Estimating the response and economic burden of rheumatoid arthritis patients treated with biologic disease-modifying antirheumatic drugs in Taiwan using the National Health Insurance Research Database (NHIRD)

Background Previous studies in Taiwan utilizing the Taiwan’s National Health Insurance Database (NHIRD) have estimated the direct healthcare costs of RA patients, but they have not focused on patients on bDMARDs, or considered patients’ response to therapy. Objectives The objective of this study was to estimate the rate of inadequate response for patients newly treated with biologic disease-modifying antirheumatic drugs (bDMARDs) as well as their costs and resource use. Methods Data were from the catastrophic illness file within the NHIRD from 1/1/2009 to 12/31/2013. Patients with RA, which was categorized by the presence of a catastrophic illness card, that were previously bDMARD-naïve, were included in this study if they initiated their first bDMARD during the index period. The index period included all of 2010, a pre-index period consisting of the index date– 365 days, and a follow-up period including the index date to 365 days post-index, were also included. Previously biologically-naïve patients were indexed into the study on the date of their first claim for a bDMARD. A validated algorithm was used to examine the rate of inadequate response (IR) in the biologically-naïve cohort of patients. Inadequate responders met one or more of the following criteria during their year of follow-up: low adherence (proportion of days covered <0.80); switched to or added a second bDMARD; added a new conventional synthetic DMARD (csDMARD); received ≥1 glucocorticoid injection; or increased oral glucocorticoid dosing. All-cause mean annual direct costs and resource use were measured in the year of follow-up. Costs were converted from NT$ to USD using 1 NT$ = 0.033 USD. Results A total of 818 patients with RA initiated their first bDMARD (54% etanercept and 46% adalimumab) in 2010. After one year of follow-up, 32% (n = 258) were classified as stable, 66% (n = 540) had an IR, and 2% (n = 20) were lost to follow-up. During the follow-up period mean annual total direct costs were $16,136 for stable patients compared to $14,154 for patients with IR. Mean annual non-medication direct costs were $937 for stable patients and $1,574 for patients with IR. Mean annual hospitalizations were higher for patients with IR (0.46) compared to stable patients (0.10) during the one year follow-up period. Conclusions The majority of patients that were previously naïve to bDMARDs had an IR to their first bDMARD during the year of follow-up. Patients with an IR had numerically increased all-cause resource utilization and non-medication costs during the follow-up period compared to patients with stable disease. This level of IR suggests an unmet need in the RA treatment paradigm.

