Tumor-necrosis Factor Inhibitors Improve Aortic Stiffness in Patients With Longstanding Rheumatoid Arthritis

Background. Major cardiovascular disease (CVD) benets of disease-modifying anti-rheumatic drugs (DMARDs) therapy occur in early RA patients with treat-to-target strategy. However, it is unknown whether long-term DMARDs treatment in established RA could be useful to improve CVD risk prole. Methods. Ultrasound aortic stiffness index (AoSI) has to be considered a proxy outcome measure in established RA patients. We measured AoSI in a group of RA patients on long-term treatment with tumour necrosis factor inhibitors (TNFi) or conventional synthetic DMARDs (csDMARDs). Eligible participants were assessed at baseline and after 12 months; changes in serum lipids, glucose and arterial blood pressure were assessed. All patients were on stable medications during the entire follow-up. Results. We included 107 (64 TNFi and 43 csDMARDs) RA patients. Most patients (74%) were in remission or low disease activity and had some CVD risk factors (45.8% hypertension, 59.8% dyslipidemia, 45.3% smoking). The two groups did not differ signicantly for baseline AoSI (5.95±3.73% vs 6.08±4.20%, p=0.867). Follow-up AoSI was signicantly increased from baseline in the csDMARDs group (+1.00%; p<0.0001) but not in the TNFi group (+0.15%, p=0.477). Patients on TNFi had signicantly lower follow-up AoSI from baseline than the csDMARD group (-1.02%, p<0.001; ANCOVA corrected for baseline AoSI, age and systolic blood pressure). Furthermore, follow-up AoSI was signicantly lower in TNFi users with 1-2 or >2 CVD risk factors than in those without. Conclusion. Long-term treatment with TNFi was associated with reduced aortic stiffness in patients with established RA and several CVD duration, body weight and height, medical history and CV and RA medications. Recruited patients were then referred for aortic stiffness assessment that was performed within 2 weeks. Laboratory tests including inammatory markers, serology, lipids and glucose levels were performed within 2 weeks before or after aortic stiffness assessment. Follow-up assessments at 12 months were performed between March 2015 and March 2017. Patients who had started CV or RA medications within 6 weeks from rst assessment were excluded from this analysis. Rheumatoid arthritis-associated factors. Data on disease duration, anti citrullinated peptides antibodies (ACPA) and rheumatoid factor (RF) were collected. Serum biomarkers of RA-related inammation (C-reactive protein CRP, and ESR) were measured. RA disease activity was evaluated by the clinical disease activity index (CDAI) score and disease-activity score in 28-joints (DAS28). Patients were dened as having remission, low, moderate or high disease activity according to CDAI values. Current immunomodulating agents including conventional synthetic DMARDs and biologic DMARDs, glucocorticoid use and dose (in prednisone-equivalent milligrams daily), and NSAIDs use were recorded. Intraclass correlation coecient (ICC) with two-way random model was used to absolute reliability and measurement in 50 patients. ICC values (95% CI) were 0.91 (0.86– 0.94) for AoS, 0.93 for AoD, 0.92 for SBP and 0.94 for DBP respectively. ICC for calculated AoSI was 0.92. TNFi group. We found no signicant correlations between changes in AoSI and serum lipids, glucose, or arterial blood pressure. disease activity score c-reactive


Introduction
Rheumatoid arthritis (RA) is a chronic immune-mediated and in ammatory disease characterized by a 48% increased risk of cardiovascular (CV) events and a 50% higher incidence of cardiovascular disease (CVD)-related mortality compared with the general population. (1,2) There is growing evidence that increased arterial stiffness may account for the excess risk of CVD in RA. (3)(4)(5)(6)(7)(8) Arterial stiffness is one of the earliest detectable manifestations within the atherosclerotic vessel wall (9,10), and it acts as a strong independent predictor of CV events and all-cause mortality in various populations (11). Aortic stiffness was signi cantly increased in RA patients (12)(13)(14), and it was associated with worse CVD outcomes (15). Interestingly, treatment with conventional synthetic diseasemodifying anti-rheumatic drugs (csDMARDs) or tumor-necrosis factor inhibitors (TNFi), appeared to be the effective strategy to improve aortic stiffness in early RA patients. (12,16,17) Most clinical trials have been successful at demonstrating a bene cial effect of csDMARDs and TNFi on CV outcomes in RA of short duration, (18)(19)(20)(21) when CVD risk pro le is still favorable and in ammation is at its highest. However, patients with RA are subject to great accumulation of CVD risk factors in a disproportionate manner than the general population and this can happen even when RA patients receive long-term therapy with good outcomes in terms of disease activity control. (21)(22)(23)(24) In such patients with longstanding and established disease, whether csDMARDs and TNFi can still have an effect on aortic stiffness is largely unknown. This knowledge could encourage retention of csDMARDs or TNFi for their CV bene t beyond the control of in ammation. The aim of this study was to comparatively describe aortic stiffness progression in such RA patients treated with csDMARDs and TNFi.

