Risk and cost of infection-related hospitalizations in medicare beneficiaries with comorbid rheumatoid arthritis treated with abatacept versus other targeted disease-modifying anti-rheumatic drugs

Abstract

Objective

This study evaluated infection-related hospitalization risk and cost in tumor necrosis factor inhibitor (TNFi)-experienced and targeted DMARD (tDMARD) naïve rheumatoid arthritis (RA) patients that were treated with abatacept, TNFi, or other non-TNFi.

Methods

This retrospective study used 100% Medicare Fee-for-Service claims to identify patients ≥65 age, diagnosed with RA, and were either 1) TNFi-experienced, who switched from a TNFi to another tDMARD (subsequent tDMARD claim served as index), or 2) tDMARD naïve (first therapy claim served as index), who initiated either abatacept, TNFi, or non-TNFi as their first tDMARD, between 2010 and 2017. Follow-up ended at the date of disenrollment, death, end of study period, or end of index treatment, whichever occurred first. Infection-related hospitalizations included pneumonia, bacterial respiratory, sepsis, skin and soft tissue, joint or genitourinary infections. A Cox proportional hazard model and two part generalized linear model were developed to estimate adjusted infection-related hospitalization risk and costs. Costs were normalized to per-patient-per-month (PPPM) and inflated to 2019 US$.

Results

The infection-related hospitalizations rate was lower during follow-up than during baseline periods for abatacept users, but was reversed for both TNFi and other non-TNFi users in both TNFi-experience and tDMARD naïve (p value < .001 based on Breslow-Day test for homogeneity of odds ratios). Infection-related hospitalization PPPM cost was significantly lower in abatacept treated patients compared to TNFi (TNFi-experienced: by $74; tDMARD naïve: $42) and other non-TNFi (TNFi-experienced: by $68; tDMARD naïve: $60). The adjusted infection-related hospitalization risk was significantly higher for RA patients treated with TNFi (TNFi-experienced HR: 1.48; 95% CI: 1.26–1.75, p < .0001; tDMARD naïve HR:1.59; 95% CI: 1.43–1.77, p < .0001) and other non-TNFi (TNFi-experienced HR:1.46; CI:1.28–1.66; tDMARD naïve HR:1.63; 95% CI: 1.44–1.83) than with abatacept.

Conclusion

RA Medicare Fee-For-Service beneficiaries who either switched or initiated abatacept have a lower infection-related hospitalization risk and cost compared to patients who switched to or initiated other tDMARDs.

Keywords:

• Rheumatoid arthritis
• TNFi
• abatacept
• other non-TNFi
• infection-related hospitalization
• risk
• healthcare costs

JEL CLASSIFICATION CODES:

• I10
• I19

Introduction

Rheumatoid arthritis (RA) is one of the most common autoimmune diseases, affecting nearly 1.3 million people in the USA¹. RA is characterized by chronic inflammation of the joints, which can ultimately lead to cartilage and bone destruction². Though not directly life-threatening, RA severely impacts patients’ quality of life and imparts a major economic burden on healthcare systems and society³. RA contributes $19.3 billion (in 2005 US$) in direct and indirect costs in the USA annually⁴.

The treatment of choice in RA is a group of medications known as disease-modifying antirheumatic drugs (DMARDs). Initial treatment of active RA is typically with a conventional DMARD (cDMARD) such as methotrexate, sulfasalazine, or leflunamide^5,6. Patients who are intolerant or show an inadequate response to cDMARDs are often treated with targeted disease-modifying antirheumatic drugs (tDMARDs). There are multiple classes of tDMARDs including tumor necrosis factor inhibitors (TNFis) (adalimumab, certolizumab pegol, etanercept, golimumab, and infliximab), anti-interleukin-6 agents (tocilizumab), anti-CD20 agents (rituximab), T-cell costimulation modulators (abatacept), and Janus kinase (JAK) inhibitors (tofacitinib) ⁷.

