Article Text

Extended report
Comparative effects of biologics on cardiovascular risk among older patients with rheumatoid arthritis
  1. Jie Zhang1,
  2. Fenglong Xie2,
  3. Huifeng Yun1,2,
  4. Lang Chen2,
  5. Paul Muntner1,
  6. Emily B Levitan1,
  7. Monika M Safford3,
  8. Shia T Kent1,
  9. Mark T Osterman4,
  10. James D Lewis4,
  11. Kenneth Saag2,
  12. Jasvinder A Singh2,5,6,
  13. Jeffrey R Curtis1,2
  1. 1Department of Epidemiology, School of Public Health, University of Alabama at Birmingham, Birmingham, Alabama, USA
  2. 2Division of Clinical Immunology and Rheumatology, School of Medicine, University of Alabama at Birmingham, Birmingham, Alabama, USA
  3. 3Division of Preventive Medicine, School of Medicine, University of Alabama at Birmingham, Birmingham, Alabama, USA
  4. 4Division of Gastroenterology, Center for Clinical Epidemiology and Biostatistics; Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania, USA
  5. 5Medicine Service, Birmingham Veterans Affairs Medical Center, Birmingham, Alabama, USA
  6. 6Department of Orthopedic Surgery, Mayo Clinic College of Medicine, Rochester, New York, USA
  1. Correspondence to Dr Jeffrey R Curtis, Department of Immunology and Rheumatology, University of Alabama at Birmingham, FOT 802, 510 20th Street South, Birmingham, AL 35294, USA; jcurtis{at}uab.edu

Abstract

Objectives To compare the coronary heart disease risk among patients with rheumatoid arthritis (RA) initiating common biologic disease-modifying antirheumatic drugs of different mechanisms.

Methods We conducted a retrospective cohort study of patients with RA enrolled in Medicare, a public health plan covering >90% of US residents 65 years or older, from 2006 to 2012 who (1) initiated a biologic, (2) had complete medical and pharmacy coverage for at least 12 months before biologic initiation and (3) were free of coronary heart disease at the time of initiation. We compared the incidence rates (IRs) of (1) acute myocardial infarction (AMI) and (2) a composite outcome of AMI or coronary revascularisation and used multivariable adjusted Cox regression models to examine the associations between the type of biologic and the two outcomes.

Results We identified 47 193 eligible patients with RA with mean age 64 (SD 13) years; 85% were women. Crude IRs for AMI ranged from 5.7 to 8.8 cases per 1000 person-years (PYs). AMI risk was significantly elevated among antitumour necrosis factor (anti-TNF) initiators overall (adjusted HR (aHR) 1.3; 95% CI 1.0 to 1.6) and individually among etanercept (aHR 1.3; 95% CI 1.0 to 1.8) and infliximab (aHR 1.3; 95% CI 1.0 to 1.6) compared with abatacept initiators. Crude IRs for the composite outcome ranged from 7.6 to 14.5 per 1000 PYs. Tocilizumab initiators were at reduced risk of the composite outcome compared with abatacept initiators (aHR 0.64, 95% CI 0.41 to 0.99).

Discussion Findings from this observational study of patients with RA suggested that anti-TNF biologics may be associated with higher AMI risk compared with abatacept.

  • Rheumatoid Arthritis
  • DMARDs (biologic)
  • Cardiovascular Disease
View Full Text

Statistics from Altmetric.com

Introduction

Patients diagnosed with rheumatoid arthritis (RA) are at increased risk of cardiovascular (CV) disease and CV mortality compared with the general population without RA.1–3 The increased CV risk is thought to be largely attributable to systemic inflammation, and anti-tumour necrosis factor α (anti-TNF) biologics significantly reduce inflammation in these patients. Many studies have therefore examined the impact of anti-TNF biologics on CV risk in patients with RA. While the conclusions from these studies are not entirely consistent, most studies have found anti-TNF biologics to be associated with no change4 or reduced CV risk,5–7 with one study reporting reduced CV risk only in patients who responded clinically to anti-TNF biologics.8 Additionally, emerging evidence has suggested favourable changes in low-density lipoprotein cholesterol (LDL-C) subfractions and vascular and atherogenic profiles (eg, the arterial stiffness marker pulse wave velocity) associated with anti-TNF therapies among patients with RA.9 ,10

