Abstract
Objectives
Rheumatoid arthritis (RA) and atherosclerosis share many common inflammatory pathways. We studied whether a multi-biomarker panel for RA disease activity (MBDA) would associate with changes in arterial inflammation in an interventional trial.
Methods
In the TARGET Trial, RA patients with active disease despite methotrexate were randomly assigned to the addition of either a TNF inhibitor or sulfasalazine+hydroxychloroquine (triple therapy). Baseline and 24-week follow-up [18F]fluorodeoxyglucose–PET/CT scans were assessed for change in arterial inflammation measured as the maximal arterial target-to-blood background ratio of FDG uptake in the most diseased segment of the carotid arteries or aorta (MDS-TBRmax). The MBDA test, measured at baseline and weeks 6, 18 and 24, was assessed for its association with the change in MDS-TBRmax.
Results
Interpretable scans were available at baseline and week 24 for 112 patients. The MBDA score at week 24 was significantly correlated with the change in MDS-TBRmax (Spearman’s rho = 0.239; P = 0.011) and remained significantly associated after adjustment for relevant confounders. Those with low MBDA at week 24 had a statistically significant adjusted reduction in arterial inflammation of 0.35 units vs no significant reduction in those who did not achieve low MBDA. Neither DAS28-CRP nor CRP predicted change in arterial inflammation. The MBDA component with the strongest association with change in arterial inflammation was serum amyloid A.
Conclusion
Among treated RA patients, achieved MBDA predicts changes in arterial inflammation. Achieving low MBDA at 24 weeks was associated with clinically meaningful reductions in arterial inflammation, regardless of treatment.
Keywords: rheumatoid arthritis, biomarkers, cardiovascular disease, atherosclerosis, prediction, disease modifying antirheumatic drugs, tumour necrosis factor inhibitors, positron emission tomography, inflammation, serum amyloid A
Rheumatology key messages.
Whether improvements in atherosclerotic plaque inflammation can be predicted with circulating biomarkers was previously unknown.
The multibiomarker disease activity score indicated which RA patients had a reduction in arterial inflammation.
Circulating biomarkers could be used to assess cardiovascular risk reduction after starting DMARD therapy.
Introduction
On average, people with rheumatoid arthritis (RA) have a greater burden of atherosclerosis and a heightened risk of cardiovascular disease (CVD) events compared with the general population [1]. Inflammatory pathways upregulated in RA likely contribute to accelerated atherogenesis and potentiate atherothrombosis [2]. However, the immunobiology of RA is complex, and the specific inflammatory contributors to CVD in RA have not been well established.
Modifying and reducing these inflammatory contributors to CVD with immunomodulator treatments has the potential to ameliorate CVD in RA patients, an effect that was explored in the Treatments Against RA and Effect on FDG-PET/CT (TARGET) Trial [3]. In this multicentre, randomized, single-blinded, active comparator trial, RA patients with active disease despite methotrexate (MTX) were randomized to adding sulfasalazine (SSZ) and hydroxychloroquine (HCQ, i.e. triple therapy) vs the addition of a TNF inhibitor (TNFi, either etanercept or adalimumab) to their background MTX. The primary outcome was the change in the uptake of [18F]fluorodeoxyglucose (18F-FDG) in the aorta and carotid arteries, measured with PET/CT after 24 weeks of treatment. FDG uptake into the arterial wall is a surrogate for the presence and activation status of immune effector cells in atherosclerotic plaques [4] and has been shown to correlate with future CVD events in epidemiological studies [5, 6]. In TARGET [3], both treatments were associated with significant reductions in both articular disease activity and arterial FDG uptake. However, the change in arterial FDG uptake was not significantly correlated with the change in articular disease activity (i.e. the DAS28-CRP score) nor the change in circulating CRP.
An additional a priori-specified aim of TARGET was an exploration of a validated serum multi-biomarker of RA disease activity (the MBDA score) to categorize treatment response and correlate it with an imaging biomarker of vascular inflammation. The MBDA is a commercially available CLIA-certified serum multi-marker that includes 12 protein analytes relevant to RA inflammation [7] [epidermal growth factor (EGF); IL-6; matrix metalloproteinase 1 and 3 (MMP-1 and MMP-3); serum amyloid A (SAA); TNF receptor 1 (TNF-R1); vascular cell adhesion molecule 1 (VCAM-1); vascular endothelial growth factor-1 (VEGF-1); chitinase-3-like protein 1 (YKL-40); CRP; leptin; and resistin]. These 12 analytes were derived from an initial candidate panel of 130 circulating proteins using samples from multiple RA cohorts from North America and Europe [7]. It was correlated with disease activity [8], as measured with the disease activity score for 28 joints (DAS28), and was shown to be predictive of future radiographic damage [9]. In addition, each of the MBDA components has been implicated in the processes driving atherogenesis and/or atherothrombosis [10]. As such, components of the MBDA were previously used to derive a cardiovascular MBDA score that improved the prediction of CVD events in a study of RA patients [11]. Our primary a priori hypothesis was that RA patients achieving MBDA-defined ‘low’ disease activity (using a priori defined thresholds) at 24 weeks would have less arterial inflammation than those with persistent moderate or high disease activity. Secondarily, we hypothesized that individual component(s) of the MBDA may correlate differently with the change in arterial inflammation and that the cardiovascular MBDA score may also correlate with the change in arterial inflammation.
