Non-invasive assessment of arterial stiffness in patients with rheumatoid arthritis: a systematic review and meta-analysis of literature studies.

Introduction. Patients with rheumatoid arthritis (RA) have an increased cardiovascular (CV) morbidity and mortality. Pulse-wave velocity (PWV) and augmentation index (AIx) are non-invasive methods to assess arterial stiffness, a marker of CV risk. We performed a meta-analysis evaluating the impact of RA on aortic-PWV, brachial-PWV, brachial–ankle (ba-) PWV, AIx, and AIx normalized to a 75 beats/minute heart rate (AIx@75). Materials and Methods. Studies evaluating the relationship between RA and aortic-PWV, brachial-PWV, ba-PWV, AIx, and AIx@75 were systematically searched. A total of 25 studies (1,472 RA patients, 1,583 controls) were included. Results. Compared to controls, RA patients showed a significantly higher aortic-PWV (mean difference 1.32 m/s; 95% CI 0.77, 1.88; P < 0.00001), ba-PWV (MD 198.42 cm/s; 95% CI 45.79, 342.76; P = 0.01), AIx (MD 11.50%; 95% CI 5.15, 17.86; P = 0.0004), and AIx@75 (MD 6.99%; 95% CI 4.92, 9.06; P < 0.00001), with a trend toward a higher brachial-PWV (MD 0.34 m/s; 95% CI –0.03, 0.70; P = 0.07). When analyzing studies on early RA, the difference in aortic-PWV among RA patients and controls was even higher (MD 2.30 m/s; 95% CI 1.15, 3.45; P < 0.0001). Conclusion. Meta-regression showed that a more severe inflammatory status impacted on aortic-PWV, AIx, and AIx@75. Arterial stiffness, a recognized marker of CV risk, is increased in RA patients. This alteration is associated with the severity of the inflammatory status and is present even in early-stage disease.


Introduction
Rheumatoid arthritis (RA) is a systemic autoimmune disorder that aff ects synovial joints and leads to chronic pain, bone erosions, and progressive disability (1). With a prevalence of 0.5% -1% in the general population, RA is the most common chronic infl ammatory condition (2,3).
Beyond joint disease, RA is characterized by a high prevalence of co-morbidities, such as gastrointestinal (4,5), respiratory (6,7), and renal diseases (5,8). Moreover, metabolic syndrome (MetS) and its major features (obesity, hypertension, impaired fasting glucose, hyperlipidemia) have been frequently found in RA patients (9). RA is also associated with shortened life expectancy (1), and cardiovascular (CV) disease accounts for 35% -50% of excess mortality in this clinical setting (10). Accordingly, the number of deaths from ischemic heart disease and cerebrovascular accidents is signifi cantly higher in RA patients than in the general population (11).
Th e increased CV morbidity and mortality in patients with RA cannot be entirely explained by traditional CV risk factors, and the underlying mechanisms leading to the increased CV risk in RA are not yet clearly understood (12,13). Th us, the association between RA and CV risk is still a matter of study.
Increased arterial stiff ness is one of the earliest stages of the atherosclerotic process (14), and pulse-wave velocity (PWV) is widely accepted as an accurate and non-invasive method to assess arterial stiff ness in humans (15). While PWV is a direct measure of arterial distensibility, the augmentation index (AIx) is a more complex parameter depending on vascular elasticity and peripheral resistance (16). PWV (17) and AIx (18) are currently considered independent predictors of major CV events and all-cause mortality. Th us, these surrogate markers of subclinical atherosclerosis provide important prognostic information over and above traditional CV risk factors (19).
During recent years, there has been a growing interest in the relationship between these markers of CV risk and RA. In particular, some case-control studies reported increased arterial stiff ness in RA patients (20,21). However, contrasting results have been reported in recent studies (22,23), and no meta-analytical data providing overall information about this issue are currently available.
Th e aim of the present study is to perform a systematic review and meta-analysis of all studies evaluating the impact of RA on PWV and AIx. Moreover, we implemented some meta-regression models to evaluate the eff ect of some clinical and demographic variables on these outcomes.