The immunomodulatory effects of RA, immunosuppressive treatments, and immunocompromising comorbidities all contribute to an increased risk of infection among RA patients⁸. RA patients have nearly twice the rate of infection compared with matched non-RA controls, and are at excess risk of infections requiring hospitalization^8,9. The most common infections associated with RA are pneumonia, skin and soft tissue infections, urinary tract infections, bacteremia, and sepsis¹⁰. In a study of veterans in treatment for RA, 48% of tDMARD switching was due to adverse events (AEs) and 43% was due to lack of efficacy¹¹. Adaptive cellular immunity is impaired in RA patients by a constricted T-cell receptor repertoire, which is crucial for naïve T-lymphocytes to recognize all potential harmless and harmful antigens⁹. In addition, clonal expansion of naïve T-cells in response to a previously unknown antigen is significantly reduced in RA patients, weakening protection against infectious diseases. Furthermore, neutropenia is common in RA patients due to increased pathological immune complexes or direct anti-neutrophil antibodies. These disease-related alterations of the immune system in RA patients all increase the risk of infections in RA¹⁰.

In addition to the disease itself, therapy with corticosteroids and tDMARDs may predispose RA patients to infection. Glucocorticoids significantly increase the risk of infection by a factor of 2 and 4 in RA patients receiving 7.5 mg/day to 14 mg/day and ≥15 mg/day, respectively. Medications in the TNFi class increase the risk of serious infection by up to twofold in RA patients. Meta-analyses suggest a similar increase with non-TNFi biologics. Comorbidities such as diabetes mellitus, chronic obstructive pulmonary disease (COPD), and chronic kidney disease, as well as factors such as age, female gender, and smoking, may also increase infection risk in RA patients^12,13. Accurate assessment of the risk of infection in RA patients could inform clinical decision-making to reduce and prevent the occurrence of infections. Recently validated, the Rheumatoid Arthritis Observation of Biologic Therapy (RABBIT) risk score determines the risk of serious infection in individual RA patients based on clinical and treatment information¹². Additional research must be conducted to determine the contribution of each risk factor to the overall infection risk in RA patients.

In RA patients with prior exposure to a biologic agent, exposure to etanercept, infliximab, or rituximab was associated with a greater 1-year risk of hospitalized infection compared with exposure to abatacept¹³. Among RA patients who experienced a hospitalized infection while on TNFi therapy, abatacept and etanercept were associated with the lowest risk of subsequent infection compared with other biologic therapies¹⁴. Though previous studies evaluated patients who received any biologics and/or targeted synthetic DMARD treatment, most patients in the USA use TNFi as their first DMARD. Therefore, further studies are needed to investigate patients who use TNFi in the past since.

There is limited information on the costs associated with infection-related hospitalizations in RA patients. Furthermore, potential differences in the risk of infection-related hospitalizations associated with the use of various tDMARDs in Medicare patients who are TNFi-experienced have not been evaluated in patients naïve to tDMARD treatment till date. The current study used the 100% Medicare Fee-for-Service (FFS) claims database to evaluate the impact of tDMARDs among RA patients on the risk and cost of infection-related hospitalizations during follow-up. More specifically, among patients who were either TNFi-experienced or tDMARD naïve, we evaluated the impact of initiation or switch to abatacept, TNFis, and other non-TNFis on infection-related hospitalization risk and costs in the USA.

Methods

Study design and data source

This was a retrospective observational study using 100% Medicare FFS (Part A, B & D) claims and enrolment data from 01 January 2009 through 31 December 2017. The study cohort was derived from the 100% sample of the Medicare research identifiable files, which included Part A and Part B FFS claims data, and Prescription Drug Event (PDE) data for all Part D plans. The identifiable claims data comprised all medical and pharmacy encounters including hospital claims, emergency department (ED) visits, skilled nursing facility stays, hospital outpatient services/ambulatory surgical centre services, physician office visits (including physician administered treatments), home health services/durable medical equipment, hospice care, and pharmacy utilization.