Methods
Study protocol. This was a post-hoc analysis of a wider cross-sectional study of cardiovascular assessment of non-institutionalized individuals > 18 years of age affected with rheumatic and musculoskeletal diseases, which was started in 2014 at the Division of Rheumatology, Department of Medicine, University and Azienda Ospedaliera Universitaria Integrata of Verona (Italy). The study was approved by the institutional review board of the University of Verona (1707CESC) and conformed to the ethical guidelines of the Declaration of Helsinki as revised in 2000. All patients gave written informed consent signing a speci c institutional consent form. Inclusion criteria comprised a diagnosis of RA according to the 2010 ACR/EULAR de nition, or psoriatic arthritis or spondyloarthritis according to international criteria. The following exclusion criteria were applied: previous CVD diagnoses, events, and procedures (any known CVD including myocardial infarction, stroke, coronary revascularization, transient ischemic attack, hospitalization for unstable angina, peripheral artery disease, symptomatic carotid artery disease), and uncontrolled systemic arterial hypertension). All participants underwent an evaluation of cardiovascular risk factors and were offered echocardiography examination, carotid ultrasound and aortic stiffness assessment. Follow-up and instrumental assessments were performed yearly. Since this was an observational study, any pharmacological or non-pharmacological interventions during follow-up both for RA or cardiovascular risk management were driven by local practice only.
Patient disposition. Participants were consecutively screened and recruited from March 2014 to March 2016. All recruited patients underwent a clinical evaluation by senior rheumatologists (OV, DG, LI) including assessments of disease activity, disease duration, body weight and height, medical history and CV and RA medications. Recruited patients were then referred for aortic stiffness assessment that was performed within 2 weeks. Laboratory tests including in ammatory markers, serology, lipids and glucose levels were performed within 2 weeks before or after aortic stiffness assessment. Follow-up assessments at 12 months were performed between March 2015 and March 2017. Patients who had started CV or RA medications within 6 weeks from rst assessment were excluded from this analysis.
Cardiovascular disease risk factors. The following CVD risk factors were collected: age; gender; systolic blood pressure, diastolic blood pressure and heart rate; body mass index (BMI); lipids including total cholesterol, low density cholesterol and high density cholesterol, and triglycerides; waist circumference; renal function. We de ned obesity when body mass index (BMI) ≥ 30 kg/m 2 . Dyslipidemia was de ned as levels of total serum cholesterol > 190 mg/dL and or triglycerides > 150 mg/dL or pharmacologically treated high lipid serum levels. To assess renal function, we considered the glomerular ltration rate (GFR) estimated with the CKD-EPI equation and de ned renal dysfunction as estimated GFR < 60 ml/min 1.73 m 2 .
Rheumatoid arthritis-associated factors. Data on disease duration, anti citrullinated peptides antibodies (ACPA) and rheumatoid factor (RF) were collected. Serum biomarkers of RA-related in ammation (Creactive protein CRP, and ESR) were measured. RA disease activity was evaluated by the clinical disease activity index (CDAI) score and disease-activity score in 28-joints (DAS28). Patients were de ned as having remission, low, moderate or high disease activity according to CDAI values. Current immunomodulating agents including conventional synthetic DMARDs and biologic DMARDs, glucocorticoid use and dose (in prednisone-equivalent milligrams daily), and NSAIDs use were recorded.
Aortic stiffness assessment. Aortic stiffness was evaluated by Doppler-echocardiography. All Dopplerechocardiographic studies were performed by an expert sonographer (FO) using a Alpha Esaote Biomedica machine (Florence, Italy) equipped with a 2.5-3.5 MHz annular-array transducer and following a standardized protocol. Images were stored on compact disks or magneto-optical disks and forwarded for nal interpretation to a senior cardiologist (GC) blinded to the identity of the subject. Aortic stiffness was assessed at the level of the aortic root, using a two-dimensional guided M-mode evaluation of systolic (AoS) and diastolic (AoD) aortic diameters, 3 cm above the aortic valve together with blood pressure measured by cuff sphygmomanometer. AoD was obtained at the peak of the R wave at the simultaneously recorded electrocardiogram, while AoS was measured at the maximal anterior motion of the aortic wall (25,26); for each diameter ve measurements were averaged. Values of SBP, DBP, AoS and AoD were used to calculate the aortic stiffness index (AoSI) using the following validated formula; Blood pressure was measured at the end of echocardiographic evaluation in supine position: Intraclass correlation coe cient (ICC) with two-way random model was used to assess the absolute reliability of aortic diameters and BP measurement in 50 patients. ICC values (95% CI) were 0.91 (0.86-0.94) for AoS, 0.93 for AoD, 0.92 for SBP and 0.94 for DBP respectively. ICC for calculated AoSI was 0.92. Statistical analysis. Continuous data are reported as mean values ± standard deviation (SD) or median [25th, 75th percentile] for non-normally distributed variables, or absolute numbers (percentage) for categorical variables. Treatment group comparisons of categorical variables were performed by chisquared or Fisher test as appropriate; for continuous variables, independent samples T-test was used. Paired samples T-test was used to determine signi cant changes from baseline of continuous variables including AoSI, arterial blood pressure, lipids, glucose, in ammatory markers and RA disease activity scores. Treatment group comparisons of follow-up AoSI were performed in the whole study population using two-way analysis of covariance (ANCOVA) with Sidak's correction for multiple comparisons, with treatment group and number of CVD risk factors categorized into three groups (0-1, 2-3, or > 3) as factors, and baseline AoSI, age and SBP as covariates. The choice of covariates was made upon prior data from our study. (15) All analyses were performed using the statistical package SPSS 22.0 (SPSS Inc. Chicago. Illinois), and statistical signi cance was identi ed by two-tailed p < 0.05. Figures were obtained using the GraphPad Prism software version 7.00.