There is limited information on the association between non-TNF biologics and CV risk, including rituximab (an anti-CD20 monoclonal antibody), abatacept (a selective inhibitor of T-cell co-stimulation) and tocilizumab (an interleukin (IL) 6 inhibitor). In small studies, rituximab (12 patients) was associated with greater reduction in serum levels of prothrombotic molecules (fibrinogen, D-dimer, tissue plasminogen activator (tPA) levels) compared with infliximab (10 patients with RA).11 Another study of 49 patients with RA showed that rituximab was associated with an overall improvement in the atherogenic index compared with that before treatment initiation, especially among clinical responders.12 In contrast, a study of 21 patients reported worsening aortic stiffness after 6 months of treatment with abatacept.13 CV risk has been a major safety concern for tocilizumab, which has been shown to significantly increase LDL-C levels shortly after treatment initiation; yet, tocilizumab has not been associated with increased CV event risk in randomised controlled trials or postmarketing studies.14 ,15

In light of the uncertainties concerning the potential impact of non-TNF biologics on CV risk among patients with RA, we conducted a retrospective cohort study to examine the comparative CHD event risk among older patients with RA enrolled in Medicare and initiating one of the currently approved biologics for RA.

Methods

We conducted a retrospective cohort study among Medicare beneficiaries diagnosed with RA using medical and pharmacy claims data from 1 January 2006 through 31 December 2012. Medicare is a public health plan which covers the elderly (≥65 years of age) and disabled populations in the USA. Because RA is a reason that Medicare allows individuals to qualify as disabled, the study population includes some patients younger than 65. Eligible patients were required to have: (1) ≥2 diagnosis codes for RA from a physician that were between 7 days and 12 months apart; (2) ≥12 months of continuous enrolment in Medicare Part A (inpatient care), Part B (physician services), and Part D (outpatient prescription drug coverage) and not be enrolled in a Medicare Advantage Plan; (3) initiated an anti-TNF biologic (adalimumab, certolizumab, etanercept, golimumab, infliximab) or any of the three non-TNF biologics (abatacept, rituximab, tocilizumab) and (4) no history of coronary heart disease using a validated claims-based algorithm (see online supplementary table S1).16 These requirements aimed to identify patients with RA with a complete claim history for services received in inpatient, outpatient and emergency department settings; and prescriptions filled in outpatient settings; to allow for a new user design that reduces bias and to capture the occurrence of incident CHD outcomes. Eligible participants did not have to be biologic naïve but they must not have been exposed to the same biologic agent in the ≥12 months preceding biologic initiation. They were also allowed to contribute multiple episodes of biologic initiations.

Follow-up started on the date (defined as the index date) when a patient initiated the biologic. The 12 months immediately prior to the index date was defined as the baseline period. Follow-up ended when the patient discontinued the biologic or switched to another biologic; an outcome (described below) occurred; the patient lost Part A, Part B or Part D coverage or entered a Medicare Advantage plan; died; or on December 31, 2012.

Outcome assessment

The outcomes were hospitalised acute myocardial infarction (AMI) and a composite outcome consisting of AMI, percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG). AMI was identified using the following algorithm: (1) at least one inpatient claim with discharge ICD-9 diagnoses (any position) for AMI (410 excluding 410.x2) and (2) at least one night of inpatient stay except if the patient died. The validity of AMI algorithms has been evaluated in prior studies with high positive predictive values that exceed 90%.17 PCI was identified by ≥1 Current Procedural Terminology (CPT) codes 92980–92996 or ICD-9 procedure codes 00.66 or 36.01–36.09 from inpatient, outpatient, revenue centre or carrier line file claims. CABG was identified by CPT codes 33510–33536 and ICD-9 procedure codes 36.10–36.19 from inpatient, outpatient, revenue centre or carrier line file claims.

