Possible misclassification of cardiovascular risk by SCORE in antisynthetase syndrome: results of the pilot multicenter study RI.CAR.D.A

Abstract

Objectives

To test the ability of an established traditional cardiovascular (CV) risk prediction score [Systematic COronary Risk Evaluation (SCORE)] and its EULAR modified version (mSCORE) to identify antisynthetase syndrome (ASyS) patients at high CV risk and to examine for the first time associations of CV and cerebrovascular surrogate markers with clinical and immunological ASyS parameters.

Methods

SCORE/mSCORE and the gold standard marker of aortic stiffness [carotid-femoral pulse wave velocity (cfPWV)] were examined in ASyS patients and healthy controls. Moreover, sonography of the common- (CCA) and internal- (ICA) carotid arteries was performed in subsets of both groups, evaluating carotid intima-media thickness (cIMT), plaques and Doppler sonographic cerebrovascular surrogates [resistance (RI) and pulsatility (PI) indices].

Results

We recruited 66 ASyS patients and 88 controls. According to mSCORE, 10% of the patients had high CV risk. However, cfPWV and carotid sonography revealed an increased CV risk in 21.2% and subclinical carotid atherosclerosis (SCA) in 85.7% of the patients, respectively. cfPWV and cIMT were higher in patients compared with controls (P_adj=0.021 and P_adj=0.003, respectively). In the ASyS group, cfPWV and cIMT correlated significantly with age (r = 0.679; P<0.001 and r = 0.664; P<0.001, respectively). Moreover, cfPWV correlated with BMI (P_adj=0.001) and diabetes (P_adj=0.043). CCA-RI and CCA-PI showed significant associations with creatine phosphokinase (r = 0.629; P=0.012 and r = 0.574; P=0.032, respectively) and ICA-RI and ICA-PI were higher in patients with lung involvement (both; P=0.039).

Conclusion

ASyS patients had higher aortic stiffness and SCA compared with controls, even after adjustment for confounders. SCORE/mSCORE performed poorly in identifying high-risk patients compared with cfPWV and carotid sonography. Thus, cfPWV and carotid sonography may improve CV and cerebrovascular screening in ASyS.

Introduction

Antisynthetase syndrome (ASyS) is a heterogeneous autoimmune overlap disease characterized by anti-aminoacyl-tRNA synthetase antibodies (anti-ARS) and the classic triad of arthritis, myositis and interstitial lung disease (ILD) [1, 2]. Other relevant, but less prevalent accompanying features are RP, fever and fissured hyperkeratosis of the fingers (so-called mechanic’s hands) [3]. Moreover, up to 35.3% of ASyS patients can develop a nailfold capillaroscopic pattern similar to systemic sclerosis [4], whereas the prevalence of pulmonary hypertension is relatively low (7.9% in a retrospective study) [5]. The most frequent anti-ARS is Jo-1, whereas other anti-ARS (i.e. anti-PL-7, -PL-12, -EJ, -OJ, -KS, -YRS, -Zo) are less commonly detected [1, 6].

ASyS diagnosis is often complicated through a lack of validated classification criteria, the sequential appearance of clinical features and the rare examination of non-Jo-1 anti-ARS [7, 8]. In addition, ASyS-related comorbidities have not been sufficiently examined and data regarding cardiovascular (CV) risk are lacking. In particular, no studies regarding CV disease have been performed in the setting of ASyS and the only information available on this topic arises from a few studies evaluating patients with myositis-associated disorders in general. In those explorations (i.e. Dobloug et al.), evidence of increased CV risk and mortality has been found [9, 10]. Interestingly, patients with DM and PM have a 2- to 4-fold increased risk of developing coronary artery disease [10]. Moreover, CV disease has been described as one of the major causes of death [9, 11] and CV mortality is reported to be as high as 36–55% in idiopathic inflammatory myopathies (IIM) [12].

In the general population, several CV prediction scores such as the Systematic COronary Risk Evaluation (SCORE) [13], Framingham [14] or PROCAM [15] have been proposed to classify CV risk. However, these scores do not assess inflammation-mediated effects on the CV system and can therefore misclassify CV risk in rheumatic diseases. For that reason, EULAR guidelines for CV risk management recommend the adaptation of CV risk prediction models by a 1.5 multiplication factor in the case of RA and other inflammatory arthritides [16]. In the same guidelines, EULAR recommends the use of the modified SCORE version (mSCORE) for the prediction of CV risk in patients with inflammatory arthropathies. However, there is no data concerning the diagnostic value of SCORE/mSCORE in ASyS patients.