Study population

Beneficiaries ≥65 years of age with a diagnosis of RA (≥1 inpatient claim or ≥2 outpatient (OP) medical claims on 2 different dates with an ICD-9-CM/ICD-10-CM diagnosis code for RA in any position during 12 months prior to the index event) were eligible for the study. The study cohorts consisted of (1) patients who were TNFi-experienced and switched to a subsequent different tDMARD treatment (subsequent tDMARD claim served as an index), and (2) patients who were tDMARD naïve and initiated either abatacept, TNFi or other non-TNFi as their first tDMARD (first tDMARD claim served as an index). TNFi’s included adalimumab, certolizumab pegol, etanercept, golimumab, and infliximab, while other non-TNFi’s included anakinra, sarilumab, tocilizumab, baricitinib, rituximab, and tofacitinib. Intravenously administered drugs were identified via Part B medical claims whereas orals and subcutaneously administered drugs were identified via Part B and D pharmacy claims. We used the Medicare enrollment data and patients with Medicare Advantage (Part C) enrollment indicator were excluded. Patients with medical or pharmacy claim for more than one type of tDMARD on the index date and diagnosis of autoimmune conditions (i.e. Crohn’s disease, ulcerative colitis, plaque psoriasis, psoriatic arthritis, ankylosing spondylitis, chronic lymphocytic leukemia, juvenile idiopathic arthritis, Wegener’s granulomatosis, non-Hodgkin’s lymphoma, and polyarteritis nodosa) at any time during the baseline were excluded. Furthermore, patients with evidence of cancer during baseline or follow-up period or prescription data validity issues (e.g. invalid or missing days’ supply that could not be easily imputed from available information) from pharmacy claims were excluded. Among tDMARD naïve RA patients, beneficiaries with use of any tDMARD within 12 months before the index date were excluded from the study. Both TNFi-experienced and tDMARD naïve groups were followed from index date for a period of maximum 36 months and ended at the earliest of the following: (1) patient disenrollment; (2) end of study period; (3) discontinuation or switch of index treatment; (4) death during follow-up; (5) 36 months post-index. Discontinuation was defined as a 2 times gap in days supply from end of day’s supply of index drug. Patients who discontinued the index drug, last date of follow-up was the last day of supply for index drug + 90 days (index drug + days supply + 90 days). Patients who switched to another tDMARDs during follow-up, last date of follow-up was the drug switching date.

Study outcomes

Infection-related hospitalization risk and costs among TNFi-experienced and tDMARD naïve RA patient cohorts were the primary endpoints. Further, infection-related costs were further broken down into following infection types: 1) pneumonia; 2) bacterial respiratory infection; 3) sepsis; 4) skin and soft tissue infection; 5) joint infection; 6) genitourinary infection. The medical services related to these infection types were defined as any inpatient medical services with at least one diagnosis code for these infection types in any position of the inpatient claims. To account for the variable follow-up period, direct infection-related hospitalization costs outcomes were evaluated on a per-patient-per-month (PPPM) basis. All costs were inflation adjusted to 2019 cost levels using the Consumer Price Index (CPI) medical care component. We adopted the payer perspective for estimating costs. Rate of infection-related hospitalization were reported as per 1000 patients per month (P1000PPM).

Statistical analysis

Patient profiles for included patients were evaluated during the baseline period for the study cohorts. Baseline demographics and patient characteristics including age, gender, US geographic region, Charlson Comorbidity Index (CCI) score, previous glucocorticoid use, history of infection, and comorbid conditions were collected. Descriptive statistics were used to evaluate differences in patient demographics, clinical characteristics, healthcare resource use (HCRU), and costs for the study cohorts. A p value of <.05 was considered statistically significant. For this study, infection-related hospitalization risk and costs were reported by tDMARD therapies (i.e. Abatacept, TNFi, and other non-TNFi) for each study cohort. Outcomes were compared between tDMARD treatments by using descriptive statistics (unadjusted) as well as multivariable regression analyses (adjusted). A two-part generalized linear model with a log link function and gamma distribution was used to evaluate potential differences in the costs associated with infection-related hospitalizations. The adjusted risk of infection-related hospitalization was evaluated using Cox proportional hazard model. Hazard ratios (HR) were compared between the abatacept, TNFi, and other non-TNFi therapies.

Sensitivity analysis

To test for the robustness of the results to the exclusion of patients with other autoimmune diseases (i.e. Crohn’s disease, ulcerative colitis, plaque psoriasis, psoriatic arthritis, ankylosing spondylitis, chronic lymphocytic leukemia, juvenile idiopathic arthritis, Wegener’s granulomatosis, non-Hodgkin’s lymphoma, and polyarteritis nodosa), sensitivity analysis was conducted in which patients with other autoimmune diseases were included. We evaluated the impact of allowing patients with autoimmune conditions on the infection-related hospitalization risk and costs during follow-up.