Medication exposure and covariates

Exposure to biologics and non-biologic disease-modifying antirheumatic drugs (DMARDs) (methotrexate, hydroxychloroquine, sulfasalazine, azathioprine, leflunomide, ciclosporin and 6-mercaptopurine) and oral glucocorticoids was determined based on prescription date and days of supply, and for infusions, the recommended dosing intervals. For example, a patient who received an infliximab infusion was considered exposed to infliximab for the next 56 days from the infusion date. Average daily dose of glucocorticoids was calculated for each 6-month interval and was classified as none, low dose (<7.5 mg/day) or higher dose (≥7.5 mg/day).

Other covariates measured in the baseline period and included in the multivariable Cox regression model included age, sex, race, original reason for Medicare enrolment (old age or disability), receipt of subsidised Medicare premium (a surrogate for low income), CV risk factors (diabetes, hypertension, chronic kidney disease, abdominal aortic aneurysm, peripheral arterial disease, atrial fibrillation, hyperlipidaemia, diagnosis or treatment for tobacco use, obesity), other comorbid diseases (heart failure, chronic obstructive pulmonary disorder (COPD)) and use of CV medications (antihypertensive medications categorised into ACE inhibitors, β blockers, and other; statins; non-steroidal anti-inflammatory drugs (NSAIDs)).

Statistical analysis

We computed descriptive statistics, including means, SDs and numbers (%), to characterise the patient population. We then calculated the incidence rates (IRs) of AMI and the composite CHD outcome by dividing the number of incident cases by the number of corresponding person-years (PYs) stratified by the type of biologic initiated. Proportional hazard regression was used to assess the relative CHD risk associated with each biologic, adjusting for covariates specified above, and stratified by number of prior biologics (0, 1 or ≥2). In addition to comparing the effect of each individual biologic on CHD risk, we performed additional analyses that compared anti-TNF biologics to non-TNF biologics. Patients who sequentially initiated different biologics were permitted to contribute to multiple exposure groups and standard errors were adjusted to account for correlated data from these patients. We created forest plots to present adjusted HRs and used the Kaplan–Meier method to plot outcome-free survival by the type of biologic. The proportional hazard assumption was tested. All analyses were performed using SAS V.9.3.

We performed several subgroup analyses to examine the association between biologic initiated and the risk of AMI among patients who were biologic naïve (no filled prescription for biologics in the baseline period), who had one prior biologic, and who used two or more biologics. To address the concern that we may have missed cases of sudden death due to CV causes, we performed an analysis modelling the risk of a composite outcome of acute MI or death, whichever occurred first. The Institutional Review Board at the University of Alabama at Birmingham approved the study protocol and waived the requirement for informed consent. Use of the Medicare data was governed by a Data Use Agreement (DUA) with the Centre for Medicare and Medicaid Services. As required by the DUA, any cell containing less than 11 beneficiaries was not shown.

Figure 1

Kaplan–Meier plot of acute myocardial infarction-free survival.

Results

We identified a total of 62 087 episodes of biologic initiations among 47 193 patients with RA who met inclusion criteria and initiated one of the eight prespecified biologic DMARDs, ranging from 1774 who initiated golimumab to 13 608 who used abatacept. Approximately 76% of patients contributed only one episode to the study. Their average age was 64 (SD 13) years, 85% were women and the average length of follow-up was 447 (SD 440) days. Table 1 presents the distribution of select demographic and clinical characteristics stratified by the biologic initiated. Patients who initiated injection biologics were younger, less likely to be white and more likely to be disabled, compared with those who initiated infusion biologics. Detailed data on the clinical characteristics of the patients are shown in online supplementary table S2.

Table 1

Select patient characteristics during baseline or at start of follow-up

The crude AMI IRs by specific biologic initiated ranged from 5.7 (golimumab) to 8.8 (infliximab) cases per 1000 PYs during follow-up and were not significantly different from one another (table 2). After multivariable adjustment, etanercept (HR 1.3; 95% CI 1.0 to 1.8) and infliximab (HR 1.3; 95% CI 1.0 to 1.6) were associated with elevated risk of AMI compared with abatacept (reference group) (figure 2A). When all anti-TNF biologics were grouped together, anti-TNF biologic initiators were at 28% higher risk of AMI compared with abatacept initiators (figure 2B).