Methods
Study design
Both the design [12] and the main report of the trial findings [3] of the TARGET Trial have been published. Briefly, RA patients at least 45 years old for men and 50 years old for women with a DAS28-CRP > 3.2 despite at least 8 weeks of MTX treatment at a dose ≥15 mg/week (or on at least 7.5 mg for ≥8 weeks with a documented intolerance) with a stable dose for the 4 weeks prior to screening were eligible. Use of biologic or targeted synthetic DMARDs in the prior 6 months, ever use of rituximab, and prior TNF inhibitor inadequate response were exclusions. Patients with prior CVD events were excluded. The use of low-intensity statins and HCQ was allowed. Additional inclusion/exclusion criteria can be found in Giles et al. [12]. All patients provided written informed consent prior to enrolment, and the trial was approved centrally by the MassGeneral Brigham Healthcare Human Subjects Committee or by individual study site Institutional Review Boards (IRB).
Eligible patients underwent 18F-FDG-PET/CT scanning and were randomized if no additional exclusions (i.e. incident findings on scanning that would preclude enrolment) were identified. Enrolees were randomized 1:1 to the addition of a TNFi, either etanercept 50 mg subcutaneous weekly or adalimumab 40 mg subcutaneous every other week, or triple therapy, SSZ (titrated to 1000 mg BID) and HCQ (200 mg BID or a lower dose based on weight). All DMARDs, except MTX, were provided to the subjects. The patients and their treating rheumatologists were not blinded to study treatment, but site metrologists and all central data interpreters (i.e. the primary trial investigators and imaging readers) were blinded. Enrolees returned for assessments of disease activity and drug safety at 6, 12 and 18 weeks after baseline, during which, per-protocol treatment changes could be made if the Clinical Disease Activity Index (CDAI) score was ≥10. Enrolees returned at week 24 for a follow-up 18F-FDG-PET/CT scan. The pre-specified primary outcome of the TARGET trial was the change in the maximal target-to-background ratio (TBRmax) of 18F-FDG uptake for the most diseased segment (MDS) of the index vessel (either carotid artery or aorta). Details of the scanning protocol and study visits can be found in Giles et al. [12]. In TARGET, 159 patients were randomized, of which 138 completed follow-up. There were 115 with paired FDG-PET/CT scans that were analysable for the primary outcome, among which 112 had complete MBDA data. MDS-TBRmax significantly decreased in both treatment groups, an effect that was independent of the change in articular disease activity or the change in CRP.
Laboratory assessments
Enrolees underwent whole blood collection at baseline and at weeks 6, 18 and 24. Sites were mailed collection kits, with samples processed on-site according to the manufacturer’s specifications. Samples were transferred from the site to the TARGET Central Biorepository at Brigham & Women’s Hospital before transfer to Crescendo Biosciences, where they were assayed as a single batch. The MBDA score was categorized into MBDA disease activity as defined by the manufacturer as follows: low disease activity (low MBDA) <30, moderate disease activity (moderate MBDA) 30–44, and high disease activity (high MBDA) >44 [8]. Additionally, we calculated the CVD MBDA score as published by Curtis et al. [11] using the formula:
CVD MBDA Score = (0.0314 × Age) + (0.2691 × Current Smoker) + (0.2732 × Diabetes) + (0.2694 × Hypertension) + (0.3378 × Coronary Artery Disease) − (0.1711 × logLeptin) + (0.1454 × logMMP3) + (0.5724 × logTNF-R1) + (1.6076 × hyperbolic tangent transformation of MBDA/33.0807).
Statistical analyses
Baseline characteristics were compared according to categories of the baseline MBDA score using Student’s t-test for normally distributed continuous variables, the Kruskal–Wallis test for non-normally distributed continuous variables, and chi-squared goodness-of-fit or Fisher’s exact test, as appropriate, for categorical variables. The crude correlation of MBDA score at each visit and the absolute change in MBDA between the baseline and the last visit with the change in MDS-TBRmax was assessed by calculating Spearman’s correlation coefficient. The MBDA score at the time point with the strongest correlation with the change in MDS-TBRmax was then modelled in a linear regression model with MDS-TBRmax as the dependent variable. The MBDA score was not linearly distributed at any time point and was transformed to normality with log transformation for all linear regression modelling. Next, an extended multivariable model was created by examining the associations between potential confounding covariates and the MBDA score using linear regression. Covariates associated with the MBDA score at the P ≤ 0.20 level were included in the multivariable model. Next, a reduced model was constructed by excluding non-contributory covariates using Akaike’s information criterion for nested models. Additional sensitivity models included using the categorical MBDA score (i.e. low, moderate, high) instead of the log-transformed continuous score and replacing the MBDA score with the CRP level. Throughout, the R2 statistic was used as a representation of the total variability in the change in MDS-TBRmax explained by the covariate(s) in the model.