Results

Cohort selection and baseline characteristics

Among TNFi-experienced cohort, a total of 16,647 Medicare FFS patients met the study criteria, of whom 6,343 patients switched to abatacept, 5,054 switched to another TNFi, and 5,250 switched to other non-TNFi (Supplemental Figure 1). Among tDMARD naïve cohort, a total of 30,439 Medicare FFS patients met the study criteria, of whom 6,303 patients initiated abatacept, 18,032 patients initiated a TNFi, and 6,104 initiated other non-TNFi (Supplemental Figure 1). Patient demographic and clinical characteristics among both TNFi-experienced and tDMARD naïve cohorts are shown in Table 1. For both TNFi-experienced and tDMARD naïve cohorts, abatacept (>90%) and other non-TNFi (>80%) were primarily administrated by intravenous route, while TNFi (>50%) were administered through subcutaneous route. Additionally, majority of the population were white and females located in the southern region of the USA with the average age around 73 years. Overall, the mean Charlson Comorbidity Scores (CCI) were non-differential (between 4 and 5) through all tDMARD groups. Additionally, majority of the RA patients had comorbid conditions for hypertension (>77%) and about a quarter of patient cohorts had diabetes (Table 1).

Table 1. Baseline demographic and clinical characteristics among TNFi-experienced and tDMARD naïve patient cohorts.

Patient characteristics	TNFi-experienced patient cohort								tDMARD naïve patient cohort
	Abatacept (N = 6,343)		TNFi (N = 5,054)		p Value	Other non-TNFi (N = 5,250)		p Value	Abatacept (N = 6,303)		TNFi (N = 18,032)		p Value	Other non-TNFi (N = 6,104)		p Value
Variable	N	%	N	%		N	%		N	%	N	%		N	%
Age (in years)
Mean	73.0		73.1		.25	72.7		.02	73.6		72.8		<.0001	73.1		<.0001
SD	5.6		6.0			5.6			5.9		5.6			5.7
Median	72.0		72.0			72.0			72.0		72.0			72.0
Gender
Female	5,425	85.5%	4,178	82.7%	<.0001	4,323	82.3%	<.0001	5,400	85.7%	14,645	81.2%	<.0001	4,976	81.5%	<.0001
Male	918	14.5%	876	17.3%		927	17.7%		903	14.3%	3,387	18.8%		1,128	18.5%
Region
South	2,833	44.7%	2,294	45.4%	<.0001	2,150	41.0%	<.0001	3,014	47.8%	7,906	43.8%	<.0001	2,231	36.5%	<.0001
Midwest	1,496	23.6%	990	19.6%		1,082	20.6%		1,277	20.3%	4,148	23.0%		1,493	24.5%
Northeast	894	14.1%	835	16.5%		856	16.3%		936	14.9%	2,645	14.7%		1,066	17.5%
West	1,114	17.6%	933	18.5%		1,160	22.1%		1,061	16.8%	3,321	18.4%		1,312	21.5%
Race/Ethnicity
White	5,585	88.0%	4,360	86.3%	<.0001	4,526	86.2%	.05	5,406	85.8%	15,618	86.6%	.01	5,209	85.3%	.27
Black or African American	365	5.8%	326	6.5%		360	6.9%		458	7.3%	1,272	7.1%		473	7.7%
Other	393	6.2%	368	7.2%		364	6.9%		439	7.0%	1142	6.3%		422	7.0%
Charlson Comorbidity Score
Mean	4.9		4.8		.01	5.0		.14	4.5		4.1		<.0001	4.5		.42
SD	3.0		3.0			3.0			2.8		2.7			2.8
Median	4.0		4.0			4.0			4.0		3.0			4.0
Other comorbidities
Congestive heart failure	799	12.6%	443	8.8%	<.0001	621	11.8%	.21	930	14.8%	1,801	10.0%	<.0001	979	16.0%	.05
Hypertension	4,869	76.8%	3,905	77.3%	.53	4,053	77.2%	.58	4,985	79.1%	13,919	77.2%	<.0001	4,752	77.9%	.09
COPD	1,566	24.7%	1,188	23.5%	.14	1,362	25.9%	.12	1,638	26.0%	4,436	24.6%	.03	1,704	27.9%	.02
Chronic liver disease	611	9.6%	386	7.6%	<.0001	462	8.8%	.12	622	9.9%	1,510	8.4%	<.0001	569	9.3%	.3
Neutropenia	70	1.1%	33	.7%	.01	61	1.2%	.77	70	1.1%	138	0.8%	.01	113	1.9%	<.0001
Renal disease	937	14.8%	699	13.8%	.15	822	15.7%	.19	1,076	17.1%	2,395	13.3%	<.0001	1,044	17.1%	.96
Diabetes	1,820	28.7%	1,404	27.8%	.28	1,527	29.1%	.64	1,993	31.6%	5,143	28.5%	<.0001	1,939	31.8%	.86
Infections – distribution by infection type
Genitourinary	1,748	27.6%	1,320	26.1%	.09	1,511	28.8%	.14	285	4.5%	585	3.2%	<.0001	318	5.2%	.07
Joint	75	1.2%	55	1.1%	.64	67	1.3%	.65	25	0.4%	59	0.3%	.42	26	0.4%	.8
Pneumonia	690	10.9%	407	8.1%	<.0001	586	11.2%	.63	209	3.3%	441	2.4%	<.0001	299	4.9%	<.0001
Respiratory	128	2.0%	74	1.5%	.03	107	2.0%	.94	21	0.3%	26	0.1%	<.0001	16	0.3%	.47
Sepsis	247	3.9%	126	2.5%	0	182	3.5%	.22	118	1.9%	248	1.4%	.01	165	2.7%	<.0001
Skin	599	9.4%	438	8.7%	.15	535	10.2%	.18	285	4.5%	585	3.2%	<.0001	318	5.2%	.07
Administration route for index tDMARD
Intravenous	5,817	91.7%	1,186	23.5%	<.0001	4,328	82.4%	<.0001	5,849	92.8%	8,435	46.8%	<.0001	5,471	89.6%	<.0001
Oral	–	–	–	–		676	12.9%		0	0.0%	0	0.0%		500	8.2%
Subcutaneous	526	8.3%	3,868	76.5%		246	4.7%		454	7.2%	9,597	53.2%		133	2.2%
Conventional DMARDs (cDMARDs)
Hydroxychloroquine	1,126	17.8%	838	16.6%	.02	896	17.1%	.03	1,862	29.5%	4,872	27.0%	<.0001	1,395	22.9%	<.0001
Other cDMARDS	4,631	73.0%	3,789	75.0%		3,924	74.7%		4,881	77.4%	15,190	84.2%		4,135	67.7%
Other medications
Glucocorticoids	5,199	82.0%	3,846	76.1%	<.0001	4,393	83.7%	.02	5,160	81.9%	14,667	81.3%	.35	5,030	82.4%	.43

