Article Text

Download PDFPDF

Extended report
Biomarker-related risk for myocardial infarction and serious infections in patients with rheumatoid arthritis: a population-based study
  1. Jeffrey R Curtis1,2,
  2. Fenglong Xie1,
  3. Lang Chen1,
  4. Kenneth G Saag1,2,
  5. Huifeng Yun1,2,
  6. Paul Muntner2
  1. 1 Division of Clinical Immunology and Rheumatology, University of Alabama at Birmingham, Birmingham, Alabama, USA
  2. 2 Department of Epidemiology, University of Alabama at Birmingham, Birmingham, Alabama, USA
  1. Correspondence to Dr Jeffrey R Curtis, Division of Clinical Immunology and Rheumatology, University of Alabama at Birmingham, Birmingham, AL 35294, USA; jrcurtis{at}uabmc.edu

Abstract

Background Rheumatoid arthritis (RA) disease activity and associated systemic inflammation has been associated with serious infection (SIEs), myocardial infarction (MI) and coronary heart disease (CHD) events based on a few registry studies or clinical trials. There are few data from large-scale population-based studies given feasibility challenges in conducting such investigations.

Methods Multibiomarker disease activity (MBDA) test scores (n=77 641) were linked to Medicare for US patients with RA. Outcomes of interest were hospitalised pneumonia/sepsis (SIE), MI and a composite CHD outcome. The MBDA score ranges from 1 to 100 and was analysed as time-varying. Cox proportional hazards models evaluated the association between MBDA score and SIEs, MI and CHD events, controlling for potential confounders. A sensitivity analysis excluded C reactive protein (CRP) from the MBDA score.

Results There were 17 433 and 16 796 patients eligible for the SIE and MI/CHD analyses, respectively. Mean (SD) age was 69 (11) years, 79% were women, 81% were white and 38% were disabled. Over 16 424 person-years of follow-up, there were 452 SIE events, 132 MIs and 181 CHD events. Higher MBDA scores were associated with SIEs (HR=1.32, 95% CI 1.23 to 1.41 per 10 unit MBDA score change). For MI/CHD events, a threshold effect was present; higher disease activity by MBDA score was associated with increased MI (HR=1.52, 95% CI 0.92 to 2.49) and CHD rates (HR=1.54, 95% CI 1.01 to 2.34, comparing scores ≥30 vs <30). Analyses of the MBDA score without CRP yielded similar results.

Conclusion Higher MBDA scores were associated with hospitalised infection, MI and CHD events in a large, predominantly older, US RA population.

  • rheumatoid arthritis
  • infections
  • cardiovascular disease

Statistics from Altmetric.com

Introduction

The goal of treatment for patients with rheumatoid arthritis (RA) is remission or low disease activity.1 2 While the benefits of lower disease activity states on clinical and radiographic outcomes are clear,3 4 a number of hypotheses suggest that a reduced systemic inflammatory burden may have non-articular benefits. For example, a few studies have found reduced rates of infections associated with lower RA disease activity as measured clinically.5–7 Consistent with this hypothesis, elevated C reactive protein (CRP) has been associated with higher rates of coronary heart disease (CHD) in several population-based studies of non-RA patients.8 A large randomised controlled trial showed that randomisation to a statin versus placebo lowered the cardiovascular disease (CVD) event rate in patients with elevated CRP,9 and reductions in CRP were correlated with greater CVD risk benefit.10

Population-based research in RA studying hard endpoints including hospitalised infection and myocardial infarction (MI) is challenging because the relatively low prevalence of RA and outcome event rates limits statistical power. Administrative data from health plans and payers have high validity for studying large cohorts of patients with RA.11 While these data sources often lack clinical assessments of RA, results of lab tests that measure RA disease activity may provide objective measurements that can augment claims data. One such measure is the multibiomarker disease activity (MBDA) test, a validated 12-protein biomarker assay12 that has been shown to have good correlation with the Disease Activity Score in 28 joints using the CRP (DAS28-CRP) in several RA cohorts.13

Given interest in quantifying the relationship between RA disease activity and serious adverse events (SAEs) and other health-related outcomes, we assembled a large cohort of patients with RA and Medicare coverage, and linked each patient to their MBDA lab test results. The MBDA score served as the measure of RA disease activity. Using this approach, we examined the association between the MBDA score and several outcomes including hospitalised infection, MI, CHD events and total healthcare costs.