Next, the effect of the MBDA score on MDS-TBRmax among several prespecified subgroups (i.e. randomized treatment allocation, sex, statin use and corticosteroid use) was explored by introducing subgroup-specific interaction terms into the final reduced model. Next, robust regression was used to model the associations of the log MBDA score with alternate measures of FDG uptake in the index vessel. Robust regression was used since the outcome variables were not able to be transformed adequately to meet the requirements of linear regression. Finally, the individual components of the MBDA score were modelled together in the reduced multivariable model with a change in MDS-TBRmax as the dependent variable. Spearman correlation coefficients were calculated between the MBDA components, and variance inflation factors were calculated to ensure that covariates with excess collinearity were not included together in the same model. Throughout, a two-tailed α of 0.05 was used. Stata/SE 16.1 (StataCorp, College Station, TX, USA) was used for all analyses.
Results
Baseline characteristics according to MBDA categories are summarized in Table 1. Among the 112 TARGET participants with complete FDG-PET/CT data and MBDA data, the median baseline MBDA score was 35 (ie in the moderate range). Thirty-two percent (n = 36) were classified as low MBDA at baseline, while 31% (n = 35) were classified as high MBDA. Compared with those with low MBDA, those with high MBDA were less likely to have smoked in the past and had significantly lower systolic blood pressure and total cholesterol. As expected, those with high MBDA had a significantly higher DAS28-CRP score than those with low MBDA, which was driven by higher CRP but not higher swollen or tender joint counts. Patient global assessment and HAQ were significantly higher among those with high vs low MBDA. CDAI, duration of morning stiffness, and RA therapies, both background and randomized, did not significantly differ by MBDA categories.
Table 1.
Baseline characteristics according to baseline multi-biomarker disease activity score categories
| Characteristic | Low MBDA | Moderate MBDA | High MBDA | P-valuea |
|---|---|---|---|---|
| (n = 36) | (n = 41) | (n = 35) | ||
| Age, mean (s.d.), years | 59.9 (6.1) | 60.3 (7.7) | 58.3 (9.5) | 0.39 |
| Female, n (%) | 22 (61) | 32 (78) | 26 (74) | 0.24 |
| White race, n (%) | 28 (78) | 30 (73) | 30 (86) | 0.39 |
| BMI, mean (s.d.), kg/m2 | 30.2 (6.5) | 29.6 (5.8) | 30.8 (5.5) | 0.68 |
| Past smoker, n (%) | 13 (36) | 11 (27) | 5 (14) | 0.035 |
| Current smoker, n (%) | 3 (8) | 3 (7) | 6 (17) | 0.31 |
| Never smokers, n (%) | 20 (56) | 27 (66) | 24 (69) | 0.26 |
| Diabetes, n (%) | 0 (0) | 1 (2) | 1 (3) | 0.49 |
| Hypertension, n (%) | 16 (44) | 18 (44) | 17 (48) | 0.73 |
| SBP, mean (s.d.), mm | 132 (14) | 131 (21) | 125 (13) | 0.031 |
| DBP, mean (s.d.), mm | 79 (10) | 77 (10) | 77 (7) | 0.36 |
| Hyperlipidaemia, n (%) | 8 (22) | 8 (20) | 7 (20) | 0.58 |
| Total cholesterol, mean (s.d.), mg/dl | 208 (38) | 213 (36) | 191 (30) | 0.049 |
| LDL-C mean (s.d.), mg/dl | 105 (25) | 111 (30) | 97 (22) | 0.16 |
| HDL-C mean (s.d.), mg/dl | 52 (21) | 61 (16) | 55 (14) | 0.53 |
| Triglycerides, median (IQR), mg/dl | 120 (87–175) | 106 (84–133) | 106 (70–143) | 0.29 |
| Statin use, n (%) | 10 (28) | 6 (15) | 6 (17) | 0.28 |
| Aspirin, n (%) | 9 (25) | 8 (20) | 11 (31) | 0.83 |
| RA duration, median (IQR), years | 1.1 (0.5–5.7) | 2.2 (0.4–5.3) | 1.9 (0.6–9.6) | 0.23 |
| RF or anti-CCP, n (%) | 20 (56) | 29 (71) | 23 (66) | 0.38 |
| DAS28-CRP, median (IQR), units | 4.4 (4.0–5.4) | 4.6 (3.7–5.1) | 5.3 (4.5–6.1) | 0.022 |
| Swollen joint count (28), median (IQR) | 7 (3–12) | 5 (2–8) | 7 (2–11) | 0.72 |
| Tender joint count (28), median (IQR) | 11 (5–16) | 8 (3–13) | 10 (6–15) | 0.87 |
| CRP, median (IQR), mg/l | 2.8 (1.0–3.9) | 2.7 (1.3–4.6) | 13.4 (7.6–28.9) | <0.001 |
| Patient global (0–100), median (IQR) | 55 (32–71) | 51 (42–74) | 60 (50–89) | 0.049 |
| CDAI, median (IQR), units | 31 (18–41) | 26 (17–35) | 30 (21–42) | 0.84 |
| HAQ, median (IQR), units | 0.88 (0.25–1.31) | 1.38 (0.50–1.62) | 1.50 (0.88–2.00) | 0.004 |
| AM Stiffness, median (IQR), min | 60 (30–90) | 45 (10–60) | 45 (15–120) | 0.74 |
| Randomized to TNFi, n (%) | 19 (53) | 20 (49) | 17 (49) | 0.72 |
| MTX dose mean (s.d.), mg/week | 20 (5) | 19 (6) | 20 (4) | 0.95 |
| Prednisone, n (%) | 11 (31) | 13 (32) | 12 (34) | 0.74 |
| NSAIDs, n (%) | 17 (47) | 17 (41) | 11 (31) | 0.23 |
Mean (s.d.) or median (interquartile range) depicted unless otherwise stated.