During recent years, our research group and others have explored CV surrogate markers [17] and their predictors in patients with several rheumatic conditions, in an attempt to find valid tools to assess CV risk (Supplementary Material, available at Rheumatology online) [18–21]. However, until now no study group has examined established markers of CV risk, such as arterial stiffness or carotid atherosclerosis in ASyS. Particularly, stiffness of the aorta, a modifiable, independent predictor of CV risk, which can be measured via carotid-femoral pulse wave velocity (cfPWV) and carotid sonography, have not been examined systematically in this setting. The predictive value of cfPWV regarding CV events has nevertheless been repeatedly shown in the general population and this marker is considered to be the gold standard for the measurement of aortic stiffness [22]. Moreover, carotid sonography is an established tool of carotid atheromatosis, compliance and resistance assessment.

Thus, the aims of this pilot prospective multicentre study [RIsk of CARdiovascular Disease in Antisynthetase syndrome (RI.CAR.D.A.)] were to compare aortic stiffness and carotid status of ASyS patients with healthy controls and to explore a possible association of cfPWV and carotid sonography indices with ASyS parameters and traditional CV risk factors. Moreover, we sought to examine the feasibility of SCORE/mSCORE to identify ASyS patients at high CV risk, compared with well-validated CV surrogate markers [23, 24].

Methods

Study populations

A total of 66 ASyS patients being treated in six German rheumatology centres were recruited in a prospective cross-sectional multicentre study setting. As control subjects, 88 hospital co-workers without underlying inflammatory diseases were recruited, who freely responded to an open call for study participation. All of the patients and controls underwent a cfPWV examination. Moreover, B-mode and Doppler sonography of the common- (CCA) and internal- (ICA) carotid arteries were conducted in a subset of patients and controls. SCORE was assessed in all eligible patients and controls (age: 40–70 years). Exclusion criteria in both groups were: malignancy, pregnancy, age <18 years, kidney failure and BMI>45 kg/m².

According to criteria described by the American and European Network of ASyS (AENEAS) we included patients with positive anti-ARS (anti-Jo-1, -PL-12, -PL-7, -EJ, OJ, -asparaginyl-tRNA synthetase) and one or more manifestations among arthritis, ILD and myositis [25]. Patients gave their informed consent and the assessment was reviewed and approved by the local standing committee for ethical conduct, in adherence to the Helsinki Declaration.

Data collection

Next to epidemiological and anthropometric data, we documented the use of immunosuppressants, antihypertensive drugs, statins and nicotine in both groups.

Muscle involvement was confirmed by MRI and/or muscle biopsy. Pulmonary function tests and chest X-rays in two planes were performed in order to screen for ILD. In case of pathologic results in these two examinations and/or presentation of symptoms/signs of ILD (i.e. cough, dyspnoea), a high-resolution CT scan was additionally performed. Arthritis occurrence (joint swelling or tenderness) was evaluated clinically. Presence of RP was assessed via patient’s history and/or by a cold test. Myositis, arthritis and ILD were considered positive if they were present at any time point during the disease course. Laboratory parameters, such as creatine phosphokinase (CPK), inflammation markers (CRP, ESR) and glomerular filtration rate (GFR) were assessed on the day of the CV examinations. Finally, immunoserological markers were assessed by well-established methods: ANAs were detected on the HEp-2 cells by indirect immunofluorescence. ELISA and line blot testing were used for the assessment of serum concentrations of antibodies such as RF, double-stranded DNA, SSA-Ro, SSB-La, Sm, Scl-70, U1-RNP. Anti-ARS were finally assessed by radioimmunoprecipitation or line blot assays (Supplementary Material, available at Rheumatology online) [26, 27].

cfPWV and carotid sonography

cfPWV was assessed by trained blinded medical stuff using a validated non-invasive oscillometric device (Vicorder^®, SMT medical, Wuerzburg, Germany), as described previously [20].