Methods

We conducted a longitudinal cohort study of older individuals in the USA with RA and Medicare coverage who received MBDA testing as part of standard of care from their treating rheumatologist.

Data source

We used national Medicare fee-for-service claims data made available from the Centers for Medicare and Medicaid Services (CMS) from 2010 to 2014 for this analysis. The CMS data contain all inpatient (part A), outpatient (part B) and prescription medication use (part D) claims for Medicare beneficiaries in the USA. The MBDA test is reimbursed nationally from a single laboratory provider who bills Medicare directly for tests ordered by rheumatologists as part of a patient’s routine care. Results of the MBDA tests for all Medicare beneficiaries were obtained directly from the laboratory provider (Crescendo Biosciences, South San Francisco, California, USA) for this analysis. The data provided included the MBDA score, leptin and CRP values, along with patients’ birth date, sex, state of residence, blood sample collection date, referral physician’s national provider identification (NPI) number and dates of service. A linkage between MBDA test and Medicare claims (billed as Healthcare Common Procedure Coding System (HCPCS) codes 84999, 84179 and 84190) was considered successful if a unique match was made between the CMS data and the laboratory’s database on these factors: (1) full birth date, (2) sex, (3) NPI number and (4) date of service. The study was governed by a Data Use Agreement from the CMS.

Cohort selection

To be included in the study, a patient must have had (1) at least one MBDA score linked to Medicare claims that was considered valid (see definition below), and (2) at least 365 continuous days of Medicare with part D (pharmacy) coverage before the first valid MBDA test date. The later of these two dates was defined as the start of follow-up (‘index date’). This 365-day baseline period was used to assess comorbidities and other baseline characteristics. Patients were excluded if (1) there was an International Classification of Diseases (ICD)-9 diagnosis code for ankylosing spondylitis, inflammatory bowel disease psoriasis, psoriatic arthritis, systemic lupus erythematosus, malignancy (ignoring non-melanoma skin cancer), polymyalgia rheumatica and giant cell arteritis in the 12-month baseline period; and (2) they had initiated any non-tumour necrosis factor (TNF) biologic or synthetic targeted DMARDs including abatacept, anakinra, rituximab, tocilizumab, ustekinumab or apremilast or tofacitinib in the 183 days before the index date. This latter restriction was applied given the possibility that some biologics might have a differential influence on the biomarker profile.14 For the MI and CHD outcomes, patients with prior MI (ICD-9 diagnosis code 410.xx or 412.xx) or any HCPCS code for coronary artery bypass grafting (CABG) or percutaneous coronary intervention (PCI) during the baseline period were also excluded.15 Given the possibility that the MBDA scores could be affected by an evolving outpatient infection or MI event, patients were excluded from the analysis of MI/CHD events if events occurred in the 7 days following the MBDA test; for the serious infection events (SIE) analysis, patients were excluded if an SIE event occurred in the 14 days after an MBDA test. The analysis of total healthcare costs was conducted among the subgroup of patients who met cohort inclusion criteria and had 1 year of follow-up after the index date.

Exposure: RA disease activity as measured by the MBDA score

The MBDA score has been studied extensively and undergone rigorous validation in numerous RA cohorts. It has been shown to correlate with the DAS28-CRP both in cross-sectional and longitudinal analyses13 16–18 and with radiographic and other musculoskeletal imaging outcomes.19 20 It consists of 12 biomarkers including TNF-RI, serum amyloid A (SAA), interleukin 6 (IL-6), matrix metalloproteinase-1 (MMP-1), matrix metalloproteinase-3 (MMP-3), epidermal growth factor (EGF), vascular endothelial growth factor (VEGF), YKL-40, vascular cell adhesion molecule-1 (VCAM-1), resistin, leptin and CRP.12 These are weighted according to a validated formula and combined into a single score that ranges from 1 to 100, with higher scores reflecting more RA disease activity. In this analysis, the MBDA score was time-varying, and its value and associated disease activity category (low, moderate, high) was updated with each new test result. It was analysed as a continuous variable, in quartiles, and using established RA disease activity cutpoints (low <30, moderate 30–44, high >44).12 13

Given the established association between CRP and CHD events, and because CRP is one of the 12 biomarkers in the MBDA test, we assessed the association between the other 11 MBDA biomarkers and the outcomes of interest by recalculating a biomarker score without CRP (ie, setting its weight to 0). This new biomarker score (‘MBDA score without CRP’) was analysed in quartiles, as there are no established clinical cutpoints for this score. Given our hypothesis that low disease activity would be associated with lower rates of the outcomes of interest and the possibility of a threshold effect, we also compared the lowest quartile with quartiles 2–4, collapsed into a single category.