P-values are from statistical tests comparing the low vs high MBDA groups. CDAI: clinical disease activity index; DAS28-CRP: disease activity score for 28 joints with CRP; DBP: diastolic blood pressure; HAQ: health assessment questionnaire; HDL-C: high density lipoprotein concentration; IQR: interquartile range; LDL-C: low density lipoprotein concentration; MBDA: multi-biomarker disease activity; MTX: methotrexate; SBP: systolic blood pressure; TNFi: TNF inhibitor.
On average, the MBDA score decreased by 6 units (15%) between baseline and week 24 (95% CI: −9, −4). Although the MBDA decreased significantly in both treatment groups, the change was greater in the TNF inhibitor arm compared with the triple therapy group [−9 vs −3 units, respectively; P-value for group difference = 0.017 (data not shown)]. At week 24, 46% were classified as low MBDA, 37% as moderate MBDA and 17% as high MBDA. All of those with high MBDA at week 24 had started in either moderate or high MBDA at baseline (Supplementary Table S1, available at Rheumatology online).
The MBDA score was associated with the change in MDS-TBRmax
The MDS-TBRmax decreased from 2.67 units at baseline to 2.45 at follow-up, an average reduction of 8.2%. The MBDA score at baseline, week 6 and week 18, and the change in MBDA between baseline and week 24 were not significantly correlated with the change in MDS-TBRmax (Table 2). However, there was a significant correlation between the week 24 MBDA score with the change in MDS-TBRmax (Spearman’s rho = 0.239; P = 0.011).
Table 2.
Correlations of MBDA during the trial vs change in arterial inflammation
| Spearman’s rho | P-value | |
|---|---|---|
| MBDA at baseline | 0.087 | 0.36 |
| MBDA at week 6 | 0.092 | 0.36 |
| MBDA at week 18 | 0.051 | 0.60 |
| MBDA at week 24 | 0.239 | 0.011 |
| Average MBDA | 0.166 | 0.076 |
| Change in MBDA between baseline and week 24 | 0.139 | 0.15 |
Arterial inflammation was assessed as the maximal target-to-background ratio for the most diseased segment. MBDA: multi-biomarker disease activity.
The significant association of the week 24 MBDA score with the change in MDS-TBRmax was also observed in a linear regression model (Table 3, unadjusted model). Adjustment for relevant confounders (derived from Supplementary Table S2, available at Rheumatology online) did not alter the significant association of the MBDA at week 24 with the change in MDS-TBRmax (Table 3, extended and reduced multivariable models). Substituting the MBDA categories for log MBDA at week 24 revealed a non-linear association of MBDA at week 24 with the change in MDS-TBRmax (Table 3, alternative exposure model 1), such that low MBDA at week 24 was associated with a statistically significant difference of 0.25 TBR units compared with moderate MBDA and 0.27 TBR units compared with high MDBA (Fig. 1). After adjustment, those achieving low MBDA at week 24 had significantly decreased their MDS-TBRmax by an average of 0.35 TBR units. On the other hand, those with moderate or high MBDA at follow-up did not demonstrate significant reductions in arterial inflammation.
Table 3.
Multivariable linear regression models of change in arterial inflammation (MDS-TBRmax of the index vessel between baseline and week 24)
| Characteristic | Unadjusted model |
Extended MV model |
Reduced MV model |
Alt. exposure model 1 |
Alt. exposure model 2 |
|||||
|---|---|---|---|---|---|---|---|---|---|---|
| β | P-value | β | P-value | β | P-value | β | P-value | β | P-value | |
| Log MBDA at week 24, per unit | 0.27 | 0.020 | 0.25 | 0.032 | 0.27 | 0.017 | ||||
| MBDA categories at week 24 | ||||||||||
| Low disease activity | referent | — | ||||||||
| Moderate disease activity | 0.25 | 0.014 | ||||||||
| High disease activity | 0.27 | 0.039 | ||||||||
| Age, per year | −0.0093 | 0.14 | −0.0099 | 0.097 | −0.010 | 0.083 | −0.011 | 0.079 | ||
| Past smoker | 0.20 | 0.080 | 0.20 | 0.073 | 0.18 | 0.10 | 0.17 | 0.13 | ||
| Current smoker | −0.22 | 0.14 | −0.21 | 0.15 | −0.25 | 0.095 | −0.21 | 0.14 | ||
| Hypertension | −0.040 | 0.68 | ||||||||
| Seropositivity | 0.064 | 0.51 | ||||||||
| Randomized to TNFi | −0.028 | 0.77 | ||||||||
| MTX dose, per mg | −0.026 | 0.007 | −0.027 | 0.005 | −0.028 | 0.003 | −0.027 | 0.004 | ||
| Log CRP at week 24, per unit | 0.063 | 0.056 | ||||||||
| R 2 | 0.048 | 0.110 | 0.129 | 0.135 | 0.112 | |||||
β coefficients represent the change in the change in TBRmax MDS over 24 weeks per 1 unit higher patient characteristic, adjusting for the other covariates in the model. The R2 statistic represents the total explainable variability in TBRmax MDS accounted for by the covariates in the model. Alt.: alternative; MBDA: multi-biomarker disease activity: methotrexate; MV: multivariable; TBRmax MDS: maximal target-to-background ratio for the most diseased segment; TNFi: TNF inhibitor; MTX.
