study to report comparable results in the unique setting of the Veterans Health Administration.

Approximately 1.3 million adults in the United States have rheumatoid arthritis (RA), an inflammatory autoimmune condition that primarily affects the synovial membrane in the joints of the body,

 causing inflammation in the joints and destruction of bone and cartilage.

 Biologics including abatacept, adalimumab, certolizumab pegol, etanercept, golimumab, infliximab, rituximab, tocilizumab, and anakinra play an important role in the treatment of RA by helping to regulate the body’s inflammatory process and inhibiting joint damage.

 The biologics differ in their method of administration, dosing level, dosing schedule, and whether they are approved for first-line use after treatment failure with a nonbiologic disease-modifying antirheumatic drug (DMARD).

Comparisons of effectiveness of different biologics and their costs in RA are critical for formulary decision makers to develop evidence-based formularies. Economic comparisons of biologics in RA have been published using managed care data,

 but only limited work has focused on the Veterans Health Administration (VHA).

 The VHA has a number of characteristics that make it difficult to generalize from managed care data, including a unique set of demographic characteristics (members are predominantly older men), physicians who are salaried rather than reimbursed on a fee-for-service basis, medication prices that are governed by the Federal Supply Schedule (FSS) and BIG 4 (Veterans Affairs [VA], Department of Defense, Coast Guard, and Public Health Service) pharmaceutical pricing rather than wholesale acquisition costs, and small or no patient co-payments.

 For these reasons, a detailed analysis of this population was expected to be informative for VHA decision makers, with potential application to other health systems.

Previous cost analyses for biologics in RA have been limited by the lack of information about effectiveness. Lower effectiveness leading to worse long-term outcomes with a biologic could offset any cost savings associated with that biologic. Curtis et al

 developed and validated a claims-based effectiveness algorithm using clinical outcomes in patients in the Veterans Administration Rheumatoid Arthritis (VARA) registry receiving biologic and nonbiologic DMARD therapy.

The algorithm uses data commonly available in claims data sets to estimate the effectiveness of biologics. The algorithm evaluates 6 criteria: adherence to the biologic, increases in biologic doses, switching biologics, adding a new nonbiologic DMARD, starting or increasing the dose of oral glucocorticoids, and the use of multiple intra-articular glucocorticoid joint injections. In the validation, the gold standard for effectiveness was a disease activity score using 28 joint counts (DAS28) of ≤3.2 units (the threshold for low disease activity) or improvement in DAS28 >1.2 units after 1 year of DMARD therapy. The DAS28 is a widely accepted metric of disease activity in RA that is commonly used both in clinical trials and in routine real-world settings. The sensitivity of the effectiveness algorithm was 75% and its specificity was 90%.

The positive predictive value and negative predictive value of the effectiveness algorithm were 76% and 90%, respectively. Recent studies applied the algorithm to commercial claims databases to estimate effectiveness and biologic cost per effectively treated patient,

 but this approach has not been applied to the VHA, which is a large purchaser of biologic agents.

The objectives of this study were to compare the estimated effectiveness of different biologics in veterans with RA by using a published algorithm based on administrative data

 to assess the reasons for treatment failure and estimate biologic costs per effectively treated patient.

This study was approved by the University of Utah Institutional Review Board and the VA Research Service. All data were from the VA Medical System. Patient data, procedures, and text notes were obtained from the Corporate Data Warehouse (CDW) and accessed through the Veterans Informatics and Computing Infrastructure.

Pharmacy dispensing data were obtained from CDW, Decision Support Services (DSS), and VA Pharmacy Benefits Management (PBM) Services. Because discrepancies were found in DMARD dispensing across the VHA data marts, a harmonization algorithm was developed to resolve such discrepancies that prioritized dispensing data found in multiple data sources based on source, with CDW first, PBM second, and DSS third. Priority was given to CDW because it is now considered production data; PBM now acquires their data from CDW and then applies data-cleaning algorithms; and DSS because it is now considered legacy data.

