Article Text
Abstract
Background Cardiovascular mortality is increased in patients with rheumatoid arthritis (RA). RA is associated with an increased left ventricular mass index (LVMI), a strong marker of cardiovascular mortality, and vessel abnormalities. Experimental studies have suggested that tumour necrosis factor α (TNFα) may induce LV hypertrophy.
Objective To study the effect of medium-term (3- and 6-months) treatment with the TNFα inhibitor etanercept (ETN) and synthetic disease-modifying antirheumatic drugs (sDMARDs) on LV morphological features and arterial stiffness in patients with RA.
Methods Consecutive female patients with active RA requiring treatment with ETN (n=28) or sDMARDs (n=20) were included. Clinical and biological monitoring, echocardiography and pulse wave velocity (PWV) assessment were performed at inclusion and at 3 and 6 months after the start of treatment. Paired t tests and multivariate linear regression analysis were used.
Results Mean LVMI tended to be higher at baseline in the ETN group than in the sDMARD group (96.5±19.8 vs 84.3±26.8 g/m2; p=0.11 for the ETN and sDMARD groups, respectively). In patients with ETN treatment, mean LVMI was significantly decreased at 3 and 6 months (−6.3±7.6 and −14.2±9.3 g/m2; p<0.001), with no change from baseline for patients with sDMARD treatment (−2.2±10.9 and −2.7±10.2 g/m2, respectively). Blood pressure (BP) and aortic PWV were not changed by either treatment.
Conclusions ETN induced a significant decrease in LVMI with medium-term treatment with no change in BP or PWV. TNFα may be an important factor of LV hypertrophy, which may explain the benefit of TNF inhibitors on cardiovascular morbidity and mortality in RA. These results need to be confirmed by larger studies and with other TNF inhibitors.
Statistics from Altmetric.com
Introduction
Rheumatoid arthritis (RA) is the most common systemic autoimmune disease and affects 0.3–1% of the developed world.1 Patients with RA have about a twofold increased risk of heart failure and mortality.2 ,3 RA is associated with increased left ventricular mass (LVM) and diastolic dysfunction but preserved LV function.4–10 LV hypertrophy (LVH) predicts cardiovascular events in the general population independently of traditional risk factors11 and may also contribute to the early cardiovascular morbidity and mortality seen in patients with RA.
Experimental studies suggest that tumour necrosis factor α (TNFα) may induce LV remodelling.12 Arterial stiffness is an independent predictor of cardiovascular risk in various patient groups,13 ,14 and increased arterial stiffness, accessed by pulse wave velocity (PWV), has been reported in RA.15 TNFα inhibitors may decrease cardiovascular morbidity and mortality in RA,16 by reducing LVM or arterial stiffness.
We aimed to assess the medium-term effect (3 and 6 months) of the TNFα inhibitor etanercept (ETN) on LV morphological features and arterial stiffness in female patients with RA. We also looked at the effect of synthetic disease-modifying antirheumatic drugs (sDMARDs) in patients with early RA. Then, we assessed the association of potential cardiovascular risk factors on ETN-induced variation in LVM over time.
Patients and methods
Study population
We enrolled consecutive patients with RA requiring treatment with ETN (50 mg subcutaneously, once a week) or a sDMARD (methotrexate, sulfasalazine, leflunomide) from September 2008 to September 2011. All patients met the 2010 American College of Rheumatology/European League Against Rheumatism criteria for RA.17 Since definition and mechanisms behind development of LVH are different between men and women,18 ,19 and since RA is much more common in women, we decided to include only women. Exclusion criteria were diabetes, obesity, previous cardiovascular events, cardiopathy (valvulopathy, impaired systolic function with LV ejection fraction <55%), uncontrolled hypertension (systolic blood pressure (BP) >140 mm Hg and/or diastolic BP >90 mm Hg), alcohol abuse (>50 g/day) and renal disease. Use of corticosteroids, non-steroidal anti-inflammatory drugs (NSAIDs) and antihypertensive drugs was permitted if doses were stable for at least 4 weeks, as were sDMARDs for the ETN group if doses were stable for at least 3 months. For the sDMARD group, we excluded patients who had received these drugs before the study. For the ETN group, we excluded patients who had received any biological treatment.
Clinical evaluation
All patients were clinically evaluated at baseline and at 3 and 6 months by the same observer (CID or JM), including disease activity score in 28 joints (DAS28) and body mass index (BMI). The study was approved by the local ethics committee (Montpellier, France), and informed consent was obtained from each patient in accordance with the Helsinki Declaration of 1975 (revised in 1983).