Abbreviations. COPD, chronic obstructive pulmonary disease; N, number of patients; SD, standard deviation; TNFi, tumor necrosis factor inhibitor.

Infection-related hospitalizations at baseline and follow-up

Among TNFi-experienced cohort, infection-related hospital visits at baseline were significantly higher in those who received abatacept at index (9.8%) in comparison to patients who received another TNFi (7.1%, p < .0001), however, they were non-differential when compared to patients who received other non-TNFi (9.7%, p = .77) (Table 2). During baseline, a significant number of patients had infections related to genitourinary (TNFi-experienced: 26–29%; tDMARD naïve: 3–5%), pneumonia (TNFi-experienced: 8–11%; tDMARD naïve: 2–5%), and skin-related (TNFi-experienced: 9–10%; tDMARD naïve: 3–5%) (Table 1). During follow-up, infection-related hospital visits were significantly lower in those who received abatacept (7.5%) in comparison to patients who received other non-TNFi (10.7%, p < .0001), however, they were non-differential when compared to patients who received another TNFi (8.1%, p = .21) (Table 2). Average number of infection-related hospital visits P1000PPM were also lower in abatacept users (6.5) in comparison to TNFi (8.0; p = .21) and other non-TNFi (9.7; p = .05) groups (Table 2). Additionally, only patients who received abatacept have a decline in infection frequency during follow-up period compared to baseline. Percentage of patients with infection-related hospitalization varied by infection type. During both baseline and follow-up, majority of infection-related hospitalization visits were for genitourinary, pneumonia, or skin-related infection types (Figure 1).

Figure 1. TNFi-experienced patient cohort: distribution of percent patients with infection-related hospital visits at baseline and follow-up by infection type. *indicate values were significant at <0.0001 level of significance. Note: The infection sites were not mutually exclusive. Patients may have experienced more than one type of infection during the hospital stay. Patients with multiple infections during follow-up were counted only once for the estimation of the total percent.

Table 2. Infection-related hospital visits at baseline and follow-up among TNFi-experienced and tDMARD naïve patient cohorts.

