Meta-Analysis of Asymmetric Dimethylarginine Concentrations in Rheumatic Diseases

Raised circulating concentrations of asymmetric dimethylarginine (ADMA), an endogenous inhibitor of nitric oxide synthase (NOS), have been reported in several rheumatic diseases (RDs). However, the strength of this relationship is unclear. Therefore, the aim of this systematic review and meta-analysis was to evaluate the magnitude and the robustness of the association between ADMA concentrations and RDs. We calculated standardized mean differences (SMD, with 95% confidence intervals, CI). Study heterogeneity was evaluated by meta-regressions and sensitivity analyses according to type of RDs, conventional cardiovascular risk factors, inflammatory markers, and type of ADMA assessment methodology. Thirty-seven studies with a total of 2,982 subjects (1,860 RDs patients and 1,122 healthy controls) were included in our meta-analysis. Pooled results showed that ADMA concentrations were significantly higher in patients with RDs than in healthy controls (SMD = 1.27 µmol/L, 95% CI 0.94–1.60 µmol/L; p < 0.001). However, the between-studies heterogeneity was high. Differences in ADMA concentrations between controls and RDs patients were not significantly associated with inflammatory markers, increasing age, lipid concentrations, body mass index, blood pressure, or methodology used to assess ADMA. Furthermore, subgroup analysis showed no difference across RDs. This meta-analysis showed that, in the context of significant between-study heterogeneity, circulating concentrations of ADMA are positively related to RDs.

The term "systemic rheumatic diseases" (RDs) encompasses a broad spectrum of chronic inflammatory disorders of joints and internal organs that are characterized by tissue destruction, disability and increased cardiovascular and all-cause mortality. The dysregulation of innate and adaptive immunologic response, with overproduction of pro-inflammatory cytokines, represents the pathogenic hallmark of RDs 1 .
Compared with the general population, patients with RDs suffer from a significantly reduced life expectancy that is mainly due to atherosclerotic cardiovascular disease 2,3 . This excess of cardiovascular mortality, only partially predicted by conventional risk factors, has been linked to early endothelial dysfunction and accelerated arterial stiffening [2][3][4] . In fact, peripheral and central endothelial dysfunction, a measure of nitric-oxide (NO) availability and vasodilatory function, has been shown to be more prevalent in RDs than in the general population [5][6][7] .
A strong body of clinical and experimental evidence collected in the last decade has convincingly supported a potential pathogenetic role for the accumulation of asymmetric dimethylarginine (ADMA), an endogenous inhibitor of NO-synthase 8 , in the occurrence and progression of endothelial dysfunction, arterial stiffening, and atherosclerotic cardiovascular disease 9 . Moreover, studies in healthy subjects and different patient groups have confirmed the independent prognostic role of basal plasma ADMA concentrations on cardiovascular and all-cause mortality [10][11][12] . Therefore, raised circulating ADMA concentrations might be a surrogate measure of endothelial dysfunction, early atherosclerosis and increased risk of future cardiovascular events.
Heterogeneity of SMD across studies was tested with the Q statistic (significance level at p < 0.10). The I 2 statistic was also calculated (I 2 < 25%, no heterogeneity; I 2 between 25% and 50%, moderate heterogeneity; I 2 between 50% and 75%, large heterogeneity; and I 2 > 75%, extreme heterogeneity) 39,40 . Statistical heterogeneity was defined as an I 2 statistic value ≥ 50% (38). In analyses in which heterogeneity was high, a random-effects model was used.
The influence of each single study on effect size was evaluated by sequentially excluding one study at the time, through sensitivity analysis. Publication bias was assessed by means of Begg's adjusted rank correlation test and Egger's regression asymmetry test at the p < 0.05 level of significance 41,42 . Duval and Tweedie "trim and fill" procedure was performed to identify and correct for funnel plot asymmetry resulting from publication bias 43 . The aim of this method is to remove (trim) smaller studies responsible for funnel plot asymmetry, use the trimmed funnel plot to evaluate the true centre of the funnel, then replacing the omitted studies with the new calculated "missing" studies around the centre (filling). This allows to estimate the number of missing studies and recalculates the effect size including the filled studies. Statistical analysis was carried out using MedCalc for Windows, version 15.4 64 bit (MedCalc Software, Ostend, Belgium) and Stata 14 (STATA Corp., College Station, TX, USA) This meta-analysis was performed according to PRISMA statements.