Biological variables
Fasting blood samples were collected, immediately centrifuged and stored at −80°C until analysis. C-reactive protein (CRP) level was determined by an immunoturbidimetric method with the Olympus AU2700 biochemistry analyser (Beckman-Coulter, Villepinte, France) and Randox full-range CRP reagent (Randox Laboratories, Mauguio, France). Low-density lipoprotein cholesterol level was determined by the Friedwald's formula for a triglyceride level <4.5 mmol/l. Other lipid levels were determined by an enzymatic method with Architect c8000 (Abbott, Rungis, France). ETN concentrations were measured by an ELISA Lisa Tracker (BioDiagnostic, Marne-la-Vallée, France). TNFα was determined using a proteomic approach on an Evidence Investigator biochip system (Randox, Mauguio, France).
Echocardiography
The same observer (GdC) performed all echocardiography with a Sequoia (Siemens, St Denis, France) ultrasound system.20 Two-dimensional images were acquired at the highest frame rate possible (>100 frames/s). Tracings were analysed using an offline station by a reader who had no knowledge of patient group and sequence treatment. Details of all echocardiography measurements obtained are given in supplementary file (available online only).
LVM was calculated by the Penn cube method as described.21 As previously published, the coefficient of variation and intraobserver variability for LVM measurement were respectively 9% and 8%.22 The LVM index (LVMI) was defined as LVM/body surface. LVH was defined as LVMI>110 g/m2 and concentric remodelling as relative wall thickness >0.43.23 LV end-diastolic volume and ejection fraction were determined by two-dimensional echocardiography according to the recommendations of the American Society of Echocardiography.24 Transmitral inflow was analysed for measurement of peak velocity of early (E) and late atrial filling (A), E/A ratio. Mitral annulus velocities obtained by pulsed tissue Doppler imaging (DTI) were recorded from the basal lateral segment, and DTI peak myocardial systolic (Sa) and early diastolic (Ea) velocities were measured. Unfortunately, DTI was not of high quality for all patients and measurements were obtained only in 18 in the ETN group and 14 patients in the sDMARD group.
Determination of arterial function
The same observer (PF) measured all haemodynamic variables with the patients in a quiet room maintained at constant temperature (21°C). BP was measured every 3 min with an automatic device (model 8800, Colin Corp, Komaki, Japan)25; reported values were the mean of at least 10 measurements after a 10 min period of rest with the patient in the supine position. Femoral, radial and carotid pressure waveforms were obtained using applanation tonometry (SPC-301; Millar Instruments, Houston, Texas, USA). The central aortic waveform and the augmentation index (AIx) were derived by applying a validated transfer system (SphygmoCor System; AtCor Medical, Sydney, Australia) to recordings of the arterial pressure waves at the radial artery. Carotid-to-femoral PWV and AIx were determined as previously described.26 ,27 Details of the methods are detailed given in an online supplementary file.
Statistical analysis
Patient characteristics are described with percentages for categorical variables and mean±SD or median (range) for continuous variables. Because of skewed distribution, the CRP level was log-transformed before further analysis. At baseline, categorical variables were compared by χ2 test or Fisher's exact test and continuous variables by Student t test or Mann–Whitney test. The relationships between LVMI at baseline and other quantitative baseline parameters were assessed using the Pearson correlation coefficient, and adjusted using the partial Pearson correlation coefficient. At months 3 and 6, variables were assessed by change from baseline by paired Student t test or signed-rank test. The evolution of LVMI during the study was analysed using a linear mixed model. To determine variables predicting the evolution of LVMI over time, we used several mixed models, adjusting for the evolution of several potentially explanatory factors. These linear mixed models included a subject-specific random intercept, LVMI was the dependent variable, fixed effects were the time after inclusion (estimate shown in table 4) and, successively, the potentially explanatory factors. The effect of these parameters as intermediates factors was estimated by quantifying the proportion of decrease after adjustment in the interaction coefficient between time and LVMI (proportion of the evolution of LVMI with time explained by the evolution of the factor).
Analyses were carried out with SAS V.9.1 (SAS Inst, Cary, North Carolina, USA). Two-tailed p<0.05 was considered statistically significant.
Results
Baseline characteristics of patients
We included 20 patients in the sDMARD group and 28 in the ETN group. Baseline characteristics are in table 1. At baseline, the two groups were comparable in age, DAS28, BMI, mean BP, controlled hypertension and frequency of smoking. However, as compared with the ETN group, the sDMARD group had shorter disease duration, less frequent positivity for rheumatoid factor (RF) and anti-citrullinated protein antibodies, fewer joint erosions and lower steroid use. The groups had similar biological variables, including CRP level, glucose and lipid profiles.