Methods
To identify all available studies, a detailed search pertaining to RA and the markers of CV risk (i.e. PWV and AIx) was conducted according to PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) guidelines (24). A systematic search was performed in the electronic databases (PubMed, Web of Science, Scopus, EMBASE), using the following search terms in all possible combinations: rheumatoid arthritis , arterial stiffness , pulse-wave analysis , pulse-wave velocity , and augmentation index . Th e last search was performed on 5 March 2015. Th e search strategy was developed without any language or study publication year restriction.
In addition, the reference lists of all retrieved articles were manually reviewed. In case of missing data, study authors were contacted by e-mail to try to retrieve original data. Two of the authors (P. A. and M.N.D.D.M.) analyzed each article and performed the data extraction independently. In case of disagreement, a third investigator was consulted (M. T.). Discrepancies were resolved by consensus. Selection results showed a high inter-reader agreement ( κ ϭ 1) and have been reported according to PRISMA fl ow chart (Supplementary online material to be found online at http://informahealthcare.com/doi/abs/ 10.3109/ 07853890.2015.1068950).

Data extraction and quality assessment
According to the pre-specifi ed protocol, all studies evaluating the impact of RA on the markers of CV risk were included. Casereports, case-series without a control group, reviews, and animal studies were excluded. To be included in the analysis, a study had to provide values (means with standard deviation or standard error) of carotid -femoral PWV (aortic-PWV), carotid -radial PWV (brachial-PWV), brachial -ankle PWV (ba-PWV), AIx, and/or AIx normalized to a 75 beats/minute heart rate (AIx@75) among RA patients and controls. Th e included studies were classifi ed as having a case-control design or a cohort design.
In each study, data regarding sample size, major clinical and demographic variables, values of aortic-PWV, brachial-PWV, ba-PWV, AIx, and AIx@75 in RA patients and controls were extracted.
Devices and methods for PWV and AIx assessment used in included studies are reported in the Supplementary online material to be found online at http://informahealthcare.com/doi/abs/ 10.3109/07853890.2015.1068950. Given the characteristics of the included studies, the evaluation of methodological quality of each study was performed with the Newcastle -Ottawa Scale (NOS), which is specifi cally developed to assess the quality of non-randomized observational studies (25). Th e scoring system encompasses three major domains (selection, comparability, exposure), and the resulting score ranges between 0 and 8, a higher score representing a better methodological quality. Results of the NOS quality assessment are reported in the Supplementary online material to be found online at http://informahealthcare.com/doi/abs/10.3109/07853890. 2015. 1068950.

Statistical analysis and risk of bias assessment
Statistical analysis was carried out using Review Manager (Version 5.2, Th e Cochrane Collaboration, Copenhagen, Denmark) provided by Th e Cochrane Collaboration.
Diff erences among cases and controls were expressed as mean diff erence (MD) with pertinent 95% confi dence intervals (95% CI).
Aortic-PWV and brachial-PWV have been expressed in meters per second (m/s), ba-PWV in centimeters per second (cm/s), while AIx and AIx@75 have been expressed as percentages (%).
Th e overall eff ect was tested using Z scores, and signifi cance was set at P Ͻ 0.05. Statistical heterogeneity between studies was assessed with chi-square Cochran ' s Q test and with I 2 statistic, which measures the inconsistency across study results and describes the proportion of total variation in study estimates that is due to heterogeneity rather than sampling error. In detail, an I 2 value of 0% indicates no heterogeneity, 25% low, 25% -50% moderate, and 50% high heterogeneity (26).
Publication bias was assessed by Egger ' s test and represented graphically by funnel plots of the standard diff erence in means versus the standard error. Visual inspection of funnel plot asymmetry was performed to assess for possible small-study eff ects, and Egger ' s test was used to assess publication bias, over and above any subjective evaluation. A P Ͻ 0.10 was considered statistically signifi cant (27). In case of a signifi cant publication bias, Duval and Tweedie ' s trim and fi ll method with the random-eff ect model was used to allow for the estimation of an adjusted eff ect size (28).
In order to be as conservative as possible, the random-eff ect method was used for all analyses to take into account the variability among included studies.