The cfPWV examination protocol was carried out in accordance with the manufacturer’s instructions and the expert consensus document on arterial stiffness [22]: cfPWV was calculated from the device software by dividing (0.8 ×) travelled pulse wave distance (right CCA to the right femoral artery) by pulse transit time [in metres per second (m/s)]. The average value of three measurements was documented. cfPWV values >10 m/s were considered as an indicator of increased CV risk [22].

Carotid sonography was performed by an experienced examiner [K.T., rheumatologist, certified by the German Society of Ultrasound in Medicine (DEGUM)], who was blinded for other study measurements. cIMT measurements were performed using a linear transducer (4–13 MHz) of a MyLab70-US device (Esaote^®) operating at a frequency of 13 MHz. cIMT was measured in the end-diastolic phase at three subsequent spots of both CCA, ∼1 cm proximal to the carotid bulb, and the maximum value of these examinations was used for further analysis. Moreover, IMT measurements were performed bilaterally in the carotid bulbs, in order to control for plaque presence, defined as localized thickenings >1.2 mm [28, 29]. Patients with IMT>0.9 mm in the CCA and/or ≥1 plaque(s) were classified as having subclinical carotid atherosclerosis (SCA) [28, 30].

Finally, duplex of both mid-CCA and proximal ICA was performed bilaterally (transverse and longitudinal planes). Peak systolic (PSV) and end-diastolic velocity (EDV) were assessed by Doppler analysis: a pulsed wave flow spectrum was recorded and frozen after 5 s of multiple identical waveforms. Resistance (RI) and pulsatility indices (PI) were calculated automatically by the US software according to the Pourcelot [RI = (PSV – EDV)/PSV] and Gosling’s formulas [(PSV – EDV)/mean flow velocity (MFV)], respectively. Maximum Doppler measurements of both CCA and ICA were documented.

SCORE calculation

SCORE was calculated according to the European guidelines on CV disease prevention [13, 31]. In the case of RA, EULAR suggests the multiplication of SCORE by a factor of 1.5 in order to assess the true CV risk (mSCORE). As there are no specific guidelines for CV risk assessment in ASyS, we applied this multiplication strategy. As proposed by the guidelines, patients with SCORE/mSCORE values >5% were considered to be high risk [30].

Statistical analysis

The assumption of normality of distribution was evaluated by the Shapiro-Wilk numerical test and quantile-quantile plots. Continuous variables were presented as mean (s.d.) when they were normally distributed and as median (25th/75th percentiles) when they were skewed. Categorical variables were summarized as absolute (n) and relative (%) frequencies. A comparison of categorical variables was performed through a chi-squared test. Diagnostic performance of SCORE/mSCORE in comparison to cfPWV and carotid sonography was examined by receiver operating characteristics (ROC).

The differences of cfPWV, PI, RI, cIMT and SCORE between ASyS patients and controls were evaluated through t test when the variables were normally distributed and Mann–Whitney U test when they were skewed. These tests were also used to evaluate the association between all examined CV surrogate markers and categorical variables with two categories. In order to assess the correlation between CV surrogates and continuous characteristics, the Spearman’s (rho) or Pearson’s (r) correlation coefficients were used.

Furthermore, difference of CV surrogate markers between ASyS and control group after controlling for possible confounding factors was examined by multiple linear regression analysis. To search for confounding factors we calculated the B coefficient of the corresponding marker before and after the inclusion of different variables in the multivariate regression analysis model. We considered a variable to be a confounder if its inclusion caused a change of the B coefficient of ≥10% (assumed as the maximum superior limit) [32]. A probability value below 0.05 was considered statistically significant. All statistical calculations were performed using IBM SPSS^® 23.0 software (USA).

Results

We performed cfPWV measurements in 66 ASyS patients and 88 healthy controls. A subset of the patients (n = 28) and the control subjects (n = 36) underwent an additional carotid sonography. A total of 50 patients and 63 controls were eligible for SCORE assessment. Descriptive characteristics of patients with ASyS and controls and their statistical differences are seen in Tables 1 and 2.

Association between group status (ASyS vs control): cfPWV, Doppler indices and score

cfPWV average was significantly higher in the patient group compared with the control group [8.58 (1.89) vs 6.89 (1.30); P<0.001] (Fig. 1A). Age, diabetes and heart rate could be statistically identified as potential confounders. Conversely, treatment with antihypertensive drugs or statins, mean arterial pressure (MAP) and traditional CV risk factors, such as nicotine abuse, hyperlipidaemia, BMI, did not show a significant confounding effect on the results. The age-, diabetes- and heart-rate adjusted logistic regression model revealed that cfPWV remained significantly higher in the patient group in comparison to the control group (0.393, 95% CI: 1.061, 2.067; P_adj=0.021). Thus, in the present study ASyS patients had higher cfPWV values than controls.