Since this study required careful tracking of average weekly dose to determine if a dose escalation occurred, accurate estimates of dosing were needed. In typical pharmacoepidemiological research, the dispensed quantity, unit strength, and days’ supply are used to estimate the prescribed average weekly dose. This approach, however, is unreliable for biologic DMARDs dispensed by the VHA system that are subcutaneously administered at home and for medications with complex weekly dosing schedules, such as methotrexate, because of known inconsistencies in pharmacy dispensing data.

The primary problem in the VHA setting is that the quantity dispensed is not well standardized; for example, the quantity dispensed for etanercept is at times listed differently at various VHA medical centers—dispensing of a box of 4 syringes of etanercept 50 mg may be listed as a quantity of 4 (meaning 4 syringes), or 1 (meaning 1 box of 4 syringes), or 200 (meaning 200 mg total per box). Fortunately, the VA CDW and PBM data contain some information on dispensed units in addition to the narrative medication schedules from prescriptions (sigs).

Determining dispensed quantity from sigs on a large scale is labor- and time-intensive, so an automated Natural Language Processing (NLP) tool was developed to extract and standardize the prescribed average weekly dose from the medication sig.

 This information was used with pharmacy dispensing data on days’ supply, unit dose, and quantity to generate the best estimate of average weekly dose. The process was validated by medical record review and comparison with the VARA registry.

Similar to the limitations in the VHA discussed above for subcutaneous biologics, outpatient infusions of medications such as intravenous biologics are not consistently documented in structured VHA pharmacy dispensing or nurse Barcode Medication Administration data.

 Healthcare Common Procedure Coding System procedure codes intended for administrative billing provide information on outpatient infusions but they are not used consistently, which makes it difficult to compute the total infused dose.

 An NLP tool was developed to search for outpatient infusion notes and extract total infused dose of biologic DMARDs.

 The process was validated by chart review and comparison with the VARA registry.

US veterans receiving care within the VHA were included in the analysis if they had at least one claim between January 1, 2008, and December 31, 2011, for abatacept, adalimumab, anakinra, certolizumab, etanercept, golimumab, infliximab, rituximab, or tocilizumab. Patients with a claim for subcutaneous abatacept were removed from the analysis because abatacept was approved only for intravenous use before the start of the study period. Subcutaneous tocilizumab and intravenous golimumab were not included in the analysis because they were not approved at any time during the study period.

The first claim for one of these biologics provided the definition for the patient’s index date and the index biologic. Patients were included only if they were enrolled in the VHA for at least 365 days before their index date. Patients were required to be aged at least 18 years on the index date and to have a diagnosis of RA (

International Classification of Diseases, Ninth Revision, Clinical Modification

]: 714.0x) on at least 1 claim in the 365 days prior to or 30 days after the index date.

Patients were excluded from the analysis if they had psoriasis (

: 556.xx), juvenile idiopathic arthritis (

: 200.xx, 202.xx), or chronic lymphocytic leukemia (

: 204.1x) on any claim in the 365 days prior to or 30 days after the index date. To limit the analysis to patients who newly initiated biologic therapy, established users of biologics (patients with any other biologic claim in the 180 days before the index date) were excluded from the analysis.

Other exclusion criteria were an index claim for a biologic before it received FDA approval for treatment of RA, claims for more than 1 biologic on the index date, and sparse use of VHA services (no claim within 180 days before the index date and between 10 and 24 months from index date, as a proxy for leaving VHA care). Biologic agents with a sample size <100 patients were not evaluated since the sample size was deemed too small to provide reliable estimates.

 the following 6 criteria were applied to each patient’s claims for the 1-year follow-up period (365 days) after the index date:

Low adherence to the index drug: a) for subcutaneous biologics (adalimumab and etanercept): medication possession ratio <80%, defined as the total days’ supply in the follow-up period divided by 365 days; and b) for intravenous biologics: fewer infusions than recommended in the FDA-approved dosing schedules (ie, <12 abatacept infusions, <7 infliximab infusions, or <4 rituximab infusions).