Table 2

Number of events and crude incidence rate by different types of biologic exposure among patients with RA for MI and composite CHD event

Figure 2

Adjusted HRs for (A) acute myocardial infarction comparing individual biologics to Abatacept; (B) acute myocardial infarction comparing biologics with different mechanisms to Abatacept; (C) composite coronary heart disease (CHD) outcome comparing individual biologics to Abatacept; and (D) composite CHD outcome comparing biologics with different mechanisms to Abatacept. (A–D) Adjusted for age, gender, race, original reason for Medicare enrolment (old age or disability), receipt of subsidy for Medicare premium (surrogate for low income), cardiovascular risk factors (diabetes, hypertension, chronic kidney disease, abdominal aortic aneurism, peripheral arterial disease, atrial fibrillation, hyperlipidaemia, diagnosis or treatment of or treatment for tobacco use, obesity), other comorbid diseases (heart failure, chronic obstructive pulmonary disorder, use of cardiovascular medications (antihypertensive medications categorised into ACE inhibitors, β blockers and other; statins; prescription non-steroidal anti-inflammatory drugs), and exposure to non-biologic disease-modifying antirheumatic drugs and oral glucocorticoids. LCL, lower confidence limit; TNF, tumour necrosis factor; UCL, upper confidence limit.

The IRs of the composite CHD outcome ranged from 7.6 (golimumab) to 14.5 (infliximab) cases per 1000 PYs; the IR among tocilizumab (8.1 per 1000 PYs; 95% CI 5.31 to 12.25) initiators was significantly lower than that among infliximab initiators (14.5; 95% CI 12.86 to 16.26 per 1000 PYs) (table 2). After adjusting for covariates, tocilizumab was associated with 36% lower risk of experiencing the composite CHD outcome compared with abatacept, but other biologics were not associated with statistically significantly different risk than abatacept (figure 2C, D). The Kaplan–Meier plot suggested minimal difference in AMI-free survival by the type of biologic (figure 1). The proportionality assumption was not violated.

Other patient characteristics significantly associated with increased risk of AMI included demographics (age, gender, race, original reason for Medicare enrolment), comorbid conditions (diabetes, chronic kidney disease, peripheral arterial disease, COPD) and medication use (antihypertensives and oral glucocorticoids). Use of oral glucocorticoids was significantly associated with increased risk of MI, both at a lower dose of <7.5 mg/day (HR 1.23; 95% CI 1.02 to 1.49) and a higher dose of ≥7.5 mg/day (HR 1.74; 95% CI 1.38 to 2.19) compared with no use, whereas only the use of high-dose glucocorticoids was associated with increased risk of the composite CHD outcome (HR 1.31; 95% CI 1.09 to 1.57). Other patient characteristics significantly associated with increased risk of the composite CHD outcome were similar to those for MI. Detailed data are presented in online supplementary table S3.

Results from subgroup analyses are presented in online supplementary table S4. In the analysis stratified by prior biologic use, the only statistically significant association was that between TNF biologic and increased MI risk in all patients and in biologic-naïve patients compared with abatacept. Tocilizumab was associated with increased MI risk among biologic-naïve patients. However, due to reduced sample sizes (from 17 events among tocilizumab initiators in the main analysis to <11 in the analysis among biologic-naïve patients), the association was not statistically significant and had broad CIs, which made it difficult to interpret. When we examined the composite outcome that included death and acute MI, the number of cases quadrupled and both rituximab and TNF biologic initiators were at increased risk of developing the composite outcome compared with abatacept initiators.

Discussion

In this retrospective analysis of more than 47 000 patients with RA, we observed a higher risk of AMI associated with anti-TNF biologics, particularly etanercept and infliximab, compared with initiators of abatacept. We did not find an elevated risk among tocilizumab users either for AMI or the composite CHD outcome that included AMI and revascularisation procedures. To our knowledge, this study is unique compared with the previous literature addressing the potential effect of biologics on their CHD risk in patients with RA in that: (1) it assessed CV risk among patients receiving biologics that are not anti-TNF antagonists, and (2) it compared across different types of biologics instead of comparing biologic users to patients not using biologics. As a result, the findings have distinctive clinical implications as they suggest that abatacept may have as favourable a risk–benefit profile as anti-TNF agents, if not more, and that CHD event risk appears at least on par for tocilizumab as other biologic therapies.