Figure 1.
Associations of categories of the MBDA score at week 24 with the change in 18F-FDG-PET detected arterial inflammation (MDS-TBRmax). Crude and adjusted means and 95% CIs are presented. Adjusted for age, current and former smoker, and methotrexate dose. FDG: fluorodeoxyglucose; HDA: high disease activity; LDA: low disease activity; MBDA: multi-biomarker disease activity score; MDA: moderate disease activity; MDS-TBRmax: maximal target-to-background ratio of the most diseased arterial segment; ns: non-significant
Substituting log CRP at week 24 for the MBDA at week 24 resulted in a non-significant association of log CRP with MDS-TBRmax (Table 3 alternative exposure model 2). The association of log MBDA at week 24 with the change in MDS-TBRmax did not significantly differ according to the randomized treatment, sex, statin use or use of corticosteroids (Table 4). The log MBDA at week 24 was not significantly associated with other arterial FDG uptake outcomes of the index vessel (Supplementary Table S3, available at Rheumatology online). Neither the MBDA CVD score at baseline nor that at week 24 was associated with the change in MDS-TBRmax (data not shown).
Table 4.
Stratified analyses: adjusteda associations of the multi-biomarker disease activity score with change in arterial inflammation between baseline and week 24 among subgroups
| Log MBDA at week 24 |
Interaction | Log average MBDA |
Interaction | |||
|---|---|---|---|---|---|---|
| β | P-value | P-value | β | P-value | P-value | |
| Triple therapy (n = 57) | 0.20 | 0.23 | 0.54 | 0.23 | 0.22 | 0.41 |
| TNFi (n = 58) | 0.33 | 0.035 | 0.45 | 0.021 | ||
| Women (n = 82) | 0.29 | 0.029 | 0.84 | 0.42 | 0.007 | 0.31 |
| Men (n = 33) | 0.25 | 0.28 | 0.11 | 0.67 | ||
| Statin use (n = 22) | 0.32 | 0.13 | 0.78 | 0.47 | 0.048 | 0.48 |
| No statins (n = 93) | 0.26 | 0.056 | 0.28 | 0.081 | ||
| Corticosteroids (n = 38) | 0.23 | 0.23 | 0.81 | 0.36 | 0.11 | 0.89 |
| No corticosteroids (n = 77) | 0.29 | 0.038 | 0.33 | 0.046 | ||
Adjusted for age, past smoker, current smoker and methotrexate dose. β coefficients represent the change in the change in MDS-TBRmax over 24 weeks per 1 unit higher patient characteristic, adjusting for the other covariates in the model. Alt.: alternative; MBDA: multi-biomarker disease activity; MDS-TBRmax: maximal target-to-background ratio for the most diseased segment; MV: multivariable; TNFi: TNF inhibitor.
Association of MBDA components with change in MDS-TBRmax
The correlation of the component analytes of the MBDA score at week 24 with the total score ranged from a Spearman’s rho of −0.374 (EGF at week 24 with MBDA at week 24) to Spearman’s rho of 0.764 (CRP at week 24 with MBDA at week 24) (Supplementary Table S4, available at Rheumatology online). There was also the variable correlation of individual components with each other. However, the calculation of variance inflation factors did not reveal sufficient collinearity to prohibit co-modelling the individual components of the MBDA, as the mean variance inflation factor (VIF) was 1.57 and the highest VIF was 2.78 (data not shown). When included in the same model, the MBDA component analyte with the strongest independent association with MDS-TBRmax was SAA (Supplementary Table S5, available at Rheumatology online). Shown graphically (Fig. 2A), those in the lowest tertile of SAA at week 24 had an adjusted change in MDS-TBRmax of −0.374 TBR units, which was significantly greater than no change compared with those in the highest tertile, in which the change in MDS-TBRmax was only −0.085 TBR units and was not significantly different from no change. This association was more pronounced among those achieving articular low disease activity (LDA) at week 24 (Fig. 2B), in which those in the lowest tertile of SAA at week 24 who had achieved articular LDA had a change in MDS-TBRmax from baseline of −0.450 TBR units compared with only −0.019 TBR units among those in the highest tertile of SAA who had also achieved articular LDA. In contrast, there was no association of SAA with change in MDS-TBRmax among those who had not attained articular LDA (P-value for interaction = 0.067). A similar interaction was not observed when CRP at week 24 was substituted for SAA at week 24 (Fig. 2C).