MBDA scores were not considered valid for this analysis and were therefore excluded (n=10 996) if patients had an outpatient infection (based on receipt of outpatient antibiotics), vaccination (pneumococcal, influenza or herpes zoster, based on HCPCS codes) or hospitalisation (based on any inpatient claim from Medicare) within the 21 days prior to the index date, given the possibility that these events might affect the MBDA score. All MBDA tests were ordered as part of rheumatologists’ standard of care for their patients with RA. Most patients (72.9%) had only one MBDA test result available for the current analysis, while 18.2% had two MBDA tests, and 8.9% had three or more test results.

Outcomes

The five outcomes of interest included (1) hospitalised serious infectious event for pneumonia or sepsis with a discharge diagnosis code in the primary position, generally indicating the main reason for hospitalisation (SIE-primary); (2) hospitalised infection for pneumonia or sepsis (primary or non-primary position discharge diagnosis code, SIE-all); (3) hospitalised MI (primary or non-primary position discharge diagnosis code); (4) CHD events including MI and PCI and CABG procedures; and (5) total costs and medical costs as paid for by the Medicare programme. The algorithms to identify the hospitalised infections and MI events as defined above have been previously shown to have high validity, with positive predictive values in the 85%–95% range.21–23 Direct medical costs were obtained from the Medicare data and calculated as 1-year incremental costs, subtracting each person’s costs during the 12-month baseline period. Indirect medical costs were not estimated.

Statistical analysis

Baseline characteristics of the RA cohort were assessed at the time of the first MBDA test, stratified by MBDA category (low, moderate and high disease activity). Event rates and 95% CIs were computed for each of the four outcomes.

Cox proportional hazards models with MBDA score as a time-dependent variable were used to evaluate the association between the MBDA score and the first occurrence of each outcome of interest, controlling for potential confounders that were selected based on their hypothesised associations with the MBDA score and the outcomes under study. These covariates were measured at baseline, and included age, sex and race, and baseline history of heart failure, stroke, abdominal aortic aneurysm, peripheral arterial disease, atrial fibrillation, diabetes, hyperlipidaemia, hypertension, obesity, smoking, chronic kidney disease, chronic obstructive pulmonary disease, pneumonia, sepsis, fibromyalgia, peptic ulcer disease, fracture and skin ulcer, all ascertained using diagnosis codes from physician office visits or hospitalisations. Covariates also included health-seeking behaviours, including cancer screening including prostate-specific antigen, Papanicolaou smear and mammography. RA factors were controlled for baseline hydroxychloroquine, leflunomide, sulfasalazine, biologic use, methotrexate (MTX) dose in the preceding 4 months, and glucocorticoid dose in the preceding 6 months, measured by summing the cumulative prednisone-equivalent dose dispensed over this period, dividing by 183 days, and categorised as none, ≤7.5 mg/day and >7.5 mg/day. A sensitivity analysis censored follow-up at 12 months after each MBDA test so as to avoid misclassification of the score over time.

After examining the correlation between the MBDA score and CRP and a recalculated biomarker score without CRP, the analyses were repeated using the MBDA score without CRP as the main independent variable. Additionally, as obesity has been associated with CHD events in some studies, we conducted sensitivity analyses that also adjusted for leptin, which has been shown to be a proxy for obesity and fat mass.24–26 Leptin was log-transformed to be more linearly related to outcomes. The proportional hazard assumption was tested using the method described by Lin et al 27 and no violations were present. All analyses were conducted on SAS V.9.4.