In the same manner, cIMT and calcification area values were statistically significantly higher in the ASyS subgroup than in the control subgroup [0.97 (0.17) vs 0.78 (0.13) and 0.25 (0.15–0.43) vs 0.10 (0.06–0.17); P<0.001 and P =0.002, respectively] (Table 2, Fig. 1B). Age, MAP and high-density lipoprotein (HDL) showed a confounding effect on cIMT. Logistic regression analysis showed that cIMT was statistically significantly higher in the patient than in the control group, even after adjustment for these confounding factors (P_adj=0.003).

Moreover, examined carotid Doppler indices (RI and PI of the ACC and ACI, respectively) were significantly higher in patients compared with controls (Table 2, Fig. 1B). Conversely, SCORE and mSCORE were not significantly different between patients and control group [2 (1–2.25) vs 1 (1–2) and 3 (1.5–3.38) vs 1.5 (1.5–3.00), respectively (both; P =0.051)].

CV events and associated factors of cfPWV and carotid sonography

A total of 9/66 patients (13.6%) had past CV events (2/9 myocardial infarction, 2/9 ischaemic stroke, 4/9 pulmonary embolism, 1/9 arterial thrombosis). Count of these patients was unfortunately low for a statistical analysis. However, mean cfPWV values, mSCORE and SCA were higher in patients with CV events compared with patients without [9.25 (1.88) m/s vs 8.48 (1.87) m/s, 3.86 (2.27) vs 2.89 (2.02) and 91.7% vs 75%, respectively].

cfPWV

Among ASyS patients, unadjusted statistical analyses showed a moderate association of cfPWV with age (rho = 0.567; P<0.001) and a poor association with disease duration (rho = 0.299; P=0.015) (Table 3). Moreover, patients with diabetes showed higher cfPWV values than patients without [8.35 (1.87) vs 10.07 (1.43); P=0.015]. Some inverse correlations of cfPWV [GFR (rho = -0.443; P<0.001), CPK (rho = -0.26; P=0.047)] could also be seen.

Among controls, cfPWV correlated moderately with age (rho = 0.679; P <0.001), MAP (rho = 0.413; P <0.001), BMI (rho = 0.348; P <0.001) and SCORE/mSCORE (r = 0.4; P =0.002), respectively. Moreover, subjects receiving antihypertensive drugs had higher cfPWV values compared with their antihypertensive drug-free counterparts [7.75 (1.31) vs 6.66 (1.21); P =0.002] (Table 4, Supplementary Material, available at Rheumatology online).

To control the established significant correlations, statistical adjustments for age (a known crucial influence factor of cfPWV) using a multivariate logistic regression model were performed. Association between cfPWV and the presence of diabetes remained statistically significant in the patient group (−1.21, 95% CI: −2.385, −0.041; P_adj =0.043). Moreover, correlations of cfPWV with MAP and BMI remained statistically significant in the control group (0.025, 95% CI: 0.005, 0.044; P_adj=0.012 and 0.076, 95% CI: 0.032, 0.121; P_adj=0.001, respectively).

Carotid sonography parameters

Results of correlation analyses between cIMT, Doppler indices (RI and PI for CCA, ICA) and patient characteristics can be seen in Tables 3 and 4. Corresponding values for the control group can be seen in Supplementary Tables S1 and S2, available at Rheumatology online.cIMT associated in the ASyS subgroup strongly with age (r = 0.697; P <0.001) and moderately with disease duration (r = 0.396; P =0.037). CCA-RI was higher in patients receiving antihypertensive drugs 0.80 (0.76–1.08) vs 0.72 (0.70–0.79) (P =0.02). Moreover, CCA-RI correlated moderately strongly with CPK values (r = 0.629; P =0.012). CCA-PI correlated also moderately with CPK (r = 0.557; P =0.039) and was found to be higher in female patients [2.13 (2.09–2.14, IQR) vs 1.76 (1.46–1.84, IQR); P =0.035].