Figure 2.
Associations of categories of tertiles of SAA (A, B) and CRP (C) at week 24 with the change in 18F-FDG PET detected arterial inflammation (MDS-TBRmax). Crude and adjusted means and 95% CIs are presented. Adjusted for age, current and former smoker, and methotrexate dose throughout. FDG: fluorodeoxyglucose; LDA: low disease activity; MBDA: multi-biomarker disease activity score; MDS-TBRmax: maximal target-to-background ratio of the most diseased arterial segment; ns: non-significant; SAA: serum amyloid A
Discussion
In the context of a clinical trial of two standard RA treatment strategies, we observed a statistically significant association of the MBDA score with the change in arterial inflammation, as represented by the change in 18F-FDG uptake in the aorta and carotid arteries, such that those who achieved low MBDA had a markedly greater reduction in arterial inflammation compared with those who did not achieve low MBDA, even after accounting for relevant confounders. Notably, neither the DAS28-CRP nor its components were significantly associated with the change in arterial inflammation to the extent that the MBDA score was. There were no significant subgroup differences based on randomized treatment group, sex, statin treatment or concomitant corticosteroids. SAA at week 24 was the component of the MBDA score with the strongest association with the change in arterial inflammation, such that those who achieved the lowest SAA levels by the end of the trial had achieved the greatest reduction in arterial inflammation.
The immunopathologies of RA and atherosclerosis share multiple common pathways, with many of the inflammatory cytokines, chemokines, growth factors and adipokine mediators that are typically elevated in both conditions and also mediators of disease activity in the synovium and arterial wall [13]. Each of the components of the MBDA score has been implicated, either directly or indirectly, in the pathways underlying atherogenesis and/or atherothrombosis [10]. Thus, our finding that the MBDA score was associated with arterial inflammation was not unexpected and was hypothesized a priori. The MBDA score was also a stronger predictor of arterial inflammation than CRP in our study, potentially due to the representation of the broad array of inflammatory mediators relevant to arterial inflammation that are part of the MBDA score. And although the MBDA score and CRP were correlated with each other, they were not so highly correlated as to be interchangeable in their associations with the change in arterial inflammation.
Neither the MBDA score at baseline nor the change in MBDA score was as strongly associated with the change in arterial inflammation as the MBDA score at the time of the final scan. The groups with the smallest decrease in arterial inflammation were those who ended the trial in moderate or high MBDA despite treatment. Interestingly, the change in articular disease activity, as represented by the DAS28-CRP score or its components, was not associated with the change in arterial inflammation (as reported in Solomon et al. [3]). This raises the possibility that the inflamed tissues of RA patients, such as synovium and artery wall, may respond differently to treatments, with different biomarkers predicting response differently, depending on the tissue.
On average, MDS-TBRmax decreased by 0.22 TBR units (−8.2%) for the total study population, which is consistent with the effect observed with statin therapy [14]. However, this reduction was not uniform, as illustrated by the fact that those who did not attain low MBDA did not experience the same reduction in MDS-TBRmax. In contrast, the mean reduction observed in those who did achieve or remained in low MBDA was 0.35 TBR units (−13%), consistent with the reduction observed with high-intensity statin therapy [14]. Although our study does not prove that observing the same reduction in arterial inflammation as high-intensity statin use would translate to the same CVD event and mortality benefits conveyed by statins, these data do provide support for the premise that RA pharmacotherapies may provide similar cardioprotective benefits to statins in the context of an effective overall systemic immunological response.
SAA was the component of the MBDA score with the strongest association with arterial inflammation in our study, and multiple lines of evidence support a causative role for SAA in atherogenesis and atherothrombosis. Atherosclerosis-prone mice that overexpress SAA have a greater burden of atherosclerosis [15], and suppression of SAA was associated with reduced atherogenesis [16]. SAA is detectable in atherosclerotic plaques of humans [17], and has been shown to activate the NLRP3 inflammasome in macrophages [18], a mechanism known to be proatherogenic. SAA was shown to upregulate the expression of vascular adhesion molecules on the endothelium leading to endothelial activation [19, 20], and effects of SAA on tissue factor may confer additional prothrombotic potential [21]. Additionally, as a component of the lipoprotein cargo of high-density lipoprotein (HDL), SAA may interfere with the anti-atherogenic functions of HDL, such as reverse cholesterol transport [22]. In RA, SAA is an acute phase reactant with levels that are correlated with articular disease activity, CRP, and circulating levels of many inflammatory cytokines of RA [23]. Our finding that a higher SAA level after the study was independently associated with no improvement in vascular inflammation, even in the setting of attaining articular LDA, suggests a unique contribution of SAA on the vasculature of RA patients that warrants additional mechanistic study. Next steps in evaluating SAA as a CVD predictor include external validation of its performance as a predictor of other intermediate CVD outcomes, and ultimately, of CVD events. SAA could also represent a target for therapeutics with the potential to reduce CVD risk in RA or other inflammatory conditions. However, due to its correlation with multiple inflammatory pathways with atherogenic potential, it could also be a bystander with no direct causative effect.