LVMI was not significantly different between the two groups (96.5±19.8 vs 84.3±26.8 g/m2; p=0.11 for the ETN and sDMARD groups, respectively). LVH frequency was greater in the ETN group (n=7 vs n=3). Concentric remodelling was more frequent with ETN than DMARDs (56% vs 21%, p=0.04); PWV was higher (9.1±2.2 vs 7.7±2.0 m/s; p=0.03). AIx was not significantly different between the two groups (9.7±6.4 vs 6.4±7.4; p=0.10). Ejection fraction was similar in the two groups (66±8 vs 67±8% respectively in the ETN and sDMARD group, NS). Diastolic function was also comparable in both groups (Ea: 0.090±0.02 vs 0.078±0.02 m/s, NS, respectively in the sDMARD and ETN groups).
LVMI at baseline
At baseline, LVMI for all patients was positively correlated with age (r = 0.68, p<0.001), pulse pressure (r = 0.56, p<0.001), systolic BP (r = 0.48, p=0.003), BMI (r = 0.35, p=0.03), triglycerides level (r = 0.53; p=0.001) and PWV (r = 0.52; p<0.001) (table 2). No correlation was found with TNFα levels (r=−0.05; p=0.79). After adjustment for age, smoking status and pulse pressure, PWV and BMI were no longer correlated with baseline LVMI. Triglycerides remained significantly correlated with LVMI (r=0.57, p=0.001). RF positivity did not influence LVMI at baseline (87±24 vs 96±22 g/m2, p=0.21, respectively for RF− and RF+).
Change in clinical and biological variables with 6-months’ treatment
One of the two patients with controlled hypertension in the sDMARD group changed his antihypertensive drug during follow-up from a β blocker to another (from propanolol to bisoprolol) (table 3). None of the hypertensive drugs of the five patients with hypertension in the ETN group were modified during follow-up. DAS28 was significantly decreased at 3 months with both sDMARDs and ETN (−0.8±0.9 and −1.2±1.5, both p<0.01, respectively) and at 6 months with ETN (−1.5±1.3, p<0.001). Steroid dose was significantly decreased at 3 months with ETN (−1.6±2.9 mg/day, p<0.01) but not at 6 months. CRP level was significantly decreased at 3 and 6 months with ETN (1.2±1.1 and −1.0±1.1 log[mg/l], p<0.05, respectively). CRP level was decreased and low-density lipoprotein cholesterol level increased at 3 months with DMARDs (−0.8±1.3 log[mg/l]; 0.1±0.2; p<0.05). TNFα significantly increased after ETN at 3 and 6 months (3.7 (0.7–10.2), p<0.01 and 5.8 (0.2–8.7) pg/ml, p<0.01; respectively), whereas no change was seen in the sDMARD group (−0.6 (−2.98 to 0.07), NS and 0.27 (−0.12 to 1.72), NS, respectively).
Change in cardiac and vessel variables with 6-months’ treatment
LVMI was significantly decreased at 3 and 6 months with ETN (−6.3±7.6, and −14.2±9.3 g/m2, p<0.001; respectively) but not in the sDMARD group (−2.2±10.9 and −2.7±10.2 g/m2, respectively) (figure 1). When modelling the evolution of LVMI during the study using a linear mixed model, LVMI was decreased by 2.16 g/m2/month with ETN (p<0.01) and 0.36 g/m2/month with sDMARDs (p=0.44). The frequency of LVH was significantly decreased with ETN (0 patients at 6 months vs 7 at baseline, p<0.01), with no significant change with sDMARD (2 patients after treatment vs 3 at baseline, p=0.17). Ejection fraction was not modified (66±8, 67±7, 67±8%, respectively, at M0, M3 and M6, NS in the ETN group and 67±8, 66±7, 67±6%, respectively, at M0, M3 and M6, NS in the sDMARD group). Improvement of early diastolic velocity (Ea) was found at 6 months in patients treated with ETN (p<0.05) (see supplementary file). No significant change was seen in aortic arterial stiffness. Of note, radial waveforms were also performed and no significant change of radial PWV or AIx was seen in either group (data not shown).
Predictors of ETN-decreased LVMI
Interactions between age at baseline, systolic BP at baseline or RF status at inclusion and evolution of LVMI during the 6-months’ treatment with ETN were not significant (β age×time=−0.027, SEM=0.028, p=0.32; β systolic BP×time=−0.007, SEM=0.020, p=0.71 and β RF status×time=0.63, SEM=0.81, p=0.44, respectively) (table 4).