ICA-RI and ICA-PI were higher in patients with hyperlipidaemia [0.73 (0.06) vs 0.64 (0.667); P =0.005 and 1.61 (1.25–1.74, IQR) vs 1.07 (0.94–1.35, IQR); P =0.013, respectively] and ILD [0.69 (0.07) vs. 0.60 (0.06); P =0.039 and 1.41 (1.06–1.63) vs 1.02 (0.90–1.15); P =0.039, respectively].

Comparison of SCORE/mSCORE with cfPWV and SCA

A total of 28 ASyS patients underwent carotid sonography. SCA could be diagnosed in 24/28 (85.7%) of the patients. A total of 23/28 of the patients were eligible for the calculation of SCORE/mSCORE and only 0/23 (0%) and 1/23 (4.3%) showed SCORE >5% and mSCORE >5%, respectively.

A total of 66 ASyS patients received a cfPWV measurement and 14/66 (21.2%) patients had cfPWV > 10 m/s. In total 50/66 patients were eligible for the calculation of SCORE/mSCORE. Of these patients 1/50 (2%) and 5/50 (10%) showed SCORE >5% and mSCORE >5%, respectively.

ROCs were performed in order to test the diagnostic performance of SCORE/mSCORE in comparison to cfPWV >10 m/s and SCA, respectively. The area under the curve (AUC) was 0.562 (95% CI: 0.392, 0.731) for the ROC between cfPWV >10 m/s and SCORE/mSCORE, pointing to a poor agreement between the two parameters (Fig. 3). In the same sense AUC was 0.631 (95% CI: 0.3, 0.962) for the ROC between SCA and SCORE/mSCORE (Fig. 2).

Discussion

Surrogate markers of angiopathy and CV risk

In this study we could show that ASyS patients had higher aortic stiffness and SCA than healthy controls, even after adjusting for confounding factors of cfPWV and cIMT, respectively. Moreover, we found that SCORE/mSCORE agreed poorly with these markers and could thus underestimate CV risk, and that carotid Doppler indices correlated with a myositis activity marker (CPK) and with the presence of ILD.

To our knowledge, this is the first study to examine the gold standard CV surrogate markers in ASyS and also the first to evaluate the feasibility of SCORE/mSCORE in detecting ASyS patients at high risk.

In general, data concerning markers of CV risk in ASyS are scarce. In particular, CV surrogate markers have been examined solely in a few studies regarding IIM. Given the fact that antibody status is not always reported, one cannot be sure whether ASyS patients were included in these explorations. But even in the case of reported inclusion of anti-ARS-positive patients, the counts of patients are low. Moreover, no subgroup analyses that could allow valid conclusions regarding the specific entity of ASyS have been performed. For example, in the study by Vincze et al. [33], brachial artery flow-mediated dilatation (FMD) (an endothelial dysfunction marker), cIMT, augmentation index and PWV were examined in 27 IIM patients. Among them, only six patients were anti-ARS (anti-Jo-1) positive. FMD, cIMT and PWV were not significantly increased in the IIM group. Nevertheless, FMD was found to be significantly decreased in the DM subgroup, compared with the control group. One of the anti-Jo-1-positive patients had developed severe pulmonary arterial hypertension followed by fatal heart failure. The remaining five ASyS patients did not show any clinical signs of CV disease.

Further IIM-related studies of CV risk examined mainly surrogate markers of cardiac (and not arterial) involvement, such as cardiac magnetic resonance [34], traditional echocardiography with tissue Doppler imaging [35] or speckle-tracking echocardiography [36]. In these studies, myocardial inflammation and reduced left ventricular function [34], early diastolic dysfunction [35] and subclinical systolic impairment [36] could be respectively shown in patients with IIM. However, in all three studies no antibody status was reported [34–36].

In ASyS we are aware of only one study having addressed a CV-associated aspect indirectly, by examining and finding a higher prevalence of metabolic syndrome (MetS) [37]. This exploration focussed on the documentation of traditional CV risk factors in ASyS.

Clinical cardiac manifestations and mortality in IIM and ASyS

In our study, ASyS patients showed statistically higher cIMT and calcification areas as markers of carotid atherosclerosis, compared with control subjects. High calcification scores could also be found in 76 IIM patients studied by Diederichsen et al. who, however, focussed on the coronary arteries [38]. Interestingly, 44% of these patients had myositis-specific antibodies, with the vast majority being anti-ARS.

