Introduction

Individuals with adult onset systemic lupus erythematosus (SLE) were first identified as having an increased risk of cardiovascular disease (CVD) over 35 years ago by Urowitz et al. [1]. In subsequent years, the increased risk of cardiovascular (CV) events among adults with SLE has been confirmed in a number of prospective studies [2–4] which have demonstrated that individuals with SLE have a higher increased risk of CV events than their unaffected peers.

The increased risk of CVD in chronic inflammatory diseases is well recognised, and several adult studies have shown an abnormal vascular phenotype in SLE. There is evidence to suggest that the change in phenotype may be due not only to disease-specific risk factors (such as severity and duration of disease [5–7], renal impairment [8–10] and medications [1, 11–17]) but also to a clustering of traditional risk factors (such as increasing age, obesity, smoking and reduced physical activity [15, 18–20]).

Children with SLE tend to have a more aggressive disease than adults, with a higher incidence of arthritis, lupus nephritis (LN) and neurological involvement [21]. Since disease severity and duration are closely linked to CVD in the adult SLE population, it is reasonable to assume that the paediatric population would similarly be at increased risk of CVD. Despite the wealth of data available on CVD risk in adults with SLE, there is a dearth of information about CVD risk in children with juvenile-onset SLE (JSLE), and most of our knowledge is extrapolated from adult studies. Although childhood lupus is often a more aggressive disease than its adult form, the greater plasticity of children’s vessels may protect them from vascular disease, making extrapolations from adult studies difficult; published studies [22–27] to date have reported conflicting results. The cross-sectional study reported here was designed to establish the baseline CV risk of a cohort of children with JSLE prior to subsequent longitudinal follow-up studies. In addition, since children should be free of the classical risk factors for CVD, studying their risk profile may add valuable information to the adult SLE literature.

Methods

Recruitment

Children with JSLE, who fulfilled ≥4 of the American College of Rheumatology classification criteria [28] and who attended the nephrology or rheumatology services in Great Ormond Street Hospital for Children NHS Foundation Trust were invited to take part in the study (referred to as the CV JSLE Cohort Study). Those aged <5 years and diagnosed with end-stage kidney disease or within 6 weeks of a major surgical procedure were excluded.

The study was approved by the Institute of Child Health and Great Ormond Street Hospital research ethics committee and The Children’s University Hospital, Dublin.

Measurement of carotid artery intima media thickness

In this study carotid artery intima media thickness (cIMT) was measured by B-mode ultrasound of both common carotid arteries using a single ultrasound machine (Vivid7, GE Medical, Horton, Norway) equipped with a 12-MHz linear-array transducer [29]. Following visualisation of the common carotid artery, a 15-heartbeat cine-loop recording was stored for offline analysis using Digital Imaging and Communications in Medicine software (DICOM; National Electrical Manufacturers Association, Rosslyn, VA). The mean IMT of a 0.5- to 1-cm-long segment was measured 1 cm proximal to the carotid bifurcation, using an automated edge detection system, on two separate images of the right and left common carotid artery from true end-diastolic frames, as determined by minimum vessel diameter. Approximately 120 individual measurements were made across a 1-cm length of artery; this measurement was repeated for three different cardiac cycles and the average of the three measurements from right and left was recorded.

Measurement of pulse wave velocity

Pulse wave velocity (PWV) was determined by using a pressure sensor over the carotid and femoral arteries and calculated as the pulse wave travel distance divided by the time difference between the rise delay of the distal and proximal pulse, as measured by the R wave on the electrocardiogram. In this project, PWV was calculated by SphygmoCor software (AtCor Medical, Sydney, Australia) using the intersecting tangent algorithm. The pulse travel distance was calculated as the distance from the suprasternal notch to the recording point at the femoral artery via the umbilicus minus the distance from the suprasternal notch to the recording point at the carotid artery [30].

Data collection

A pilot version of the questionnaire was trialed in The Children’s University Hospital, Dublin and was used to record demographic data, treatment information and risk factors for CVD. Routine blood samples were processed by the departments of Haematology, Immunology and Clinical Chemistry at Great Ormond Street Hospital for Children NHS Foundation Trust.