ADMA and RDs.
A total of 1,860 RDs patients (41% males) and 1,122 healthy controls (34% males) were evaluated. Overall, the mean age of participants across all studies was 44.6 years both in RDs patients and in controls (Tables 1 and 2).
The forest plot for ADMA concentrations in RDs patients and controls is shown in Fig. 2. Due to the extreme heterogeneity between studies (I 2 = 93.2%, p < 0.001), random-effects models were used to perform the analysis. Pooled results showed that ADMA concentrations were significantly higher in patients with RDs (SMD = 1.27 µmol/L, 95% CI 0.94-1.60 µmol/L; p < 0.001).
Sensitivity analysis showed that the effect size was not modified when any single study was in turn removed (effect size ranged between 1.17 µmol/L and 1.33 µmol/L, Fig. 3).
Since the Begg's (p < 0.001) and Egger's tests (p < 0.001) evidenced a significant publication bias, we applied the trim-and-fill method to correct the results. Twelve potential missing studies were added on the left side of the www.nature.com/scientificreports www.nature.com/scientificreports/ funnel plot to ensure symmetry (Fig. 4). The adjusted SMD was decreased, but remained significant (0.51 µmol/L, 95% CI 0.15-0.88 µmol/L, p = 0.006).
To explore possible sources of heterogeneity, we investigated differences in SMD across RDs (Table 2). Extreme heterogeneity was observed in all disease types (Table 2). RA, SSc and AS but not BD, FMF, and PsA patients showed significantly higher SMD when compared to healthy controls. However, no significant differences in SMD values were observed between different disease types by meta-regression analysis (p > 0.05).