Data regarding CV-associated mortality can be indirectly drawn from a retrospective analysis of a British myositis database (46 IIM patients, 7/44 Jo-1-positive) [39]. Six patients from this cohort died during follow-up (21 years). Causes of death were CV disease-associated in 50% of the cases and the 5-year and 10-year survival rates were 95% and 83.8%, respectively. Nevertheless, no subset analysis focussing on the Jo-1-positive patients was reported. In a further study examining 76 patients with IIM, 44 patients (58%) were tested for the presence of Jo-1 antibodies and 23% of these were positive. In this study, ECG changes (33%) and congestive heart failure (25%) were reported as the most important CV manifestations. A total of 13 patients died during follow-up, with an overall case fatality rate of 17.1%.

Established associations of CV surrogate markers among patients and controls

In ASyS patients and controls, cfPWV and cIMT correlated directly with age. This correlation has been described in healthy persons [40, 41] and patients with different rheumatologic diseases [18, 19, 21]. In fact, age is one of the most important factors that can influence cfPWV, and cIMT and atherosclerosis are known to increase with advancing age [42–44]. Moreover, in our control group, an association between cfPWV and MAP could be found. Stretching of the artery wall through high blood pressure leads to an increase in stiffness, which is represented by higher cfPWV values [45]. cfPWV correlated also with BMI in both groups and with diabetes in the ASyS group, signalling the effect of metabolic factors on aortic stiffness and thus overall CV risk.

Furthermore, in the Doppler subgroup, both CCA-RI and CCA-PI associated significantly with CPK values. Moreover, ICA-RI and ICA-PI were higher in patients with ILD. Until today, no study has examined the significance of these markers in ASyS or IIM. However, carotid PI has been found to associate positively with cerebral small-vessel disease and higher risk of stroke in the general population [24, 46]. Interestingly, a recently published registry-based study including all Swedish patients with newly diagnosed IIM found a higher risk of both haemorrhagic and ischaemic stroke compared with general population comparators [47]. Moreover, Johnson et al. found that risk of death was statistically significantly higher among IIM patients with ILD in a retrospective cross-sectional analysis [48]. Given the fact that 50% of the ILD patients with available antibody status data were anti-ARS-positive, one could postulate that ASyS was present in this cohort and was associated with higher mortality rates. Moreover, RI is a combined marker of arterial resistance and compliance [49] and carotid RI has been shown to correlate with the degree of generalized atherosclerosis [23]. Thus, the established associations between Doppler indices and both myositis activity and ILD in our multicentre cohort seem plausible and PI/RI may have value as CV and cerebrovascular screening tools.

This study has some limitations. First of all, SCORE/mSCORE were only compared with CV and atherosclerosis surrogates and not with morbidity or mortality data. Thus, underestimation of CV risk through SCORE can only be hypothetized. However, numerous studies have shown the predictive value of cfPWV and carotid sonography in the general population and these markers are suggested as additional tools for risk stratification (level of evidence A, Recommendation IIa) [50]. Another limitation was the relatively low count of patients in the Doppler subgroup and study subjects who had diabetes. However, ASyS is a rare rheumatologic condition and cohort size cannot been characterized as small compared with other studies on this disease. Moreover, presence of CV risk factors in both groups and of immunosuppressants/glucocorticoids in the patient group may have had an influencing effect on the results. Despite the fact that adjusted statistical analyses for the most important confounding factors were performed, future control and confirmation of these results is important.

To conclude, this is the first report of higher aortic stiffness and carotid atherosclerosis in patients with ASyS. This finding is of significance, since traditional CV assessment scores such as the mSCORE may underestimate CV risk. Novel surrogate markers in this setting could assist detection, valid monitoring and early initiation of therapy in ASyS patients and thus improve disease outcome.

Special thanks to the medical assistants Mrs Nicole Dirvonskis and Mrs Susanne Dietz of the ACURA Rheumatology Center Rhineland-Palatinate, Germany and to all participating medical centres. Results described in this manuscript are part of the doctoral thesis of Ms. Anna Klonowski.

Acknowledgments

Funding: No specific funding was received from any bodies in the public, commercial or not-for-profit sectors to carry out the work described in this manuscript.

Disclosure statement: The authors have declared no conflicts of interest.

Supplementary data

Supplementary data are available at Rheumatology online.

References