Body mass index (BMI) and percentiles were calculated using the Centre for Disease Control and Prevention BMI-for-age growth chart (www.cdc.gov). A positive family history of CVD was defined as per the Expert Panel on Integrated Guidelines for Cardiovascular health and Risk Reduction [31] as a CV event in a male relative before 55 years of age or in a female relative before 65 years of age. Family history included siblings, parents, uncles, aunts, grand-parents, grand-aunts and grand-uncles. Estimated glomerular filtration rate (eGFR) was calculated by the Schwartz formula with a K value of 34, as calculated locally [32].

pBILAG–2004 score

Many of the children recruited to this study (CV JSLE Cohort Study) were also co-recruited to the UK JSLE Cohort Study [21]. As part of the protocol the paediatric adaptation of the 2004 British Isles Lupus Assessment Group (pBILAG–2004) disease activity index was measured at 3-month intervals and included in the analysis where available. The pBILAG–2004 index was developed as an intention-to-treat index; it has been validated in a UK paediatric cohort [33] and shown to be more sensitive than the SLE Disease Activity Index (SLEDAI)–2000 to detect active disease [34]. It is an ordinal scale index with nine systems (Constitutional, Mucocutaneous, Neuropsychiatric, Musculoskeletal, Cardiorespiratory, Gastrointesinal, Opthalmic, Renal and Haematological) which divides disease activity into five categories where A represents very active disease, B represents moderate disease, C represents mild stable disease, D represents previous disease and E represents no current or previous disease. A continuous total score has been developed and validated [35] and was used in this study to represent systemic disease burden.

Calculation of physical activity

The International Physical Activity Questionnaire (IPAQ) [36] was used to estimate the physical activity of the study cohort. This short version of this questionnaire was chosen as a well-developed, internationally recognised instrument which enables scoring of the cohort by metabolic equivalents of task (MET). This questionnaire proposes three levels of physical activity: “low” where there is no activity or <600 MET min/week are recorded; “moderate” where 600–1500 MET min/week are recorded; “high” where >1,500 MET min/week are recorded. This scoring protocol was recently used by Katz et al. [37] to show that 28 % of 138 women with SLE with a mean age of 18 years scored <600 MET min/week.

Statistics

For normally distributed data results are reported as the mean ± standard deviation (SD), and comparisons between means were made using unpaired t test results (2-tailed). For non-normally distributed data, results are reported as the median ±  interquartile range (IQR), and comparisons between medians were made using the Mann–Whitney test. Correlations were made using Pearson’s correlation co-efficient or Spearman’s test depending on data distribution.

Within this project a p value < 0.05 was determined to have reached significance. Statistical analyses were performed in GraphPad Prism version 6.0b for Mac OS X, and multiple regression analysis was performed using GraphPad InStat version 3.10 32bit for Windows (GraphPad Software, La Jolla, CA; www.graphpad.com).

Results

Demographics

Forty-five children were recruited to our study, among whom 39 were girls. The median age of the children was 9.75 (IQR 7.23–12.6) years at diagnosis and 14 (IQR 12.2–15.5) years at recruitment. Forty-four consented to have all vascular scans performed, and one patient was unable to remain sufficiently still for the PWV scanning. Thirty-five children of these children were also recruited to the UK JSLE Cohort Study; pBILAG-2004 scores were calculated and vascular assessments were performed for 28 of these children [Electronic Supplementary Material (ESM)]. The control group comprised subjects who had participated in previous studies [38] from our centre and who were matched for historical age. Demographic details of patients and controls are outlined in Table 1.

Table 1 Demographic, clinical and anthropometric characteristics of children and young adults with juvenile-onset systemic lupus erythematosus and controls
Full size table

Nineteen children identified themselves as Afro-Caribbean, of whom 13 identified themselves as of African descent, two as Black British and four as Black Caribbean. Thirteen children identified themselves as Caucasian and ten as Asian (5 Indian, 1 Pakistani 1 Chinese, 3 “other Asian”). One child was Afghani, one was Turkish Cypriot and one declined to answer this question. The ethnic composition of our patient group differs from that of the overall UK JSLE Cohort, as outlined in Table 2, where the majority of children were Caucasian and almost one-third were Black Afro-Caribbean. By contrast, of the 45 children enrolled in our study (CV JSLE Cohort Study in Table 2), 19 were Black Afro-Caribbean and 13 were Caucasian, which reflects the ethnic diversity of London relative to the rest of the UK.