On univariate analysis, variation in LVMI with ETN was positively correlated with change in DAS28 in the first 6 months (r = 0.63; p<0.001; figure 2). On multivariate linear mixed model, adjustment for change in TNFα, DAS28, ETN concentration at 6 months or change in CRP level reduced the coefficient of evolution of LVMI with time by 37%, 25%, 21% and 10%, respectively. Adjustment for change in CRP levels and ETN concentration at 6 months decreased the estimate by 32% (−1.46 g/m2/month). Adjustment for baseline RF positivity decreased the estimate by 22%.
Discussion
We studied the medium-term effect (3 and 6 months) of treatment with the TNFα inhibitor ETN in established RA, or sDMARD in early RA, on LV morphological features and arterial stiffness in female patients with RA, and the predictors of ETN-induced variation in LVM. A 6-month course of ETN significantly decreased LVMI and corrected LV morphological features. This decrease was related in part to ameliorated DAS28 and ETN concentration. No effect on LVMI was seen after a 6-month course of sDMARDs.
In RA, increased LVM and impaired diastolic function have been reported.4–10 In our study, in comparison with reference values of a control population of patients,28 we found an increase of LVMI in the ETN group and arguments in favour of diastolic dysfunction in both groups (decrease of Ea). However, the way in which RA induces changes in LV structure is not well established. We found the main determinants of LVMI at baseline to be age, BP, BMI and triglycerides, as described for the general population. BP and BMI are slightly increased in RA and might therefore explain, in part, the increase in LVMI in these patients.29–31 Previous studies found no modification of triglycerides in RA.32 ,33 RA duration was not correlated with LVMI, as was previously found.7
Proinflammatory cytokines, especially TNFα, are known to play an important part in the pathogenesis of RA and might explain the augmentation of LVMI in this disease. TNFα is increased in various cardiovascular diseases, including myocarditis, myocardial infarction and congestive heart failure. In patients with diabetes, LVM is associated with circulating TNFα level.34 TNFα promotes LV remodelling and myocyte hypertrophy in mouse models.12 ,35–39 TNFα-knockout mice which underwent aortic banding showed attenuated cardiac apoptosis, hypertrophy, inflammatory response, reparative fibrosis and a reduced level of cardiac matrix metalloproteinase-9 activity as compared with wild-type mice.35 The effect of TNFα might be explained, in part, by its role in mediating the effects of angiotensin II, which promotes cardiac hypertrophy.40 Proinflammatory cytokines could also be associated with subclinical myocarditis, which was described and recently confirmed with MRI in RA.41 ,42 In our study, TNFα levels at baseline did not correlate with LVMI. However, measurements of TNFα in serum samples are not reliable.
Arterial stiffness was correlated with LVMI at baseline but was not an independent determinant of LVMI after adjusting PWV for age and mean BP. In this study, inflammation-induced arterial stiffness does not seem to explain LVH as it has been shown with lupus patients.43 However, this study has not enough power to rule out this hypothesis and larger studies are required. Moreover, no significant change of PWV was seen in the ETN group. Several studies have previously reported a decrease of aortic PWV after treatment with TNF inhibitors.15 ,44–46 Therefore, the absence of a significant decrease might be because the study had insufficient power.
The control of disease activity with ETN may explain in part the decrease in LVMI. Change in LVMI with ETN was correlated with change in DAS28 (r=0.63; p<0.001). On multivariate analysis, adjusting for DAS28 decreased the magnitude of the LVMI variation (25% decrease). However, DAS28 could explain the observed effect only in part because the LVMI decrease remained significant after DAS28 adjustment. We found no change in BP, BMI or triglyceride level in our study, so these parameters may not explain LVMI changes in RA. In patients with hypertension, correction of BP with antihypertensive drugs led to a smaller and slower decrease of LVMI as compared with our results.47–49
With the known relation between TNFα and cardiac remodelling, LVMI decrease could be directly due to inhibition of TNFα. In a randomised controlled trial of patients who underwent cardiac allografting, treatment with TNFα inhibitors decreased the occurrence of LVH as compared with placebo.50 In our study, TNFα levels in patients treated with ETN were significantly increased, which strongly suggests that they do not reflect active TNFα. Thus it is difficult to interpret the decrease in LVMI estimate induced by adjustment of TNFα levels. However, the decrease in LVMI was influenced by ETN concentrations. Lack of a control group does not allow us to reach a conclusion about a specific effect of anti-TNF drugs.