Notable strengths of the study include the use of a validated and standardized multi-biomarker that includes analytes relevant to both RA and atherosclerosis. The study occurred in the context of a carefully conducted clinical trial in which RA disease activity and the MBDA score were measured at multiple time points.18F-FDG-PET/CT is a well-accepted surrogate that corresponds histologically with vulnerable and inflamed plaques [4] and predicts future CVD events [5]. However, it is an intermediate outcome for CVD risk, and it is unknown whether arterial inflammation detected on FDG-PET/CT correlates with CVD events in RA patients. Among limitations, there were several scans for which the TBR could not be reliably measured. However, there were no clinical or serological differences between the patients with and without interpretable scans. In addition, the previously published MBDA CVD score did not correlate with the change in arterial inflammation. This could be due to the differences in the outcomes and patient groups studied, as the MBDA CVD score was derived from a real-world population with CVD events as the outcome [11], while our study used a clinical trial population entering with moderate to high RA disease activity with arterial inflammation as the outcome. Both may be valid for assessing differing forms of CVD risk in RA patients, and additional study into the clinical utility of using the MBDA score or its components for CVD prediction and risk stratification in RA is warranted.
In summary, achieving or remaining in low disease activity by the MBDA at 24 weeks was associated with a clinically meaningful reduction in arterial inflammation, like high-intensity statins, in a way not predicted using other RA disease activity measures, suggesting that treatment-associated improvements in arterial inflammation may be indicated by specific biomarkers that overlap those used to track articular disease activity.
Supplementary material
Supplementary material is available at Rheumatology online.
Supplementary Material
Acknowledgements
The TARGET Trial Consortium includes Yousef Ali, Joshua Baker, Marcy Bolster, Vivian Bykerk, Christina Charles-Schoeman, Cong-Qiu Chu, Stanley Cohen, Jeffrey Curtis, Jack Cush, Christina Downey, Margarita Fallena, Nazanin Firooz, Brigid Freyne, Jonathan Graf, Maria Greenwald, Diane Horowitz, Elaine Husni, Rajesh Kataria, Edward Keystone, Alan Kivitz, Joel Kremer, Robert Levin, Kristine Lohr, Elena Massarotti, Alan Matsumoto, Philip Mease, Barbara Mendez, Jeffrey Miller, Larry Moreland, Birh Nguyen, Deborah Parks, William Rigby, Jose Scher, Elena Schiopu, Beth Scholz and Guillermo Valenzuela. Scientific and financial support for the Foundation for the NIH Biomarkers Consortium TARGET Biomarker Study were made possible through grants, direct and in-kind contributions provided by: Amgen Inc., The Arthritis Foundation, Merck & Co., Regeneron Pharmaceuticals, Inc., Laboratory Corporation of America and Rules Based Medicine, Inc.
Contributor Information
Jon T Giles, Division of Rheumatology, Cedars-Sinai Medical Center, Los Angeles, CA, USA.
Daniel H Solomon, Division of Rheumatology, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA.
Katherine P Liao, Division of Rheumatology, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA.
Pamela M Rist, Division of Preventive Medicine, Department of Medicine, Harvard Medical School and Brigham and Women's Hospital, Boston, MA, USA.
Zahi A Fayad, BioMedical Engineering and Imaging Institute, Icahn School of Medicine at Mount Sinai, New York, NY, USA.
Ahmed Tawakol, Cardiology Division, Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA.
Joan M Bathon, Division of Rheumatology, Cedars-Sinai Medical Center, Los Angeles, CA, USA.
Data availability
All data from the trial will be available through dbGaP (https://www.ncbi.nlm.nih.gov/gap/).
Funding
NIH-NIAMS U01-AR068043; Abbvie and Amgen supplied the study drug; Crescendo Biosciences funded and performed the assays for the MBDA testing. Scientific and financial support was supplied by the Foundation for the NIH.
Disclosure statement: D.H.S. receives research support through his institution from Abbvie, Amgen, CorEvitas, Janssen, and Moderna. He receives royalties from UpToDate on unrelated chapters. J.T.G. has been a consultant for AbbVie, Pfizer, Eli Lilly and Company, Bristol Meyers Squibb, Gilead, Novartis in the last 3 years and received an unrestricted grant from Pfizer. The remaining authors have declared no conflicts of interest.