Table 2 Comparison of patient demographic data for the UK Juvenile-onset Systemic Lupus Erythematosus (JSLE) Cohort Study and the Cardiovascular (CV) JSLE study
Full size table

Median disease duration was 39 (IQR 20.5–64.5) months. Disease activity was assessed using the pBILAG–2004 disease activity score within 3 months of recruitment. The median pBILAG-2004 score was 3.5 (IQR 1.25–5.75), which is in line with that of the treated national cohort group (UK JSLE Cohort Study) which had a median pBILAG–2004 score at 12 months after diagnosis of 2 (IQR 1–4).

Traditional CV risk factors

No children admitted to cigarette smoking, and one child had started taking the oral contraceptive pill 1 month previous to recruitment. Ten children had a positive family history of CVD as defined by Expert Panel on Integrated Guidelines for Cardiovascular health and Risk Reduction.

The mean serum cholesterol level was 4.4 (IQR 3.95–5.2) mmol/l, serum triglycerides were 1.04 (0.76–1.93) mmol/l, serum high-density lipoprotein (HDL) was 1.28 ± 0.46 mmol/l, low-density lipoprotein (LDL) was 2.35 (2.14–2.99) mmol/l and very-low-density lipoprotein was 0.64 ± 0.4556 mmol/l. Nine children had a serum cholesterol >5 mmol/l, eight had a serum HDL of <1 mmol/l and four had a serum LDL of >3.5 mmol/l.

For those children with a serum cholesterol of >5 mmol/l, mean cIMT was 0.47 ± 0.05 mm and mean PWV was 5.49 ± 0.7 m/s; for those with a serum cholesterol of <5 mmol/l, mean cIMT was lower at 0.45 ± 0.04 mm and PWV was also lower at 5.24 ± 0.8 m/s. Neither result reached statistical significance. Abnormalities in serum lipids were correlated with duration of disease, younger age of diagnosis, mean cIMT, PWV, eGFR, increased systolic blood pressure (SBP) and increased BMI. Those children with a serum cholesterol of >5 mmol/l were on a lower mean dose of prednisolone (0.26 ± 0.27 mg/kg) than those with a serum cholesterol of <5 mmol/l (4.28 ± 5.9 mg/kg) (p = 0.054), possibly indicating that higher cholesterol was associated with disease activity rather than secondary to medications.

Mean BMI percentile was 65.63 ± 28.8. The median physical activity score was 1773 (IQR 676–2854) METs, which correlated with duration of disease (r 2 = 0.33, p = 0.04). The median physical activity score was 1,773 (IQR 676–2,854) METs, which correlated with duration of disease (r 2 = 0.325, p = 0.038). These results correspond to a high level of activity, but there was significant heterogeneity within the group. Interestingly, children with a positive family history of CVD had a higher median MET score than those with no family history of CVD [1,738 (IQR 677–2,741) vs. 568 (IQR 0–1,689) METs, respectively], but this difference did not reach statistical significance. Not surprisingly, there was a strong correlation between BMI percentile and physical activity. The median BMI percentile for children with a low MET score, i.e. <600 MET, was 88.2 (IQR 49.9–97.5) and for those with a high MET score, i.e. >1500 MET, it was 65.8 (32.5–85.5); this difference did not reach statistical significance.

Vascular phenotype

Carotid intima media thickness

Forty-four children with JSLE had cIMT measurements; the control group consisted of 40 patients. Mean cIMT for the control group and JSLE group was 0.37 ± 0.06 and 0.45 ± 0.04 mm, respectively (Fig. 1). Children with JSLE had a significant increase in mean cIMT of 0.08 mm when compared with the controls (p < 0.0001.) There was no statistically significant association between cIMT and upper or lower quartiles of age, METs, pBILAG–2004 score, pBILAG–2004 renal score or disease duration. An increase in cIMT was also not associated with male gender or a family history of CVD. However, cIMT did correlate weakly with the number of sitting min/week (r 2 = 0.234, p = 0.176) and BMI percentile (r 2 = 0.276, p = 0.077).