Results
Baseline characteristics of csDMARDs and TNFi patients The study population consisted of 107 white RA individuals, 43 patients in the csDMARDs group and 67 in the TNFi group. All patients had established RA and disease duration longer than 2 years. Most patients (74%) were in remission or low-disease activity, while disease activity was moderate only in 26% and high in none. High values of ESR (> 40 mm/h) or CRP (> 10 mg/L) were found in 11.8% and 8% only, respectively. Excluding age and sex, 92% of RA patients had at least one CVD risk factor, 58% two or more and 26% three or more. Patients in the csDMARDs and TNFi groups were equally balanced for the proportion of CVD risk factors and medications, and there were no signi cant differences in baseline SBP, DBP, or serum lipids. With regard to RA characteristics, the two groups differed only for a greater use of hydroxychloroquine in the csDMARD group (Table 1). Decreased aortic stiffness with TNFi compared to csDMARDs The two groups did not differ signi cantly for baseline AoSI (5.95 ± 3.73% vs 6.08 ± 4.20%, p = 0.867). However, follow-up AoSI was signi cantly increased in the csDMARDs group (mean difference 1.00%, 95% CI 0.59, 1.42; p < 0.0001) but not in the TNFi group (mean difference 0.15%, 95% CI -0.28, 0.60, p = 0.477). Patients on TNFi had signi cantly lower follow-up AoSI than the csDMARD group (aMD − 1.02, 95% CI -1.581, -0.457, p < 0.001; ANCOVA corrected for baseline AoSI, age and SBP).