References
- 1. Avina-Zubieta JA, Choi HK, Sadatsafavi M. et al. Risk of cardiovascular mortality in patients with rheumatoid arthritis: a meta-analysis of observational studies. Arthritis and Rheumatism 2008;59:1690–7. [DOI] [PubMed] [Google Scholar]
- 2. Kinlay S, Selwyn AP, Libby P, Ganz P.. Inflammation, the endothelium, and the acute coronary syndromes. Journal of cardiovascular pharmacology 1998;32(Suppl 3):S62–S6. [PubMed] [Google Scholar]
- 3. Solomon DH, Giles JT, Liao KP, TARGET Trial Consortium et al. Reducing cardiovascular risk with immunomodulators: a randomised active comparator trial among patients with rheumatoid arthritis. Ann Rheum Dis 2023;82:324–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Tawakol A, Migrino RQ, Bashian GG. et al. In vivo 18F-fluorodeoxyglucose positron emission tomography imaging provides a noninvasive measure of carotid plaque inflammation in patients. J Am Coll Cardiol 2006;48:1818–24. [DOI] [PubMed] [Google Scholar]
- 5. Figueroa AL, Abdelbaky A, Truong QA. et al. Measurement of arterial activity on routine FDG PET/CT images improves prediction of risk of future CV events. JACC Cardiovasc Imaging 2013;6:1250–9. [DOI] [PubMed] [Google Scholar]
- 6. Moon SH, Cho YS, Noh TS. et al. Carotid FDG uptake improves prediction of future cardiovascular events in asymptomatic individuals. JACC Cardiovasc Imaging 2015;8:949–56. [DOI] [PubMed] [Google Scholar]
- 7. Centola M, Cavet G, Shen Y. et al. Development of a multi-biomarker disease activity test for rheumatoid arthritis. PLoS One 2013;8:e60635. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Curtis JR, van der Helm-van Mil AH, Knevel R. et al. Validation of a novel multibiomarker test to assess rheumatoid arthritis disease activity. Arthritis Care Res (Hoboken) 2012;64:1794–803. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Curtis JR, Weinblatt ME, Shadick NA. et al. Validation of the adjusted multi-biomarker disease activity score as a prognostic test for radiographic progression in rheumatoid arthritis: a combined analysis of multiple studies. Arthritis Res Ther 2021;23:1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Tedgui A, Mallat Z.. Cytokines in atherosclerosis: pathogenic and regulatory pathways. Physiol Rev 2006;86:515–81. [DOI] [PubMed] [Google Scholar]
- 11. Curtis JR, Xie F, Crowson CS. et al. Derivation and internal validation of a multi-biomarker-based cardiovascular disease risk prediction score for rheumatoid arthritis patients. Arthritis Res Ther 2020;22:282. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Giles JT, Rist PM, Liao KP. et al. Testing the effects of disease-modifying antirheumatic drugs on vascular inflammation in rheumatoid arthritis: rationale and design of the TARGET trial. ACR Open Rheumatol 2021;3:371–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Raggi P, Genest J, Giles JT. et al. Role of inflammation in the pathogenesis of atherosclerosis and therapeutic interventions. Atherosclerosis 2018;276:98–108. [DOI] [PubMed] [Google Scholar]
- 14. Tawakol A, Fayad ZA, Mogg R. et al. Intensification of statin therapy results in a rapid reduction in atherosclerotic inflammation: results of a multicenter fluorodeoxyglucose-positron emission tomography/computed tomography feasibility study. J Am Coll Cardiol 2013;62:909–17. [DOI] [PubMed] [Google Scholar]
- 15. Dong Z, Wu T, Qin W. et al. Serum amyloid A directly accelerates the progression of atherosclerosis in apolipoprotein E-deficient mice. Mol Med 2011;17:1357–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Thompson JC, Wilson PG, Shridas P. et al. Serum amyloid A3 is pro-atherogenic. Atherosclerosis 2018;268:32–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Meek RL, Urieli-Shoval S, Benditt EP.. Expression of apolipoprotein serum amyloid A mRNA in human atherosclerotic lesions and cultured vascular cells: implications for serum amyloid A function. Proc Natl Acad Sci U S A 1994;91:3186–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Shridas P, De Beer MC, Webb NR.. High-density lipoprotein inhibits serum amyloid A-mediated reactive oxygen species generation and NLRP3 inflammasome activation. J Biol Chem 2018;293:13257–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Nishida E, Aino M, Kobayashi S-I. et al. Serum Amyloid A Promotes E-Selectin Expression via Toll-Like Receptor 2 in Human Aortic Endothelial Cells. Mediators Inflamm 2016;2016:7150509. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Vallejo A, Chami B, Dennis JM. et al. NFkappaB inhibition mitigates serum amyloid A-induced pro-atherogenic responses in endothelial cells and leukocyte adhesion and adverse changes to endothelium function in isolated aorta. Int J Mol Sci 2018;20:105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Page MJ, Thomson GJA, Nunes JM. et al. Serum amyloid A binds to fibrin(ogen), promoting fibrin amyloid formation. Sci Rep 2019;9:3102. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Van Lenten BJ, Hama SY, de Beer FC. et al. Anti-inflammatory HDL becomes pro-inflammatory during the acute phase response. Loss of protective effect of HDL against LDL oxidation in aortic wall cell cocultures. J Clin Invest 1995;96:2758–67. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Zhou J, Dai Y, Lin Y, Chen K.. Association between serum amyloid A and rheumatoid arthritis: a systematic review and meta-analysis. Semin Arthritis Rheum 2022;52:151943. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
All data from the trial will be available through dbGaP (https://www.ncbi.nlm.nih.gov/gap/).