Fig. 1
figure 1

Comparison of carotid intima media thickness (cIMT) between controls and patients with juvenile-onset systemic lupus erythematosus (JSLE)

Full size image

Pulse wave velocity

Forty-three children with JSLE had PWV measurements; the control group consisted of 40 children. Mean PWV for the control and patient groups was 5.34 ± 0.97 and 5.27 ± 0.88 m/s, respectively (Fig. 2). Children and young adults with JSLE had a 1.2 % lower PWV than the controls, but this difference was not statistically significant (p = 0.77). Increased PWV was associated with disease duration (r 2 = 0.37, p = 0.02), age (r 2 = 0.35, p = 0.02), SBP (r 2 = 0.35, p = 0.02), LDL (r 2 = 0.37, p = 0.048) and eGFR (r 2 = 0.331, p = 0.06).

Fig. 2
figure 2

Comparison of pulse wave velocity (PSV) between controls and patients with juvenile-onset systemic lupus erythematosus (JSLE)

Full size image

There was no statistically significant association between PWV and the upper or lower quartiles of age, pBILAG–2004 score, pBILAG–2004 renal score, prednisolone dose, METs or disease duration. An increase in PWV was also not associated with male gender or a family history of CVD (Table 3).

Table 3 Demographic, clinical and anthropometric characteristics of children and young adults with juvenile-onset systemic lupus erythematosus (JSLE) and a normal or grossly elevated urine albumin-to-creatinine ratio
Full size table

Renal function and hypertension

All 45 JSLE patients were normotensive at the time of the study. Twenty-six were not taking any anti-hypertensive medication, and 15 were on a single medication [enalapril (9 patients), amlodipine (4) furosemide (1) and irbesartan (1)], two were on two medications, one was on three medications and one patient required four medications to maintain normal blood pressure.

There was a significant association between the use of antihypertensives and an increase in cIMT. The cIMT in the normotensive group not treated with antihypertensives and in the normotensive group treated with antihypertensives was 0.45 ± 0.04 and 0.47 ± 0.05 mm, respectively (p = 0.04). The relationship between antihypertensive use and cIMT remained significant even after controlling for total pBILAG–2004 score, prednisolone dose, age, family history of CVD and gender (p = 0.04). In the absence of antihypertensive medication, the mean cIMT for patients with JSLE was significantly higher than that of controls at 0.45 ± 0.04 mm (p < 0.0001). (see Fig. 3).

Fig. 3
figure 3

Comparison of carotid intima media thickness (cIMT) between controls and patients with juvenile-onset systemic lupus erythematosus (JSLE) according to use of anti-hypertensive medication

Full size image

There was a significant association between the use of antihypertensives and an increase in PWV. The PWV in the normotensive and anti-hypertensive groups was 5.13 ± 0.6 and 5.69 ± 1.13 m/s, respectively (p = 0.043). Similar to the cIMT findings, the PWV was increased in these children compared to the controls, even in the absence of anti-hypertensives, from 5.34 ± 0.97 to 5.13 ± 0.06 m/s; however this difference was not statistically significant (p = 0.132).

Those children defined as hypertensive may only have had a low level of hypertension (HTN) and proteinuria. The median urine albumin-to-creatinine ratio (UaUc) for the JSLE cohort was 2.9 (IQR 1.2–55.3) mg/mmol. In an effort to assess the effect of renal dysfunction on the vascular phenotype we examined only those children with a grossly elevated UaUc (>20 mg/mmol) and compared their vascular phenotype with both the group as a whole and with a those with a UaUc closer to the normal range (Table 3). The median UaUc of the two subgroups were similar to that of the overall cohort with regards to age, gender, disease duration and BMI. Not unsurprisingly, the children with elevated UaUc were significantly more likely to be concurrently treated with antihypertensive medication, most commonly an angiotensin-converting enzyme inhibitor (ACEi), which was used as much for the management of proteinuria as for the management of HTN.