	TNFi-experienced patient cohort					tDMARD naïve patient cohort
Cohort name	N	Patients with infection-related hospital visit; N (%)	p Value	Number of infection-related hospital visits per 1,000 patients per month		N	Patients with infection-related hospital visit; N (%)	p Value	Number of infection-related hospital visits per 1,000 patients per month
				Mean (STD)	p Value				Mean (STD)	p Value
Baseline
Abatacept	6,343	623 (9.8%)	–	10.3 (35.0)		6,303	543 (8.6%)	–	9.5 (35.2)
TNFi	5,054	360 (7.1%)	<.0001	6.9 (27.0)	<.0001	18,032	1,128 (6.3%)	<.0001	6.5 (28.5)	<.0001
Other Non-TNFi	5,250	507 (9.7%)	.77	10.1 (34.6)	.8	6,104	655 (10.7%)	<.0001	11.3 (36.4)	<.0001
Follow-up
Abatacept	6,343	476 (7.5%)	–	6.5 (31.3)		6,303	507 (8.0%)	–	6.4 (31.7)
TNFi	5,054	411 (8.1%)	.21	8 (36.8)	.21	18,032	1,532 (8.5%)	.26	7.8 (34.7)	.52
Other Non-TNFi	5,250	561 (10.7%)	<.0001	9.7 (37.6)	.05	6,104	743 (12.2%)	<.0001	11.9 (40.5)	.01

Among tDMARD naïve cohort, a similar trend of reduced infection-related hospitalization was observed during follow-up in the abatacept arm. Infection-related hospital visits at baseline were significantly lower in those who received abatacept at index (8.6%) in comparison to patients who received non-TNFi (10.7%, p < .0001), however, they were significantly higher when compared to patients who received TNFi (6.3%, p < .0001) (Table 2). Similar to TNFi-experienced cohort, percentage of patients with infection-related hospitalization varied by infection type. During both baseline and follow-up, majority of infection-related hospitalization visits were for genitourinary, pneumonia, or skin-related infection types (Figure 2). During follow-up, infection-related hospital visits were lower in those who received abatacept (8.0%) in comparison to patients who received a TNFi (8.5%, p = .26) or other non-TNFi (12.2%, p < .0001) (Table 2). Average number of infection-related hospital visits P1000PPM were also lower in abatacept users (6.4) in comparison to TNFi (7.8; p = .52) and other non-TNFi (11.9; p = .01) groups (Table 2).

Figure 2. tDMARD naïve patient cohort: distribution of percent patients with infection-related hospital visits at baseline and follow-up by infection type. *indicate values were significant at <0.0001 level of significance. Note: The infection sites were not mutually exclusive. Patients may have experienced more than one type of infection during the hospital stay. Patients with multiple infections during follow-up were counted only once for the estimation of the total percent.

Infection-related hospitalization costs during follow-up

Among TNFi-experienced cohort, during follow-up, the direct PPPM costs associated with infection-related hospitalization were lower for abatacept users in comparison to TNFi and other non-TNFi users (Table 3). After controlling for potential confounding variables, infection-related hospitalization costs PPPM were significantly higher in patients treated with TNFi’s (cost difference: $74; 95% CI: $44–$105) and other non-TNFi’s (cost difference: $68; 95% CI: $44–$92) than in those treated with abatacept (Table 3). Furthermore, the PPPM cost was lower in follow-up than baseline periods for abatacept users, but the opposite for TNFi and other non-TNFi users (Table 3).

Table 3. Comparison of infection-related hospitalization costs by tDMARD therapies among TNFi-experienced and tDMARD naïve patient cohorts.

	Baseline	Follow-up			Sensitivity analysis
	Unadjusted infection-related hospitalization mean PPPM cost	Unadjusted infection-related hospitalization mean PPPM cost	Adjusted^a infection-related hospitalization mean PPPM cost (95%CI)	PPPM cost difference	Follow-up adjusted^a infection-related hospitalization mean PPPM cost (95%CI)
TNFi-experienced
Abatacept	$131	$77	$73 ($60–$85)	Reference	$73 ($61–$85)
TNFi	$83**	$108	$147 ($120–$174)	$74***	$145 ($120–$170)
Other Non-TNFi	$112	$141**	$141 ($119–$163)	$68***	$139 ($117–$161)
tDMARD Naïve
Abatacept	$128	$86	$75 ($62–$88)	Reference	$77 ($64–$91)
TNFi	$79**	$104	$118 ($106–$130)****	$42***	$122 ($110–$134)
Other Non-TNFi	$185	$165**	$135 ($117–$154)****	$60***	$140 ($122–$159)

Abbreviation. PPPM, per patient per month.