Introduction Rheumatoid arthritis (RA) is a chronic, systemic inflammatory disease characterized by synovial inflammation, swelling, cartilage and bone destruction, and other systemic features [1]. Within low-to-middle income countries, the prevalence of RA is estimated to be slightly lower for Southeast Asian countries than those in Europe, the Americas, and Western Pacific regions [2]. The mean age-adjusted incidence of RA in Taiwan is estimated to be 15.8 per 100,000 [3,4]. The estimated prevalence of RA in Taiwan was 99.6 per 100,000 in 2007 [4].
Patients with RA often experience morning stiffness, pain and limitations of motion, and general difficulty performing daily activities [5]. Extra-articular and systemic manifestations are commonly associated with RA patients and affect various tissues and organ systems [6,7]. These extra-articular symptoms include fever, fatigue, malaise, depression, weakness, and/or cardiovascular disease (CVD) [8,9].
Patients with RA can experience comorbidities such as depression, asthma, cardiovascular events (myocardial infarction, stroke), solid cancers, and chronic obstructive pulmonary disease. However, there is high inter-country variability in the prevalence of comorbidities [10]. RA comorbidities have been shown to shorten patient life expectancy [11,12], primarily due to the increased prevalence of CVD, a greater incidence of infections, and cancer [13][14][15][16][17]. Furthermore, osteoporotic fractures are more commonly observed in RA patients and can lead to functional decline [18,19]. A retrospective database analysis of Taiwan's National Health Insurance Research Database (NHIRD) using data from 2004 to 2010 found a substantially increased incidence rate of hip fractures in younger patients with RA vs. those without across all age groups [20].
There are three main medication classes used to treat RA, depending on the severity of disease and symptoms. These classes include nonsteroidal anti-inflammatory drugs (NSAIDs), corticosteroids, and disease-modifying antirheumatic drugs (DMARDs) [21]. Furthermore, the DMARD class of medications include conventional synthetic DMARDs (csDMARDs), the HWDC, Department of Statistics, Ministry of Health and Welfare, Taiwan (http://dep.mohw.gov. tw/DOS/np-2497-113.html).
Funding: This study was funded by Eli Lilly and Co (https://www.lilly.com/). Eli Lilly provided financial support in the form of salaries for QS, TT, CG, CL, WW. Eli Lilly paid the following authors as consultants: KL, BW, WF, CT. Elysia Group provided financial support in the form of salaries for BW and WF. The specific roles of these authors are articulated in the author contributions. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests:
The authors have read the journal's policy and have the following conflicts: QS is an employee of Eli Lilly and owns stock in Eli Lilly. KL is a paid consultant for Eli Lilly. TT is an employee of Eli Lilly and owns stock in Eli Lilly. BW is an employee of Elysia Group and a paid consultant for Eli Lilly. CG is an employee of Eli Lilly and owns stock in Eli Lilly. CL is an employee of Eli Lilly. WW is an employee of Eli Lilly and owns stock in Eli Lilly. WF is an employee of Elysia Group and a paid consultant to Eli Lilly. CT is a paid consultant for Eli Lilly. There are no patents, products in development or marketed products to declare. This does not alter our adherence to all the PLOS ONE policies on sharing data and materials.
For patients utilizing biologics, a significant proportion either do not respond, discontinue, or lose efficacy over time [23][24][25]. Response to therapy can vary for patients based on disease factors and treatment choice. Typically biologic therapy is initiated using TNF-inhibitors. However, an estimated 30% of patients have an inadequate response to them [23]. While patients will often switch to a different TNF-inhibitor rather than a different class of medication, research has shown inadequate response to an initial TNF-inhibitor to be a predictor of response and tolerance of a second TNF-inhibitor [26].
In Taiwan, patients with RA must fit the following treatment criteria: 1) fulfill ACR criteria for RA, 2) have continuously active RA, which is defined as a disease activity score (DAS) 28 score of >5.1, and 3) have failed csDMARD treatment, defined as being treated with at least two csDMARDs which resulted in no significant efficacy in order to be considered for reimbursement of a bDMARD [27]. In Taiwan, reimbursement for the continuous use of biologics is limited to three years before patients must discontinue with a mandatory break before reapplying for reimbursement [27]. Rather than limiting time on therapy, guidelines have increasingly called for a treat-to-target strategy [28]. This treatment strategy involves continuous assessment of disease activity every 1-3 months, with adjustments made to therapy if patients are not reaching either remission or low disease activity.
This study attempts to quantify the number of patients not responding to their newly initiated bDMARD therapy, their healthcare resource utilization (HCRU) and the economic burden in Taiwan, using the NHIRD. Previous studies in Taiwan utilizing the NHIRD have estimated the direct healthcare costs of RA patients, but they have not focused on patients on bDMARDs, or considered patients' response to therapy [29][30][31]. This study will provide evidence for decision makers to understand the real-world rate of response to therapy, and the differences in HCRU and costs between patients responding, and not responding to therapy.

Data source
This study utilized Taiwan's National Health Insurance Database, which is a population-based claims database covering over 99% of Taiwan's population. Claims data from 1/1/2009 to 12/ 31/2013 were included in the study. Subjects included in the catastrophic illness file, which is a registry file containing all cases of patients with major diseases, were analyzed for their RA diagnosis.
Previously bDMARD-naïve RA patients initiating a bDMARD during the index period were included in the study if they had any healthcare event with an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code 714.xx (excluding 714.3), a claim for a bDMARD in the index period, and the presence of a catastrophic illness card for RA. Patients were excluded from the study if they: 1) were aged <18 years at index; 2) were not in the database in the pre-index period (ignoring breaks of <30 days); 3) had !1 claim for any of the following conditions: comorbid psoriasis or psoriatic arthritis (696.xx), ankylosing spondylitis (720.0x), Crohn's disease (555.xx), ulcerative colitis (556.xx) and juvenile chronic polyarthritis (714.3x), at any time during the study period; or 4) had a diagnosis of any solid tumors or hematologic malignancy (140-208.xx), except for non-melanoma skin cancer (173. xx) or human immunodeficiency virus (HIV)/acquired immune deficiency syndrome (AIDS) (042.xx, 043.xx, 044.xx) at any time during the study period.