With regards to their vascular phenotype, children with an elevated UaUc had an elevated cIMT [0.455 (IQR 0.43–0.49) vs. 0.435 (0.42–0.49) mm] and a prolonged PWV (5.564 ± 0.72 vs, 5.172 ± 0.92 m/s). However, the numbers were too small to reach statistical significance.

Median eGFR was 103.3 (IQR 95.65–113.3) ml/min/1.73 m2. Five children had an eGFR of <90 ml/min/1.73 m2 (26.3, 34.5, 66.8, 84.6 and 87.2 ml/min/1.73 m2, respectively). Median cIMT for the children with an eGFR of <90  ml/min/1.73 m2 ml/min/1.73 m2 and for those with an eGFR of <70 ml/min/1.73 m2 ml/min/1.73 m2 was 0.45 (IQR 0.4–0.5) and 0.45 (IQR 0.36–0.55) mm, respectively. There was no difference when compared to children with JSLE and an eGFR >90 ml/min/1.73 m2. The difference between medians was not significant (p = 0.556). However, median PWV for children with an eGFR of <90 ml/min/1.73 m2 and those with an eGFR of <70 ml/min/1.73 m2 was 5.7 (IQR 5.5–6.6) and 6.8 (IQR 5.7–6.9) m/s, respectively. This is an increase of 1.1 m/s when compared to children with an eGFR of >90 ml/min/1.73 m2 and JSLE, and this difference did reach statistical significance despite the small numbers (p = 0.0277).

Renal involvement was also assessed using the pBILAG–2004 score for those who had performed this test, and the data are summarised in Table 4. One child was diagnosed with severe renal disease activity (category A), five were diagnosed with moderate renal disease activity (category B), five were diagnosed with stable renal disease activity (category C), 16 were diagnosed with currently quiescent but previously active renal disease (category D) and only three of the 30 children tested were defined as no renal involvement ever (category E). Those with current LN (categories A/BC) had slightly increased cIMT (0.46 ± 0.06 mm) compared with those with quiescent LN (categories D/E; 0.45 ± 0.04 mm) (p = 0.73). Similarly, there was little change in the PWV (5.46 ± 0.95 vs. 5.43 ± 0.74 m/s; p = 0.78).

Table 4 Pulse wave velocity (PWV) and carotid intima media thickness (cIMT) by the pBILAG–2004 score
Full size table

Prednisolone and the vascular phenotype

There was significant increase in cIMT between controls and patients receiving no prednisolone, and those receiving the lower and upper quartiles of prednisolone (0.087 and 0.315 mg/kg, respectively). The cIMT of those receiving no prednisolone measured 0.43 ± 0.04 mm, which increased to 0.5 ± 0.04 mm for those on low doses and to 0.45 ± 0.05 mm for those on higher doses. There was a significant difference in cIMT between controls and those on all doses of prednisolone (p < 0.05) and between those on no prednisolone and the lower quartile (p = 0.0018) (Fig. 4).

Fig. 4
figure 4

Comparison of carotid intima media thickness (cIMT) between controls and patients with juvenile-onset systemic lupus erythematosus (JSLE) according to daily dose of prednisolone [quartiles: none lower, higher (mg/kg/day)]

Full size image

Ethnicity and the vascular phenotype

When analysed by ethnicity there was a significant difference between the Asian population and both the Caucasian and Afro-Caribbean population, despite the small overall numbers. One Caucasian child was unable to remain still for the PWV reading and one Afro-Caribbean child declined to consent to vascular scanning. Mean cIMT was highest in the Afro-Caribbean cohort (0.47 ± 0.04 mm) and lowest in the Asian cohort (0.429 ± 0.04 mm); the Caucasian cohort had a median cIMT of 0.448 ± 0.04 mm. This difference was statistically significant for the Afro-Caribbean cohort compared to the Asian cohort (p = 0.017), but it did not meet statistical significance when comparing the other cohorts (Fig. 5).

Fig. 5
figure 5

Comparison of carotid intima media thickness (cIMT) between patients with juvenile-onset systemic lupus erythematosus (JSLE) erythematosus by ethnic background

Full size image

There was a similar difference noted among the PWV results. Mean PWV was highest for the Afro-Caribbean cohort (5.339 ± 0.74 m/s) but lowest for the Caucasian population (5.225 ± 1.148). The high standard deviation suggests greater spread amongst this group. Mean PWV among the Asian cohort was 5.31 ± 0.69 m/s. None of these comparisons reached statistical significance.