Sensitivity analyses
We repeated sensitivity analyses by including only the studies judged as ' high quality ' according to NOS (i.e. NOS Ն the median value found among included studies).
In order to avoid the risk of data overlap, a sensitivity analysis was performed aft er excluding studies involving the same recruitment centers and enrolling patients in the same time-period as other included studies.

Subgroup analyses
Given the potential infl uence of disease duration on the outcomes, we planned to perform separate analyses for studies on early RA (defi ned by a disease duration Ͻ 12 months) and late RA (defi ned by a disease duration Ն 12 months).

Meta-regression analyses
We hypothesized that diff erences among included studies may be aff ected by demographic variables (mean age, male gender) and clinical data related to disease activity [disease activity score in 28 joints (DAS28), rheumatoid factor (RF) positivity, C-reactive protein (CRP) levels, erythrocyte sedimentation rate (ESR)], antirheumatic treatment [therapy with non-steroidal anti-infl ammatory drugs (NSAIDs), sulfasalazine (SSZ), corticosteroids (CCS), methotrexate (MTX), or TNF α -blockers], and by the coexistence of traditional CV risk factors (hypertension, smoking habit, diabetes mellitus, obesity, hyperlipidemia). To assess the possible eff ect of such variables in explaining diff erent results observed across studies, we planned to perform meta-regression analyses aft er implementing a regression model with changes in aortic-PWV, brachial-PWV, ba-PWV, AIx, or AIx@75 as dependent variables ( y ) and the above mentioned co-variates as independent variables ( x ). Th is analysis was performed with Comprehensive Meta-analysis (Version 2, Biostat, 2005, Englewood, NJ, USA).

Results
Aft er excluding duplicate results, the search retrieved 342 articles. Of these studies, 194 were excluded because they were off the topic aft er scanning the title and/or the abstract, 116 because they were reviews/comments/case-reports or they lacked data of interest. Of three studies the online full-length version was not available, but, for two of them, data could be extracted from the abstract. Other six studies were excluded aft er full-length paper evaluation.

Study characteristics
All included studies had a case-control design. Major characteristics of case and control study populations are shown in Table I, and further data about disease activity and ongoing antirheumatic treatment of RA patients are reported in Table II.
Th e number of patients varied from 14 to 203, the mean age from 40.1 to 60.8 years, and the prevalence of male gender from 0% to 62. 5%.
In one study (21), which provided separate data for patients with early and late RA, the two groups were evaluated as two different data-sets.
Defi nition of early RA was highly variable among included studies, being defi ned as newly diagnosed RA (36,38) or as disease during less than 1 (21) or 6 years (49).
Th e NOS for quality assessment of included studies showed a median value of 6.

Publication bias
Because it is recognized that publication bias can aff ect the results of meta-analyses, we attempted to assess this potential bias using funnel plots analysis. Visual inspection of funnel plots of eff ect size versus standard error for studies evaluating aortic-PWV, brachial-PWV, and AIx@75 suggested an asymmetric distribution of studies around the mean (Supplementary online material to be found online at http://informahealthcare.com/doi/abs/10.3109/ 07853890. 2015.1068950), and Egger ' s test confi rmed a signifi cant publication bias ( P Ͻ 0.001, P ϭ 0.01, and P ϭ 0.02, respectively). Interestingly, the adjusted eff ect size, estimated by Duval and Tweedie ' s trim and fi ll method, substantially confi rmed results for aortic-PWV (MD 1.32 m/s; 95% CI 0.77, 1.88), brachial-PWV (MD 0.34 m/s; 95% CI -0.03, 0.70), and AIx@75 (MD 5.42%; 95% CI 7.28, 37.97).
In contrast, the distribution of studies evaluating ba-PWV and AIx was rather symmetrical, and no publication bias was found by Egger ' s test ( P ϭ 0.99 and P ϭ 0.61, respectively).
Interestingly, in both sensitivity analyses, the diff erence between RA patients and controls became signifi cant for brachial PWV.
Interestingly, our results on aortic-PWV were confi rmed in both early and late RA. However, the relatively low number of Table I  studies on early RA ( n ϭ 4) made this subgroup analysis unlikely to be performed for the other outcomes. Th us, results on brachial-PWV, ba-PWV, AIx, and AIx@75 could be confi rmed only in late RA patients (Table IV).