Study population
Patients identified as bDMARD-naïve were followed for one year post-index to evaluate their response. Patients were divided into three possible cohorts: 1) bDMARD stable disease; 2) bDMARD-Inadequate Response (IR); or 3) bDMARD Unknown status. Response was measured using an algorithm adapted from Curtis et al. to Taiwan [32]. The algorithm considered five criteria for response. Patients meeting all five criteria were considered to have stable disease. Patients meeting one or more of the IR criteria were considered inadequate responders, and those lost to follow-up were considered to have unknown status. The response criteria included: (1) adherence (<80% proportion of days covered); (2) a switch or add of a bDMARD; (3) the addition of a new non-biologic DMARD; (4) more than one glucocorticoid joint injection; and (5) an increase in the dose of an oral glucocorticoid. The algorithm also included the criteria of an increased dose in bDMARD to be a measure of inadequate response. In localizing this algorithm to Taiwan we did not consider this criteria as patients are not able to increase their dose per the reimbursement guidelines [27].

Statistical analysis
Baseline demographics (age, gender, comorbidities) and treatment mix for each of the cohorts were measured. Comorbidities were measured using a claims-based algorithm to estimate patients' Charlson Comorbidity Index (CCI) [33]. A total of three years of follow-up postindex were analyzed to evaluate switching patterns, HCRU, and costs. Included bDMARDs were abatacept, adalimumab, etanercept, golimumab, and tocilizumab. The tsDMARD tofactinib was not reimbursed during the study period and was therefore not included in the analyses. Direct costs were expressed as mean annual costs for each of the cohorts, and separated into major categories including: outpatient, inpatient, emergency department (ED), and medication. Resource utilization was the mean annual quantity of service (hospital admissions, days of hospitalization, outpatient visits, and ED visits) used. All outcomes were measured for each of the patient's groups. All analyses were for descriptive purposes only and no formal statistical inference testing were conducted. The software SAS (version 9.4) was used in management and descriptive analysis of data. All costs were converted to 2017 USD using an exchange rate of 1 TWD = 0.033 USD [34].

Incidence of RA patients initiating a biologic
A total of 110,645 patients with a diagnosis of RA in 2010 were included in the study. Of these patients, a total of 818 fit the study criteria to be classified as patients initiating a bDMARD in 2010 (Fig 2).
Patients initiating a bDMARD in 2010 had a mean age of 55.5 years, were 84% female, and had a mean CCI of 1.65 (Table 1). All patients either initiated adalimumab (n = 375, 46%) or etanercept (n = 443, 54%) as their first biologic therapy.
Within the bDMARD naïve cases (n = 818), after one year of follow-up, a total of 258 (32%) patients were confirmed stable, 540 (66%) patients were classified as inadequate responders, and 20 (2%) patients had the status of unknown. As shown in Table 1, patients with an inadequate response to their biologic had a higher numerically mean age than those confirmed stable (55.7 vs. 54.1 years).
Inadequate response was driven by low adherence (36%) followed by an increase in glucocorticoid dose (20%), !1 glucocorticoid joint injection (17%), addition of a new csDMARD (8%), and finally, switching or adding a biologic (3%).  Of the patients who switched to another biologic medication, the mean time to switch from first-line therapy for all bDMARD naïve patients was 782.4 days for patients that started on adalimumab and 755.6 days for those starting on etanercept. As expected, patients with an inadequate response had a shorter time to switch to a new bDMARD than the confirmed stable cohort due to the IR algorithm. (Fig 3). The difference in mean days to switch between the stable and IR cohorts was significant for etanercept users (p = 0.0031), but not for adalimumab users (p = 0.0566).

Healthcare resource utilization
Mean annual resource utilization was collected for each of the cohorts for years one through three ( Table 2). For years one through three, the bDMARD-IR cohort of patients had higher numerically mean annual hospital admissions, hospital days, and outpatient visits.

Direct costs
Direct medical costs were measured for each of the cohorts, over three years, and categorized into non-medication costs (inpatient, outpatient, and emergency department costs) and medication costs ( Table 3). All bDMARD naïve patients had mean total direct costs of $14,795, $11,460, and $10,395 from years one through three, respectively. For each of these years, medication costs were the primary driver contributing 91%, 88%, and 87% of costs in years one through three, respectively.
When responder status was considered, confirmed stable patients in general had 14% ($16,136 vs. $14,154), 23% ($13,083 vs. $10,674), and 20% ($11,734 vs. $9,742) numerically  higher mean costs than the IR patient cohort over the three years post initiation, respectively. This increase in costs was driven by medication costs, which remained high for stable patients and decreased significantly for IR patients, as the primary driver of IR was low adherence. When only non-medication costs were considered, stable patients had lower mean outpatient, inpatient, and ED costs for years one through three, post initiation. In total, mean non-medication costs were 40%, 17%, and 14% numerically lower for stable patients compared to IR patients, in years one through three, respectively.