Discussion
Besides their effect on the vascular tone, high circulating concentrations of ADMA have been reported to promote oxidative stress 58 vascular inflammation 59 and smooth muscle cell proliferation 60,61 . Therefore, ADMA is a candidate pathogenetic factor for oxidative stress-related endothelial dysfunction, accelerated atherosclerosis and vascular remodelling process 62 occurring in RDs.
Therefore, we performed a systematic review and meta-analysis to evaluate the strength of the association between circulating ADMA concentrations and RDs. Our analysis, including data from 37 studies with a total of 2,982 subjects, suggested that RDs are associated with significantly higher concentrations of circulating ADMA. www.nature.com/scientificreports www.nature.com/scientificreports/ However, we found substantial heterogeneity in studies estimating ADMA in RDs; Therefore, to explore in detail any potential sources of heterogeneity we presented data on a wide range of subgroups based on rheumatic disease phenotype and methodological aspects (Figs 5-8).
First, according to a pre-defined protocol, a subgroup analysis according to type of RDs was performed to identify the source of heterogeneity: other than expected, inclusion in the meta-analysis of different RDs did not account for the large heterogeneity reported in the pooled analysis; Even if SMD was higher in studies including RA, SSc and AS than in studies including BD, FMF and PsA (Table 2) this difference was not statistically significant across studies by meta-regression analysis.
We were also interested to understand whether specific pathogenetic and clinical features of RDs were associated with significant between studies heterogeneity. Therefore, according to pre-specified sub-group analyses, we www.nature.com/scientificreports www.nature.com/scientificreports/ stratified RDs in CTD vs no-CTD. However, this classification of RDs was not able to explain the large heterogeneity observed in the pooled analysis.
Moreover, according to Mc Goonagle et al. 64 , we stratified RDs in autoimmune diseases (e.g RA, SLE, and SSc), diseases with mixed autoimmune and autoinflammatory features (e.g, diseases associated with HLA-B27 and other MHC class I epitopes, such as AS, PsA and BD 65,66 ) and autoinflammatory diseases (e.g FMF). However, the difference in autoimmune vs autoinflammatory features did not account for the observed heterogeneity.
A significant association between raised circulating ADMA concentrations and traditional atherosclerotic risk factors (hypercholesterolemia, hypertriglyceridemia, and hypertension) have been reported [67][68][69] suggesting that ADMA is causally involved in the pathophysiology of atherosclerosis.
Interestingly, in our study, cardiovascular risk factors did not modify the ADMA SMD. These results question the currently accepted hypothesis that raised circulating ADMA concentrations, a potential measure of the presence of co-existing cardiovascular risk factors, may be associated with increased morbidity and mortality for cardiovascular events in RDs. Therefore, prospective studies with larger sample sizes are warranted to further define the significance of the relationship between ADMA and conventional atherosclerotic risk factor and the independent role of ADMA in the development of atherosclerotic disease in RDs.
Biomarkers of systemic inflammatory burden (such as CRP and ESR) have been shown to be correlated to circulating ADMA concentrations in the general population and in RDs 17,59,70 . However, in our meta-analyses, we found no significant association between ESR, CRP, disease duration and ADMA concentrations: the cross-sectional design and the small sample size of studies included in this meta-analysis may partly explain this finding.
We also performed two additional pre-specified meta-analyses to evaluate between-group differences according to the analytical techniques used and the biological matrix tested as a potential source of heterogeneity.
To date, preferred methods for the determination of plasma and serum ADMA concentrations are LC and ELISA. However, it is still a matter of debate whether these two methods are comparable to each other [71][72][73] . A recent meta-analysis reported plasma ADMA concentrations in healthy subjects ranging from 0.34 to 1.10 μmol/l 73 . On the contrary, normal serum ADMA concentrations in healthy subjects are less well defined: a recent study reported a range of normality of serum ADMA concentrations of 0.43-0.96 μmol/ 74 .
In our subgroup meta-analysis, we found no difference in ADMA SMD on the basis of the analytic method employed or biological matrix tested, suggesting that the large heterogeneity between studies is not explained by  www.nature.com/scientificreports www.nature.com/scientificreports/ techniques variability. However, as reported in Table 1, the range of mean plasma and serum ADMA concentrations in CTRLs was extremely large (from 0.10 to 3.22 μmol/L) with reported mean values not falling within the expected range in a large number of studies 30,33,34,37,52,57,75 . We think that this finding might be related to variability in the selection of healthy controls, pre-analytical and analytical issues, potentially accounting for the large heterogeneity of pooled estimates in this meta-analysis.
Finally, we explored the substantial heterogeneity of pooled analysis by evaluating whether specific studies might influence the final results. After excluding each study in turn, the significance of SMD remained stable in every meta-analysis.
Moreover, according to Begg's method and Egger's test, our meta-analyses carry a significant publication bias. Thus, the trim and fill method was used to adjust for funnel plot asymmetry: in particular, 12 studies were added to balance the funnel plot. The adjusted SMD, even if decreased, remained significant. This indicated that publication bias did not affect the result of the meta-analysis.
This meta-analysis has some limitations; First, substantial heterogeneity of the included studies might decrease the significance of the results. Second, we only searched for studies written in English, an aspect that may have introduced publication bias. Third, most of the articles did not report the type of anti-inflammatory or anti-rheumatic drugs used and their effects on outcome. This last aspect is of both clinical and experimental relevance, considering that some prospective studies have reported a positive effect in reducing ADMA concentrations of immunosuppressive therapies 76,77 .  www.nature.com/scientificreports www.nature.com/scientificreports/  www.nature.com/scientificreports www.nature.com/scientificreports/ Strengths of this systematic review and meta-analysis on the topic of ADMA in RDs, performed on a relatively large and representative population, include the evaluation of ADMA in different disease phenotypes within the 'rheumatic diseases' umbrella and extensive analysis of the source of heterogeneity including pathogenetic and clinical features of RDs, analytical techniques and biological matrixes.
Moreover, this meta-analysis provided an estimated 'effect size' target of 1.27 umol/L for future interventional studies aiming at reducing ADMA in the population of patients with RDs.  www.nature.com/scientificreports www.nature.com/scientificreports/

Conclusions
In summary, results from this meta-analysis show a significant positive association between elevated circulating ADMA concentrations and RDs.
The relatively high heterogeneity of the studies included in the meta-analysis, albeit extensively investigated with a number of additional analyses focused on clinical and demographic characteristics, analytical techniques, biological matrixes, and specific studies, requires further research to confirm, or refute, our findings.

Data Availability
All data generated or analysed during this study are included in this published article.