Discussion
Results of the present meta-analysis consistently show that RA is associated with increased arterial stiff ness. In particular, we reported an increased aortic-PWV, ba-PWV, AIx, and AIx@75 in RA patients. A trend towards a higher brachial-PWV was also documented in RA. Our fi ndings are strengthened by the sensitivity analyses and, particularly, by the subgroup analysis which confi rmed results on aortic-PWV both in early RA and in late RA patients. Moreover, regression models were able to refi ne the results further, providing the evidence that age, male gender, and a more severe infl ammatory status may signifi cantly impact on the evaluated outcomes.
Arterial stiff ening is one of the earliest stages of the atherosclerotic process (50). PWV and AIx are widely accepted as non-invasive methods to assess clinically the peripheral and central arterial stiff ness (15,16) and, in turn, as surrogate markers of subclinical atherosclerosis (19). Th e clinical relevance of increased arterial stiff ness lies in its ability to predict CV morbidity and mortality over and above other traditional CV risk factors in patients with hypertension (51), end-stage renal disease (52), elderly individuals (53), and in the general population (54,55). Th is prognostic importance has also been recently confi rmed in a meta-analysis (17).
Th us, clinical assessment of arterial stiff ness consistently suggests an elevated CV risk in RA patients, which is widely confi rmed by some epidemiological studies showing an increased incidence of major CV events in this clinical setting (10,11). Th is evidence is even more signifi cant if we consider the relatively young age of the Anti-TNF α ϭ ongoing therapy with TNF α -blockers; CCS ϭ ongoing therapy with corticosteroids; CRP ϭ C-reactive protein; DAS28 ϭ Disease Activity Score in 28 Joints (clinical remission DAS28 Ͻ 2.6); ESR ϭ erythrocyte sedimentation rate; MTX ϭ ongoing treatment with methotrexate; NA ϭ not assessed; NSAIDs ϭ ongoing therapy with non-steroidal antiinfl ammatory drugs; Pts ϭ patients with rheumatoid arthritis; RA ϭ rheumatoid arthritis; RF ϭ rheumatoid factor; SSZ ϭ ongoing therapy with sulfasalazine.  population enrolled in included studies, with a mean age ranging between 40.1 and 60.8 years. Moreover, despite the presence of several CV risk factors, the meta-regression analysis confi rmed that all results of our meta-analysis were largely independent of these co-variates. Many CV risk factors are thought to have a causal role in the atherosclerotic process (56). Although RA patients exhibit an in-creased prevalence of these CV risk factors (9), the relationship between atherosclerosis and RA seems to be more complex, and the presence of traditional risk factors might not entirely explain the accelerated atherosclerotic process in this clinical setting. Th us, other mechanisms (i.e. infl ammatory and immunological) have been proposed to explain the relationship between RA and atherosclerosis (9). Th e improvement of the CV risk profi le fol-tion/disease activity should be considered in the CV risk profi le assessment of RA patients.
Interestingly, our subgroup analyses excluded the potential infl uence of disease duration on aortic-PWV and confi rmed results on this outcome both in early RA and in late RA patients. However, considering the relatively low number of studies on early RA ( n ϭ 4), results on brachial-PWV, ba-PWV, AIx, and AIx@75 could be confi rmed only in late RA patients. Of interest, the diff erence in aortic-PWV between cases and controls was even greater when specifi cally considering the early RA subset (MD 2.30 m/s) compared with patients aff ected by late lowing the control of systemic infl ammation by anti-infl ammatory treatments (57) argues for the possibility that systemic infl ammation acts as an independent CV risk factor in RA (58).
RA (MD 1.32 m/s). Th e latter result is in line with some data from our recent meta-analysis (60), showing that the diff erence in CCA-IMT among patients and controls is higher in early RA.
As previously supposed (60), this may be due to the younger age of patients and controls recruited in the studies on early RA. In such a young clinical setting, the pro-atherogenic eff ect of RA is likely to be more signifi cant, and this supports the hypothesis that CV risk already starts to increase as soon as the fi rst signs of autoimmunity and infl ammation appear, which is oft en some months/years prior to diagnosis (61,62). Overall, our fi ndings further support the hypothesis that premature atherosclerosis may be one of the main features of RA and autoimmune disorders (63) and that chronic infl ammation plays an important role in its pathogenesis, acting independently and/or synergistically with traditional CV risk factors. In order to provide a comprehensive overview of the relationship between the attempt to overcome this potential limitation, we repeated the analyses by using standardized mean diff erence (SMD) instead of MD, this method being designed to be used when diff erent methods of measurement are analyzed together (26). Interestingly, all results were confi rmed using SMD (data not shown).