Interaction of treatment and CVD risk factors on aortic stiffness
There was a statistically signi cant two-way interaction between treatment group and number of CVD risk factors on AoSI at follow-up, whilst controlling for baseline AoSI, age, and SBP (p < 0.0001, η 2 = 0.038). Follow-up AoSI was lower in the TNFi compared to csDMARDs group (Fig. 1) both when CVD risk factors were 1-2 (adjusted mean difference, aMD − 1.143, 95% CI -2.102, -0.185, p = 0.019) and when CVD risk factors were two or more (aMD − 4.806, 95% CI -6.128, -3.484, p < 0.001). We also compared the effect on aortic stiffness of TNFi and csDMARDs therapy across RA patients according to the presence of the most prevalent CVD risk factors in our study population: hypertension, dyslipidemia, and smoking. Adjusted AoSI means at follow-up were signi cantly higher than baseline in the csDMARDs group but not in the TNFI group (Fig. 2).
Changes in lipids, glucose and blood pressure induced by DMARD therapy.
Overall, favorable changes in lipids and glucose after DMARD therapy were found (Table 2). There was a signi cant reduction in TC, VLDL, LDL and HDL in the csDMARDs group. SFG was reduced both in the csDMARDs and TNFi group. However, blood pressure (both SBP and DBP) was signi cantly increased in the csDMARDs group, whereas DBP was signi cantly decreased in the TNFi group. We found no signi cant correlations between changes in AoSI and serum lipids, glucose, or arterial blood pressure.

Discussion
The original nding of this study is showing that arterial stiffness progression can be hampered by TNFi not only in early, but even in long-standing RA. Those individuals show a greater number of CVD risk factors than early RA patients (27) and higher CVD mortality than the general population (1). Hence, control of CVD risk in such patients is the most important outcome to achieve along with control of disease activity.
Prior studies on the effect of TNFi on arterial stiffness focused on early RA patients with short disease duration and relatively low CVD risk. We noticed favorable effects in terms of reduction of aortic stiffness with TNFi compared to csDMARDs in a cohort of patients with several CVD risk factors. TNFi can reduce endothelial dysfunction and reduce carotid intima-media thickness (28). Skin microvascular responses assessed by laser Doppler imaging improved in patients with active RA and no previous history of CVD who responded to TNFi or MTX (29). Short-term treatment with TNFi also increased circulating endothelial progenitor cells concurrently with a proportional decrease of disease activity (30). Our results, along with the previous evidence, are consistent with the hypothesis that the vascular-protective effect could be effectively achieved by inhibition of TNF (31).
The effect of TNFi on arterial stiffness in RA was deemed to be independent of the reduction of systemic in ammation in patients with very high disease activity (17). Herein, we excluded patients with high disease activity and provided further evidence that the bene cial effect of TNF-alpha inhibition on arterial stiffness goes beyond the DMARDs-associated reduction of systemic in ammation. Indeed, we also found no association between in ammatory markers or disease activity scores and aortic stiffness.
However, the relationship between disease activity and AoSI was di cult to ascertain as all changes in disease activity scores were subtle and non-clinically meaningful, as they did not lead to treatment changes per study inclusion criteria.
The third result of our study is that TNFi may be more bene cial than csDMARDs in presence of some CVD risk factors such as hypertension, dyslipidemia and smoking. Although traditional CV risk factors alone do not explain the heightened risk of CVD in RA (32), a meta-analysis con rmed hypertension, type 2 DM, smoking, and hypercholesterolemia as key traditional factors increasing the risk of CVD in RA (33). Hence, we analysed changes of aortic stiffness according to the presence of each of these CVD risk factors, except for DM due to the scarcity of patients with DM in our study population.
There was a sharp cardiovascular bene t of TNFi over csDMARDs in hypertensive RA patients. Moreover, TNFi therapy signi cantly decreased DBP values at follow up, while SBP and DBP were both increased in the csDMARDs group. Essential hypertension was reported in up to 57% of patients with RA, and can predict CV events (HR 3.67, 95% CI 2.0, 6.4, p = 0.001) (34). Several small studies support the potential BP-lowering effect of TNFi in RA patients (35). Nonetheless, in a US epidemiological study of RA patients, treatment with TNFi did not reduce the risk of incident hypertension compared with non-bDMARDs (36). Interestingly, we showed that TNFi decreased AoSI and DBP also in normotensive RA patients, suggesting that the main driver of decreased BP is the TNFi-mediated favorable effect on arterial stiffness.
Patients with RA and dyslipidemia on TNFi also showed reduced arterial stiffness. Moreover, one year of therapy with TNFi did not increase blood lipids, a nding that is in line with a meta-analysis of 25 RCTs of patients with chronic in ammatory arthritis that failed to demonstrate an effect of TNFi on TC, HDL-C, and LDL-C (37). Similar results were obtained by a recent RCT investigating the cardiovascular safety of tocilizumab against etanercept (38). Conversely, there was signi cant reduction of lipids with csDMARDs despite worse results on progression of aortic stiffness, suggesting that arterial stiffness in RA may be scarcely associated with serum lipid levels. This nding can be partially explained by the higher number of patients taking HCQ in the csDMARD group. Although HCQ confers limited e cacy on disease activity and progression of RA, HCQ increases HDL and reduces levels of TC, LDL-C, and triglycerides (39). Additionally, we noticed decreased glucose across treatment groups, consistent with the lower incidence of diabetes with the use of HCQ (39,40) or TNFi (40) among RA patients.
To our knowledge, this is the rst study to demonstrate decreased arterial stiffness with TNFi therapy in RA patients who smoked. Cigarette smoking is the strongest known lifestyle or environmental risk factor for RA (23,(41)(42)(43) and RA treatment failure (44). Moreover, smoking can damage the vascular wall, possibly leading to impaired prostacyclin production and enhanced platelet-vessel wall interactions (45).
This can reduce the elastic properties of the aorta, resulting in stiffening of and trauma to the wall (46).
Study limitations and strengths. The main strength of this study consists of including a real-life cohort of RA patients with longstanding disease, several CVD risk factors and stable treatment. This kind of patient represents most patients we manage daily in our outpatient clinics. We used a prospective design, stringent entry criteria, and a reliable method for the assessment of aortic stiffness which could be easily implemented in clinical practice. With regard to study limitations, we have to underline the relatively small sample size and the cross-sectional design of the study (patients were not randomized for treatment arms). Furthermore, we certainly cannot draw conclusions on RA patients on non-TNFi biologics as they were not included. Moreover, we could not substantiate a reduction of CVD events in RA patients with decreased arterial stiffness as the study was not powered for this outcome. Finally, smoking status was recorded as a binomial variable (ever vs never) and the number of pack-years was not calculated.