LVMI was significantly reduced after 3 and 6 months of ETN treatment, for a 15% decrease at 6 months from baseline. The seven patients with LVH at baseline showed normalised LVM at 3 or 6 months (p=0.01). Improvement in diastolic function parameters was also found in the ETN group at 6 months. This result suggests that normalisation of LV morphology with ETN is associated with a reversibility of diastolic dysfunction. In a recent study of 23 patients with RA, other authors also found a significant decrease of LVM after a 1-year course of infliximab.51 Another study found no significant change in diastolic dysfunction but the methods used were different.52 In the general population, women with LVH have a 7.4% risk of cardiovascular events as compared with a 4.2% risk with normal LVM.11 In patients with hypertension, improving LVH decreases by nearly 40% the risk of cardiovascular events (HR=0.62; 95% CI 0.47 to 0.82; p=0.001).53 Therefore, the normalisation observed in our study might be related to the previously reported decrease in cardiovascular events in patients receiving TNFα inhibitors.16 ,54 These results are surprising because TNFα inhibitors are contraindicated in patients with severe heart failure. Four randomised controlled trials studied the effect of TNFα antagonists (infliximab and ETN) in patients with moderate to severe heart failure.55 Mortality was slightly, but not significantly, higher than in the placebo group only for high doses of infliximab.55 These doses are not used in RA and these populations are different from our patients with RA, who did not show impaired systolic function. In patients with hypertension or heart failure, LVH initially adapts the cardiac fibre to the mechanical overload, so LV reduction could have adverse effects. For our patients, with only diastolic dysfunction and abnormal LVH, the benefit would be more sensitive. The RABBIT study did not find any increased risk of worsening of prevalent heart failure in patients receiving TNFα antagonists.56
One limitation of this study is the lack of a control group. The sDMARD group cannot be used as a control group since its baseline characteristics differed from those of the ETN group. Mean baseline LVMI was 96.5±19.8 g/m2, with 7% and 75% of patients having LVH and/or concentric remodelling, respectively, at inclusion in the ETN group; as compared with 84.3±26.8 g/m2 with 3% and 30% patients having LVH and/or a concentric remodelling, respectively in the sDMARD group. Even if the difference were not statistically significant between the two groups, LVMI with ETN but not sDMARDs was increased as compared with healthy subjects,28 which may explain why we found no significant change in the sDMARD group. Moreover, in the sDMARD group, patients less often had autoantibodies, which are associated with cardiovascular events.57 Randomised control studies with patients starting either monoclonal antibodies or a soluble receptor directed against TNFα or another biological agent, are needed to determine if the decrease in LVMI was due to a specific effect of anti-TNF drugs and especially to ETN.
Another limitation of this study is the use of echocardiography. Cardiac MRI is more accurate and more reproducible for evaluation of cardiac size and structure than echocardiography. However echocardiography is more accessible and is not contraindicated when metallic implants are present. Other controlled studies based on cardiac MRI would be interesting to confirm our results. Then, missing data for diastolic function assessment requires caution and further longitudinal studies using imaging such as ‘echo tracking’ would be useful for a better assessment of diastolic function. Finally, NSAIDs were not recorded in this study. It has been shown that NSAIDs can induce changes in LV dimensions but these modifications seem to be transient and should not influence the results of our study at 3 and 6 months. Antihypertensive treatment might also affect LVMI. However, none of the five patients with hypertension in the ETN group had any change of their antihypertensive drugs during the 6 months’ follow-up.
In this longitudinal prospective study of patients with RA, ETN induced a significant decrease in LVM. BP and arterial stiffness were not modified by treatment and therefore could not explain the effect on LVM. TNFα may be an important factor in LVH observed in RA and may explain in part the previously reported benefit of TNF inhibition on cardiovascular morbidity and mortality in RA. These results need to be confirmed in larger controlled studies with an MRI evaluation and also with other TNF inhibitors.
Acknowledgments
This work was supported by a grant from the French Society of Rheumatology and unrestricted grants from Pfizer (France). We acknowledge Martine Delage, Ermis Parussini, Sami Slimani, Cédric Lukas and Gaël Mouterde for helpful participation in this study.
References
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.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Footnotes
-
Contributors CID, PF, GdC, VD and JM: performed clinical or cardiovascular evaluations; A-MD, J-PC: performed biological evaluations; TM: performed statistics; BC, JR, CID, PF, GdC, J-PC: designed the project; CID, PF, GdC: data collection; CID: wrote the article; PF, JM, BC, GdC: corrected the article.
-
Funding BC and JM received less than €10 000 in honoraria from Pfizer laboratory for consulting as investigator of clinical trials.
-
Competing interests None.
-
Ethics approval The local ethics committee (Montpellier, France).
-
Provenance and peer review Not commissioned; externally peer reviewed.