Results

Eligible population

A total of 17 433 patients were eligible for the SIE analysis and 16 796 patients for the MI/CHD outcome (online supplementary appendix figure 1). Overall, the mean (SD) age of participants in the cohort was 69 (11) years, 79% were women, 81% were white and 38% were disabled (bottom row). RA therapies being taken included biologics (17%), MTX (54%), other non-biologic disease-modifying antirheumatic drugs (DMARDs) (40%) and oral glucocorticoids (53%). Characteristics of patients in lower MBDA categories suggested that compared with patients with higher MBDA scores, patients were younger (3-year difference) and had a lower burden of comorbidities, with less glucocorticoid use and more biologic use (table 1).

Supplementary file 1

Table 1

Baseline patient characteristics by MBDA category

MBDA and serious infections, MI and CHD events

Over up to 16 424 person-years of follow-up, depending on the outcome, there were 452 SIE-primary, 653 SIE-all, 132 MI and 181 CHD events. Higher MBDA scores, modelled in quartiles (figure 1), were associated with increasing outcome rates, as were MBDA scores modelled per 10 unit increase or using established RA disease activity cutpoints (table 2). After multivariable adjustment, higher MBDA scores were associated with statistically significantly higher rates of SIE-primary and SIE-all events and higher MI and CHD rates. The sensitivity analysis that censored follow-up 12 months after each MBDA score yielded similar results (data not shown).

Figure 1

Incidence rates* of hospitalised infections, MI and CHD events by quartile of MBDA. Error bars represent the 95% CIs around the incidence rate. *Rates per 100 patient-years. **Primary or Secondary. CHD, coronary heart disease; MBDA, multibiomarker disease activity; MI, myocardial infarction; SIE, serious infection event.

Table 2

Crude incidence rates and adjusted HRs of serious infection events, MI and CHD associated with MBDA score

MBDA adjusting for CRP and leptin

There was a high correlation between the MBDA score and the MBDA score without CRP (r=0.97) (scatterplot, online supplementary appendix figure 2). Higher MBDA scores without CRP were associated with increased outcome rates (online supplementary appendix table). For the MI and CHD outcomes, there was a dose–response relationship between higher quartiles of the MBDA without CRP and the SIE outcomes, but for MI and CHD events quartiles 2–4 were associated with similar risk. In analyses, with and without exclusion of CRP or adjustment for leptin, the adjusted HR for MI and CHD events in the highest three compared with the lowest quartile of disease activity was between 1.5 and 1.8 (table 3). Recognizing that leptin is part of the MBDA score, leptin as a model covariate was associated with a protective effect for both MI (adjusted HR for log-transformed leptin=0.79, 95% CI 0.69 to 0.91) and CHD (adjusted HR=0.83, 95% CI 0.73 to 0.94). There was no statistically significant interaction between sex and log-transformed leptin (P=0.16 for both AMI and CHD events).

Table 3

Adjusted HRs for MI and CHD associated with MBDA score

MBDA and cost analysis

In the subgroup analysis of 10 058 people who had full medical and pharmacy coverage for 12 months after their MBDA test, analysis of healthcare costs associated with MBDA score (table 4) showed that patients with higher disease activity by MBDA score had higher baseline healthcare costs. Although not statistically significant, the incremental costs, compared with each person’s baseline costs, during the follow-up period were numerically higher among those in higher disease activity categories based on the MBDA score.

Table 4

Healthcare costs by MBDA category*

Discussion

SIE, MI and death are among the most concerning SAEs that occur in patients with RA. The role of RA disease activity and associated systemic inflammation has not been well examined as it relates to these outcomes. In this large RA population predominantly consisting of older individuals, higher MBDA scores were associated with increased risk for hospitalised infection, MI and CHD events. A strong dose–response gradient existed between MBDA scores and hospitalised infections (pneumonia and sepsis). The patterns for MI and CHD events suggested more of a threshold effect, where those with the lowest level of disease activity and inflammation were at lowest risk, but gradations with higher levels of the biomarker were relatively absent.