^a A two-part generalized linear model with a log link function and gamma distribution was used to evaluate potential differences in the cost associated with infection-related hospitalizations among the corresponding RA patients.

** indicates significance at .05 level, ***.001 and ****<.0001 p values were evaluated between abatacept and TNFis as well as between abatacept and other non-TNFis.

Similar to the TNFi-experience cohort, among the tDMARD naïve cohort, during follow up, the direct PPPM costs associated with infection-related hospitalization were lower for abatacept users in comparison to TNFi and other non-TNFi users (Table 3). After controlling for potential confounding variables, infection-related hospitalization costs PPPM were significantly higher in patients treated with TNFi’s (cost difference: $42; 95% CI: $25–$59) and other non-TNFi’s (cost difference: $60; 95% CI: $38–$81) than in those treated with abatacept (Table 3). The cost was lower in follow-up than in baseline periods for abatacept and other non-TNFi users, but the opposite for TNFi users (Table 3).

Sensitivity analysis gave similar results as the original study analysis for the TNFi-experienced and tDMARD naïve cohorts. The PPPM infection-related hospitalization costs was the lowest for abatacept users in comparison to TNFi and other non-TNFi users (Table 3).

Risk for infection-related hospitalization

Among both TNFi-experienced and tDMARD naïve cohorts, after adjusting for differences in patient demographic and clinical characteristics, Cox regression analysis showed that the risk for an infection-related hospitalization was significantly higher for RA patients treated with TNFi’s (HR for TNFi-experienced: 1.48, 95% CI: 1.26–1.75, p < .0001; HR for tDMARD naïve: 1.59, 95% CI: 1.4–1.8, p < .0001) and other non-TNFi’s (HR for TNFi-experienced: 1.46, 95% CI: 1.28–1.66, p < .0001; HR for tDMARD naïve: 1.63, CI: 1.4–1.8, p < .0001) than in those treated with abatacept (Figure 3). In the unadjusted analysis, TNFi (HR: 1.18, 95% CI: 1.03–1.37) and non-TNFi users (HR: 1.40, 95% CI: 1.22–1.60) experienced a higher risk of infection-related hospitalization compared to abatacept users in the TNFi-experienced cohort. Baseline infection-related hospitalization (HR: 1.42, 95% CI: 1.23–1.64, p < .0001), CCI score of 5+ (HR: 2.59, 95% CI: 2.13–3.14, p < .0001), and glucocorticoid use (HR: 1.31, 95% CI: 1.12–1.53, p = .0007) were independent predictors of infection-related hospitalization. Both the unadjusted and adjusted findings were consistent for the naïve cohort. Sensitivity analysis showed the results to be consistent and robust across both study cohorts (Figure 3).

Figure 3. Comparison of adjusted infection-related hospitalization risk by tDMARD therapies among TNFi-experienced and tDMARD naïve patient cohorts. Note: HR were estimated by comparing TNFi and other non-TNFi to abatacept users i.e abatacept group served as a reference. † indicate values were significant at <0.0001 level of significance.

Discussion

Rheumatoid arthritis is one of the most common autoimmune disease. RA has been previously shown to be associated with increased risk of infection¹⁵. Patients with RA have nearly twice the rate of infection compared to the general population¹⁵. Infection-related hospitalizations are a significant economic burden in the Medicare population¹⁶. Therefore, it is important to understand the clinical and economic burden associated with infection-related hospitalizations among RA Medicare FFS patients treated with tDMARDs. To our knowledge, our real-world study is the first study to use 100% sample of Medicare FFS beneficiaries to assess the risk and costs of infection-related hospitalization in RA patients who were TNFi-experienced or tDMARD naïve.

Among TNFi-experienced cohort, abatacept users had a higher percentage of patients with infection-related hospitalization in the baseline compared to TNFi users. However, during follow-up, for both cohorts, patients in abatacept group had a lower percentage of patients with infection-related hospitalization than patients in either the TNFi. Our study is consistent with previous studies that have shown a lower rate of infection-related hospitalizations for patients treated with abatacept compared to TNFis and other non-TNFis in the biologic DMARD (bDMARD)-experienced Medicare patients population^13,17. The cost of infection-related hospitalization PPPM cost was lower in patients treated with abatacept compared to TNFi. Further, the risk for an infection-related hospitalization following initiation or switch of abatacept was significantly lower in comparison to patients who initiated or switched to TNFi. The cost of infection-related hospitalization PPPM cost was lower in patients treated with abatacept compared to TNFi.