Discussion
In this population-based analysis, a total of 3,399 naïve RA patients with a catastrophic illness card initiated either a bDMARD or csDMARD in 2010. Of these patients, 2,581 (76%) started on csDMARDs and 818 (24%) started on bDMARDs. While bDMARDs are a subsequent line of therapy, our results suggest over two-thirds of patients initiating bDMARDs are inadequate responders after one year of therapy.
Our study results are in-line with previous analyses conducted using the same algorithm in different databases. The IR rate of 66% in our study was slightly lower than the original validation of the algorithm in the Veteran Health Administration's RA Registry of 72% [32]. The algorithm was also utilized using Medicare data from 2006-2010 for biologic-naïve patients with RA. At one year the rate of IR was 75% across all treatments (etanercept = 28%; abatacept = 26%; adalimumab = 24%; infliximab = 23%). When only considering adalimumab and etanercept, as in our study, the IR rate was 74% [32]. The lower rate of IR in our study can mostly be explained by the lack of dose increases in Taiwan, which is regulated through reimbursement guidelines. Additionally, our study only included etanercept and adalimumab as first-line therapies. The high concentration of patients on etanercept, which is a human soluble dimeric TNF receptor fusion protein, and lack of patients on infliximab, a mouse-human chimeric mAb against TNF could have contributed to patients remaining on their biologic for longer periods and decreasing the number of IR patients. Etanercept has been shown to have fewer patients develop anti-drug antibodies compared to adalimumab and infliximab [35]. The results of our study also reaffirm the 30% rate of inadequate response to initial TNF-inhibitors presented by Furst and Emery [23].
In the first year of initiation, patients with an inadequate response to bDMARDs had mean annual total direct healthcare expenditures of $14,154 compared to $16,136 for patients classified as stable. The higher costs in the stable cohort were driven by medication costs, as IR patients often discontinued their medication, reducing costs. When only non-medication costs were considered, All patients either initiated adalimumab (n = 375, 46%) or etanercept (n = 443, 54%) as their first biologic therapy. Resource utilization for IR patients was numerically higher than stable patients over the first year after initiation of therapy. This numerical increase in HCRU was 7-fold for hospital days (3.4  The results of this study highlight an unmet need within the treatment paradigm. In Taiwan, bDMARDs accounted for nearly 57.9% ($50M USD of $86.4M USD) of direct medical costs for RA patients in 2011 [31]. Because of the high medication costs associated with biologics, the early identification of inadequate responders can help payers to redirect these costs to treatments with better health outcomes.
Future extensions of this research could include an analysis of the response to second line therapies to better understand the impact of TNF-cycling versus utilizing a treatment with a new mechanism of action for patients with inadequate response to their first therapy. Additionally, an analysis of the societal costs associated with patients considered stable vs. those with inadequate response could be explored. The results of this study could aid in the development of cost-effectiveness and budget impact models, which could help payers to understand the value and financial impact of initiating more effective medications earlier in the treatment paradigm to reduce the number of inadequate responders.
Our study includes several limitations. First, data for this study was only available through 2013. Our cohort of patients were only initiated on either adalimumab or etanercept, both TNFs, and did not examine bDMARDs with alternative mechanisms of action. Second, this treatment algorithm to determine response was developed, and preliminarily validated in the Veterans Health Administration, which is a claims-based database within the Department of Veteran Affairs (VA) RA Registry in the United States. The applicability of the algorithm outside the US, specifically Taiwan, has not previously been examined. The algorithm also relies on a set of criteria that are related to effectiveness, but not directly tied to response. Patients initiating or escalating oral glucocorticoids were said to have inadequate response, however patients could have initiated or escalated due to a comorbidity such as COPD and not due to RA. Lastly, the dataset is a claims-based dataset, which served as a proxy for clinical outcomes. The true clinical response of patients remains unknown, while the predictive algorithm has been shown to estimate the percentage of inadequate responders fairly accurately, it may be prone to misclassification. Additionally, the claims-based dataset does not include clinical variables, which limited the variables used for inclusion/exclusion in the study and could have led to bias in the study results.

Conclusion
In conclusion, this study highlights the unmet needs within the RA treatment paradigm; patients may still experience lack of response to treatments, which incur HCRU and costs regardless of effectiveness. Of patients initiating their first biologic, 66% had an inadequate response by the end of their first year. These patients with inadequate response had numeric higher HCRU and non-medication direct healthcare costs. This analysis could provide information to key decision makers when developing reimbursement guidelines, which could help inform physicians' decisions in clinical practice.