Prior cohort studies have found that RA-related disease activity as measured clinically (eg, by the Clinical Disease Activity Index (CDAI)) correlates with risk for CVD events, hospitalised infections and other important healthcare outcomes. For example, in a US observational study conducted in the Corrona registry, a 10-point lower CDAI was associated with a 21% (95% CI 13 to 29) lower rate of CVD events.28 In a separate analysis from the Corrona cohort, there was a strong association between higher CDAI and increased rates of all-cause hospitalisation (with infections as the most common type), mortality and healthcare costs in a dose-dependent fashion.7 In an analysis of CVD events associated with various risk factors in a tocilizumab-treated RA population, changes in disease activity, but not lipid levels, were associated with CVD events.29 While some reports initially suggested that RA disease duration might impact CVD risk, this has not been uniformly confirmed in subsequent analyses, and it appears that disease activity over time is a more important determinant of CVD risk.30 Concordant with this hypothesis, a number of observational studies have found that initiation of biologics for RA reduces the risk for CVD events,31–33 and some of these studies have been able to show that this reduction is in part mediated by lower disease activity. Long-term decreases in infection risk also have been observed following biologic initiation, in part mediated through improvements in functional and reductions in disease activity and steroid dose.34

Population-based studies in RA are sometimes limited by insufficient statistical power to study systemic inflammation and its association with hard endpoints like MI. Consequently, carotid intimal medial thickness or coronary artery calcification, subclinical markers of CHD, have been studied instead of clinical endpoints.35 36 The current study is novel in that it leveraged a large administrative data source linked to a laboratory test provider database to address a question that neither data source by itself could address. Given concerns that the associations that we studied could be unduly influenced by CRP, a factor known to be related with CHD risk, we performed sensitivity analyses that excluded this biomarker. In sensitivity analyses, we analysed only the remaining 11 MBDA biomarkers and found very similar results that showed a 1.5-fold (CHD) and 1.7-fold (MI) elevated risk for patients in the higher disease activity category compared with the lowest. While the data source did not contain information on obesity, leptin was used as a surrogate to control for this potentially confounding factor, which strengthened the associations of interest.

The current results should be interpreted in light of the study design. We examined disease activity as measured by the MBDA test, but we did not address the possibility that patients subsequently started RA therapy to lower disease activity as a mediator that might reduce risk for SIE, MI and CHD. Thus, the true associations may be stronger than the associations we reported here. Outcome events were not confirmed by medical record review, although they have been shown to have high validity (eg, positive predictive value of 85%–95%) in studies comparing the approach we used with medical record review as the gold standard.23 We also recognise the possibility that despite controlling for a variety of potential confounding factors, residual confounding (eg, due to comorbidities that may both increase risk for SIEs and MI and also relate to higher MBDA scores) may have been present. The MBDA score was not used as part of a RA-specific cardiovascular risk prediction equation or ‘calculator’,37 38 although ongoing work is evaluating this possibility. This cohort included predominantly older adults with mean age of 69 years, and results may not be generalisable to younger patients with RA. However, given that the prevalence of infectious and CHD risk factors generally increases with age, we would speculate that the associations between RA disease activity and the outcomes that we studied might be even stronger in younger patients with RA. Finally, we recognise that the biomarker test itself has a cost associated with ordering it; prior work has suggested that it may be cost-effective as part of RA clinical management.39 While its availability greatly improves feasibility to conduct large-scale epidemiological studies like ours, its utility and cost-effectiveness for predicting CHD or infection risk have yet to be determined, especially compared with measuring disease activity using the DAS28 or other clinical metrics.

In conclusion, higher disease activity as measured by a panel of biomarkers was associated with higher rates of hospitalised infections, MI and CHD events. These findings add to the growing body of evidence that further strengthens the argument to strive for lower disease activity in RA. While the accepted goal of lower RA disease activity is to improve patients’ articular signs and symptoms, the current results also show that lower disease activity as measured by the MBDA score is associated with a lower rate of non-articular clinical events, including hospitalised infections and CHD events. Use of the MBDA score to risk-stratify patients to identify those with high levels of inflammatory activity and are at high risk for SAEs may be possible to help clinicians identify those at greatest risk and targets for early intervention including intensive RA management.

References

View Abstract

Footnotes

  • Handling editor Tore K Kvien

  • Contributors JRC has full access to the data and takes responsibility for all aspects of the analysis, including drafting the manuscript. All coauthors edited the manuscript for content and approved submission of the final version.

  • Funding This analysis was supported in part by Crescendo Bioscience, a Myriad Genetics Company. JRC receives support from the Patient-Centered Outcomes Research Institute (PCORI).

  • Competing interests JRC: consultant and research grants from Myriad Genetics.

  • Ethics approval UAB Institutional Review Board for Human Use.

  • 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.