Given the excess CV burden in patients with RA and evidence suggesting that RA-associated inflammation is one of the causes for the observed increase in CV risk, researchers have postulated that RA treatments may alter CV risk. A number of studies have associated anti-TNF biologics with no change in or reduced risk of CV risk and mortality compared with non-biologic DMARDs4–8 ,18 ,19 and suggest that anti-TNF therapy reduced CVD event risk between 20% and 57%.5–8 ,18 Some data have suggested that the effect is confined to the patients who respond clinically, with a reduction in systemic inflammation.8 Our results suggest that abatacept is at least comparable with anti-TNF therapy, if not more favourable.

Existing data on CV risk among abatacept users are scant and conflicting. A prior report associated abatacept with increased aortic stiffness among 21 patients with RA who received 6 months of treatment.13 The study also examined the lipid profiles before and after treatment and observed increases in total and LDL-C (but no change in overall lipid profile). However, belatacept, a molecule closely related to abatacept, was associated with lower blood pressure, better lipid profile and reduced diabetes risk when compared with other immune-suppressive agents among patients who underwent kidney transplant.20 To our knowledge, no prior study has examined the association of abatacept with more clinically important outcomes such as AMI.

The effects of tocilizumab, an IL-6 antagonist, on CV risk are of particular interest, since elevations in total and LDL-C levels have been observed among patients who initiated tocilizumab, which has raised CV safety concerns.21 ,22 For example, in a randomised trial, the LDL-C level increased from <160 to ≥160 mg/dL from baseline to 24 weeks among 16% of the patients with RA assigned to receive tocilizumab, compared with just 3% of the patients in the control group.22 Data pooled from six trials reported an AMI IR of 2.5 per 1000 PYs (95% CI 1.6 to 3.8) among patients exposed to tocilizumab,14 which is lower than those reported from anti-TNF-treated patients enrolled in the British Society for Rheumatology Biologic Register (4.8 per 1000 PYs (95% CI 3.7 to 6.1))8 but higher than that from patients enrolled in a large north American RA registry (0.5 per 1000 PYs (95% CI 0.01 to 1.0)).5 In our study, tocilizumab was not associated with increased risk of AMI compared with reference (abatacept), after adjusting for CV risk factors. In fact, we found a marginally significant protective effect. It seems quite possible that this observation is a result of confounding by indication (in this case ‘contraindication’) if prescription of tocilizumab is avoided in patients with perceived high CV risk.

Prior to our study, rituximab treatment has been reported to improve the lipid profile and reduce levels of prothrombotic biomarkers in patients with RA in two small studies.11 ,12 The study by Jin and colleagues reported greater reduction among rituximab-treated- compared with anti-TNF-treated patients.11 However, this difference may not translate to clinical effect. In our study, patients with RA who initiated anti-TNF and rituximab were not different in terms of their risk of developing AMI and the composite outcome.

To ensure appropriate interpretation of our findings, it is important to point out that this study was not intended or designed to address the effect of non-biologic DMARD or statins on CHD risk. We employed a new-user design to assess CHD risk in relation to biologics to ensure that we captured early events, to reduce bias associated with adherence and to ensure the appropriate capture of baseline covariates. Because the study was not specifically designed to assess the risk of AMI associated with the use of non-biologic DMARDs, statin or other medications, we want to caution against drawing conclusions about the influence of any of these medications with AMI or the composite CHD outcome.