Some limitations associated with this study and observational studies in general need to be acknowledged. Claims data exist mainly for billing and reimbursement purposes thus, there is a possibility for errors in documentation of medical conditions and outcomes which may lead to patient misclassification either due to miscoding or misdiagnosis. As claims data only captures those disease entities and variables that have their own specific billing codes, this study cannot be used to determine cause and effect. In comparison to cross-sectional and case–control study design, this retrospective cohort study will have higher internal validity. As for all observational studies, treatments are prescribed on the basis of clinical judgment. Therefore, comparisons of patients on different treatment regimens will likely be confounded because there are differences in disease severity and risk of the events of interest (e.g. patients receiving one tDMARD abatacept are likely to be different in many ways from patients receiving other tDMARDs). Potential confounding variables were controlled for via appropriate study design and statistical modelling. However, the possibility of residual confounding from unmeasured factors cannot be excluded. Conclusions regarding the differential risk in serious infections may not be generalizable to specific mechanism of action (MOAs) within the non-TNFi arm. Rituximab and tocilizumab accounted for majority of the treatment within the non-TNFi arm; tofacitinib was used in less than 13% of patients.

Even though there are some limitations to the study, there are several beneficial aspects of the use of data from a large, nationally representative US Medicare claims database. Overall, the dataset is comprehensive because it incorporates all medical and pharmacy claims of Medicare FFS patients and allows for longitudinal analysis of a large US patient sample. A major key strength of retrospective database studies is that it allows observation of patients who are often under-represented in clinical trials (e.g. those with comorbidities and the elderly population). Since prescribing patterns in the real-world are broader and less limiting, the retrospective analysis provides a more comprehensive picture of how medications are used by clinicians in routine practice and the adherence of treatments. Many of the medications being studied are relatively new to the market, and this database captures the utilization of these newer drugs. Our study provides a better understanding of the RA population in real-world clinical practice as compared to the controlled conditions of a clinical trial.

Overall, among both TNFi-experienced and tDMARD naïve Medicare FFS beneficiaries with RA, patients who initiated or switched to abatacept had lower risk and cost of infection-related hospitalization in comparison to patients who initiated or switched to TNFi’s or other non-TNFi’s. Results from this study favors use of abatacept in comparison to other tDMARDs such as TNFis as it can potentially help in reducing the clinical and economic burden associated with infection-related hospitalization in Medicare patients with RA. Future research in needed to compare abatacept with each drug within the non-TNFi arm at the mechanism of action level (Janus kinase inhibitors, B-cell depleting agent, interleukin-6 inhibitor).

Transparency

Declaration of funding

This research has been funded by Bristol Myers Squibb.

Declaration of financial/other relationships

The analysis was conducted independently by Avalere Health. BMS and Avalere Health collaborated on study design and interpretation of results. Vardhaman Patel and Francis Lobo are BMS employees and shareholders. Zulkarnain Pulungan, Anne Shah, Mahesh Kambhampati, and Allison Petrilla are employees of Avalere Health. All authors contributed towards the study design, analysis, interpretation of results and manuscript writing. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Author contributions

VP designed the study, interpreted the results and supported the manuscript writing. ZP, AS and MK assisted with study design, conducted the analysis and interpreted the results. AS supported the manuscript writing. FL interpreted the results and supported the manuscript writing. AP supervised the study design, analysis, interpretation of results and manuscript content development.

Previous presentations

A portion of these results were presented at the 2019 American College of Rheumatology (ACR)/Association of Rheumatology Professionals (ARP) Annual Meeting; 2019 Nov 11; Atlanta, GA.

Acknowledgements

This study was sponsored by Bristol Myers Squibb. Bristol Myers Squibb participated in the interpretation of data, review and approval of publication. Anny Wong and Sudeepti Puhuja provided research support, data interpretation, and manuscript writing for the study.

Data availability statement

The data described in this paper are sourced from CMS Medicare Fee-for-Service claims and enrollment data. The analytic file constructed for this analysis cannot be shared due to protections of patient privacy. Researchers may request use of CMS data through ResDAC.