Increase in biologic dose during the follow-up period compared with the beginning dose: a) abatacept: >100 mg dose increase or >18 infusions total; b) adalimumab: frequency increase to 40 mg once weekly; c) etanercept: frequency increase to 50 mg twice weekly; d) infliximab: >100 mg dose increase (with each dose rounded up to the nearest 100 mg) or >10 infusions total; and e) rituximab: no escalation rule was applied.

Switch to a new biologic agent during the follow-up period.

Addition of a new nonbiologic DMARD during the follow-up period.

Oral glucocorticoids: a) initiation of chronic glucocorticoids for those with no oral glucocorticoid prescriptions during the 180 days pre-index; and b) increase by >20% in oral glucocorticoid dose (based on prednisone equivalent dose

) between months 7 and 12 post index for those who received any oral glucocorticoid prescription during the 180 days pre-index.

More than 1 parenteral or intra-articular glucocorticoid joint injection on unique days between 90 days and 1 year post index.

According to the algorithm, the index biologic was considered “effective” and the patient was considered to be “effectively treated” if none of the 6 criteria were met. The index biologic was considered “not effective” if at least 1 of the 6 criteria was met. The proportion of patients (95% CI) that failed each of the criteria was analyzed. Proportions (95% CI) were also calculated for effective and not effective index biologics.

Total costs of the index biologic in the first year (ie, between the index date and 365 days after the index date, inclusive) were calculated by summing the lowest contracted amount using VHA prices from the 2013 fiscal year based on the FSS and BIG 4 (eAppendix 1 [eAppendices available at

 These were the most recent costs available at the time of the study. Intravenous abatacept was $1.58 per mg, adalimumab was $16.77 per mg, etanercept was $6.61 per mg, infliximab was $4.85 per mg, and rituximab was $4.70 per mg. The cost for each outpatient pharmacy dispensing episode for subcutaneously injected medication was $25 and the infusion costs were $169.09 per administration. Costs of other medications used in the first year, such as DMARDs or other biologics, were not included in the cost calculations.

One-year biologic cost per patient was calculated for each biologic as the total cost for the index biologic divided by the number of patients who received that index biologic. The 1-year cost per effectively treated RA patient (ie, a patient who met the algorithm effectiveness rule) was estimated for each index biologic. The metric reflected total expenditures by the VHA for the index biologic divided by the number of patients effectively treated by that biologic. It was calculated as follows:

Cost per effectively treated patient =  Total cost of index agent for all patients who index on it / Number of effectively treated patients on index agent

Subgroup analyses of effectiveness, biologic costs, and costs per effectively treated patient were stratified by the following baseline characteristics: gender (male or female), smoking status (current smoker or current nonsmoker), cyclic citrullinated peptide (CCP) (positive [>20 IU/mL] or negative), rheumatoid factor (positive [>20 IU/mL or >1:40 titer] or negative), and body mass index (≥30 kg/m

Continuous data are presented as means and 95% CIs, and categorical variables are presented as proportions and 95% CIs. The focus on 95% CIs instead of 

 values provides evidence for the stability of estimates along with statistical significance. In general, when the 95% CIs do not overlap between 2 groups 

 <.05 and the 2 groups are considered significantly different.

A total of 4696 patients met all inclusion and exclusion criteria (

), of whom 2069 (44.1%) started etanercept, 1989 (42.4%) started adalimumab, 267 (5.7%) started rituximab, 254 (5.4%) started infliximab, and 117 (2.5%) started abatacept. Eight patients who received subcutaneous abatacept in the first year were not analyzed. Fifty-seven patients who started anakinra, certolizumab, golimumab, and tocilizumab were not analyzed because of the small sample size (n <100 patients) for each of those agents.

Articles

This study applied a validated algorithm to estimate the effectiveness and costs of the index biologic used by 4696 US veterans with rheumatoid arthritis.