Our study has a number of limitations. One of the most impactful may be the lack of clinical data, such as disease severity or activity at time of treatment initiation, given that active RA is a recognised risk factor for adverse CV outcomes. If select biologics are perceived to be more effective and preferentially used in patients with more severe disease (eg, anti-TNF therapy), confounding by disease severity may have biased the results and showing less favourable outcomes for these patients. Similarly, the study lacked clinical data on important CV risk factors such as lipid levels. This limitation may be particularly relevant for tocilizumab since tocilizumab initiation leads to increases in lipid levels in clinical trials, a finding that may cause physicians to avoid the use of tocilizumab among patients with perceived high CV risk. While we were unable to adjust for these important clinical factors in our multivariable analyses, we examined the proportions of patients receiving no, low and high doses of oral glucocorticoid and prescription NSAIDs during the baseline period by type of biologic initiated, which could be proxies for RA disease severity. Similarly, data on statin use, history of diabetes and history of hypertension from the baseline period could serve as potential surrogate markers of CV risk (see online supplementary table S2). We did not detect any perceptible signal that suggests the presence of confounding by RA severity or CV risk. For example, 27% of patients initiating tocilizumab had filled a prescription for statin during the baseline period, which was within the range from 25.1% (rituximab) to 29.7% (infliximab) observed for patients initiating other biologics.

Despite the use of nationwide Medicare claims of patients with RA, this study was nevertheless limited by a relatively small sample size for biologics that were introduced more recently. For instance, we had 1574 golimumab- and 2727 tocilizumab-exposed PYs in the analysis of the composite CHD outcome; therefore, CIs were wide for these biologics. Another limitation was that patients with RA who experienced an AMI and never were hospitalised would not have been captured, although our sensitivity analysis that examined a composite of CHD events or death addressed this concern. We also had incomplete data on smoking, but it is unlikely that smoking history would differ substantially between users of different biologic therapies. In addition, we were unable to capture over-the-counter NSAIDs and prophylactic aspirin use. The inability to control for these factors may result in residual confounding.

Our study has a number of strengths. We used population-based data from a large population of older patients with RA. As a result, we were able to analyse data with more than 585 AMI events accrued from more than 74 662 exposed PYs of follow-up. We directly assessed the clinical outcomes and not surrogate outcomes. We used validated algorithms to exclude patients with prevalent coronary heart disease (CHD) and to identify incident CHD outcomes.16 ,23 We included a set of carefully chosen CVD risk factors, in addition to demographics and use of health services, in the adjusted analysis. Finally, because the study compared among patients who initiated biologics, the analysis is less likely to be biased by residual confounding compared with analysis that compared biologic users with those who used non-biologic DMARDs only.

In conclusion, the findings from this large observational study suggest a somewhat greater AMI risk among patients with RA who initiated some anti-TNF agents compared with abatacept, although the absolute rate differences were small. This study also provides evidence that tocilizumab does not appear to increase risk for CHD events. These findings, however, should be interpreted with caution given the absence of data on important CV risk factors and RA disease severity.

References

View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Handling editor Tore K Kvien

  • Contributors FX analysed the data. Everyone contributed to the conception, design and interpretation of the data. JZ drafted the article. Everyone revised the article for important intellectual content. All authors gave final approval for the version to be published.

  • Competing interests MTO is an advisory board member for Takeda, Lycera, Janssen, AbbVie and UCB, and has received research grant support from UCB. JAS has received research grants from Takeda and Savient and consultant fees from Savient, Takeda, Regeneron and Allergan. JAS is a member of the executive of OMERACT, an organisation that develops outcome measures in rheumatology and receives arms-length funding from 36 companies; a member of the American College of Rheumatology’s Guidelines Subcommittee of the Quality of Care Committee; and a member of the Veterans Affairs Rheumatology Field Advisory Committee. MMS has received funding for investigator-initiated research from Amgen, Inc. and consults for diaDexus and Medscape. EBL receives research funding from Amgen, Inc and has served as a consultant for Amgen and Robinson Calcagnie Robinson Shapiro Davis. JRC has received research grant funding and consulting monies from Abbvie, Amgen, BMS, Janssen, Pfizer, Roche and UCB. JDL has received research funding from Takeda, AbbVie, Bayer and Nestle Health Science. JDL has served as a consultant to Amgen, Takeda, Lilly, Shire, Nestle Health Science, Merck, Janssen, Immune Pharmaceuticals, AstraZeneca, Pfizer and AbbVie. STK has received research support from Amgen. KS has received research funding from Amgen, AstraZeneca, Lilly and Merck and has served as a consultant to Amgen, AstraZeneca, Bayer, Merck, Pfizer and Roche/Genentech.

  • Ethics approval University of Alabama at Birmingham IRB.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.