Finally, results on AIx should be interpreted with caution, as AIx is a more complex parameter depending not only on arterial stiff ness but also on peripheral resistance (16).
In conclusion, in our meta-analysis RA appeared signifi cantly associated with increased arterial stiff ness and, in turn, with subclinical atherosclerosis and CV risk. Th us, patients with RA may benefi t from a more meticulous screening for CV risk factors and more specifi c CV prevention strategies. However, additional and specifi cally designed studies are needed in order to establish the optimal management of these patients.
RA and arterial stiff ness, we evaluated all available methods for its clinical assessment. Th us, we also included studies on ba-PWV, which is a promising technique to measure arterial stiff ness conveniently and more suited to routine clinical use than the other more ' complicated ' approaches (64). However, we should consider that aortic-PWV still remains the most widely used technique and the ' gold standard ' measure of arterial stiff ness with a clear predictive value (18). Th e clinical relevance of our results can be better understood when we consider that the risk of major CV events increases by about 14% with each 1 m/s increase in aortic-PWV (17). Similarly, this risk increases by 31.8% with each 10% increase in AIx (18).
Further extending the hypothesis of an increased CV risk in RA patients and suggesting the need for a strict monitoring of subclinical signs of atherosclerosis in RA patients, we have recently documented an increased carotid intima-media thickness (IMT) accompanied by higher prevalence of carotid plaques in patients with RA (60). Interestingly, we also demonstrated that the increase in IMT was dependent on the severity of the infl ammatory process in patients with RA.
Overall, our results further support the need for large long-term interventional trials with CV end-points to investigate whether benefi ts in articular disease achieved by aggressive suppression of infl ammation may translate into reduced CV risk in RA.
Some potential limitations of our study need to be discussed. First, studies included in our meta-analysis have diff erent inclusion and exclusion criteria, and most of patients included in the analysis had concomitant CV risk factors (hypertension, smoking, obesity, diabetes mellitus, hyperlipidemia) and diff erent disease activity status. Since meta-analysis is performed on aggregate data and some information is missing in each study, the multivariate approach allowed for the adjustment for some (but not all) potential confounders. Th us, although results of meta-regression analyses were able to refi ne analyses by assessing the infl uence of most clinical and demographic variables on the observed results, caution is necessary in overall results interpretation.
Second, heterogeneity among the studies was generally signifi cant. Although it was not possible to ascertain conclusively the sources of heterogeneity, all results were confi rmed aft er adjustment for potential publication bias.
A further limitation is the relative lack of specifi c data on early RA and the high variability in the defi nition of early RA among studies. In the frame of a sensitivity analysis, to adjust results for the potential infl uence of disease duration on the outcomes, we were able to confi rm our results on aortic-PWV in early RA, but we could not confi rm results on brachial-PWV, ba-PWV, AIx, and AIx@75 in this clinical setting. In line with some recent data suggesting the presence of an increased subclinical atherosclerosis in patients with early RA (60), further studies specifi cally focusing on PWV and AIx in early RA patients are needed to address the hypothesis of a signifi cant increase in the CV risk starting from very early stages of the disease in patients with RA (61).
Another potential limitation of this study is that PWV and AIx measurement may be infl uenced by many confounding factors, signifi cantly limiting reproducibility of arterial stiff ness assessment and, in turn, the relevance of our results. In particular, some studies (65,66) indicate that heart rate may aff ect results, and, with the only exception of AIx@75, all the other outcomes have not been standardized to a specifi c heart rate.
Moreover, we have to consider that diff erences among assessment techniques and devices, as well as the lack of comparable age-adjusted normal values may limit the validity of arterial stiffness parameters as markers of early atherosclerosis (67). Th us, caution is necessary in overall results interpretation. However, in