Clinical Implications And Conclusions
Long-standing RA is commonly managed in rheumatology outpatient clinics. Our results seem to indicate that TNFi treatment could be associated with reduced arterial stiffness in patients with established, longstanding RA with several CVD risk factors. Although long-term TNFi therapy can be challenging due to the high CVD burden, our data encourage the assessment of AoSI in RA patients and maintain TNFi therapy whereas AoSI is abnormally high. This can be particularly relevant in such RA patients at high CVD risk.

Declarations
Authors' contributions: AG, OV, MR provided the conception of the study; AG provided literature search, data collection and interpretation, drafting the article, revised it critically for important intellectual content; GO, AF, RB provided data collection; FO and GC performed blinded assessment of ultrasound scans; GC, RB, AD, GA provided data search, interpretation of data, drafting the article, revised it critically for important intellectual content; OV, DG, LI performed clinical assessments; all authors revised the article critically for important intellectual content and gave nal approval of the version to be submitted.

Patient and Public Involvement statement:
This research was done without direct patient involvement. Patients were not invited to comment on the study design and were not consulted to develop patient-relevant outcomes or interpret the results. Patients were not invited to contribute to the writing or editing of this document for readability or accuracy.

Consent to publication:
Not applicable (the manuscript does not contain any individual person data.
Ethical approval: All patients gave written informed consent signing a speci c institutional consent form, the study was approved by the institutional review board of the University of Verona (1707CESC) and conformed to the ethical guidelines of the Declaration of Helsinki as revised in 2000.
Data sharing: Data sharing not applicable to this article, please contact the corresponding author for data requests.
Transparency declaration: The lead author (the manuscript's guarantor) a rms that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained. Disclosures: All the authors declare that no con ict of interest exists. No author received any nancial support or other bene ts from commercial sources for this work. This research did not receive any speci c grant from funding agencies in the public, commercial, or not-for-pro t sectors.
Interaction between treatment and cardiovascular disease risk factors on aortic stiffness index (two-way ANCOVA). Data are presented as adjusted estimated means of follow-up AoSI (bullets and squares) and their 95% con dence intervals (vertical error bars). csDMARDs, conventional synthetic disease modifying antirheumatic drugs; TNFi, tumor-necrosis factor inhibitors.

Figure 2
Grouped bar charts representing means (bars) and standard errors of the mean (vertical error bars) of follow-up aortic stiffness index (AoSI) values according to treatment group and cardiovascular disease