Discussion

There is strong epidemiological evidence to show that adult onset SLE is associated with an increased risk of clinical and preclinical CVD. However, the increased prevalence of traditional risk factors in this cohort, such as smoking and obesity, makes it difficult to study the effect of SLE-specific factors and interventions, since many SLE patients will already have a significant lifetime CV risk due to non-modifiable risk factors, such as a lifetime of smoking. The aim of our study was to establish the risk of CVD in a population free from these traditional risk factors and, thereby, perform a closer analysis of disease-specific risk factors.

Patients with JSLE generally have more severe disease, with a greater prevalence of LN [21], and will have a longer lifetime burden of disease. To date, the results of published studies examining CVD in JSLE have been conflicting, and many of the published studies have lacked control groups and thus relied on the correlation of risk factors with changes in vascular measures.

cIMT is a measurement of the intima and media layers lining the carotid artery. It is a measure of very early structural changes in the artery and an important surrogate marker for atherosclerosis. Carotid–femoral PWV is a measurement of the time delay between the foot of the pulse pressure arriving at the carotid and femoral arteries, and it is a widely used method to determine arterial stiffness [39, 40]. This cohort of children and adolescents with SLE and normal renal function have an increased cIMT, i.e. a structural change in the vessel wall, but without an increase in PWV, which is a functional measurement of vessel health. Indeed, the median PWV was lower in our JSLE cohort than in the controls, suggesting alternate pathways for the development of CVD in JSLE or a degree of compensation in the early stages of SLE, which may be overwhelmed in time.

In a study designed to assess the effect of atorvastatin on the progression of subclinical atherosclerosis, as measured by cIMT, Schanberg et al. [22, 24] found a mean cIMT of 0.468 mm, similar to our data. Increased cIMT was associated with both a high and low dose of prednisolone, BMI, increased age and creatinine clearance. This study included children with mildly active disease (mean SLEDAI 4.6), with 36 % having LN and 34 % having HTN. In contrast, our patients showed a greater prevalence of active LN. Secondary analysis of the data reported by Schanberg et al. [22, 24] showed that atorvastatin may slow the progression of cIMT thickening only in a subgroup of pubertal patients with higher high-sensitivity C-reactive protein [23]. In contrast to our findings, Boros et al. [27] found an increase in PWV, but no increase in cIMT or flow-mediated dilatation, among a cohort of children with SLE. Although these authors showed an increased burden of traditional and non-traditional CV risk factors among the cohort, such as HTN, these did not correlate with changes in vascular measures. When subdivided by disease activity, El Gamal et al. [26] showed increased PWV in hypertensive patients with active SLE when compared to controls, but no difference between controls and patients with inactive disease as defined by SLEDAI.

Interestingly, Sozeri et al. [25] showed significantly increased cIMT and PWV in patients with lupus when compared with controls. In this study, cIMT correlated with disease activity but not with HTN or LN. However, in the ten patients (20 %) with HTN (8 of whom had left ventricular hypertrophy), the authors found an increased PWV; they also showed a marginal increase in PWV in those with biopsy-proven LN when compared to those with normal renal biopsies.

The increase in cIMT in our cohort was most marked in the patients treated with antihypertensives. Since the commonest antihypertensive in use was an ACEi, which is also used to control proteinuria, the difference between groups may be explained by renal involvement or severity of disease. However, there we found no correlation between cIMT or PWV and either the overall or the renal pBILAG–2004 score, or indeed the prednisolone dose, which could be presumed to be a reasonable surrogate marker for disease severity.

In this study, we found no difference in BMI percentiles between patients and controls, a low prevalence of sedentary behavior and no effect of smoking. Indeed, in contrast to their adult counterparts [37], this cohort of children with JSLE showed high levels of activity, with a median of 1,773 METs per week, corresponding to a high level of physical activity. Encouraging an active lifestyle during clinical encounters is one way in which paediatricians could potentially reduce their patients’ future burden of CVD.

Disease duration has been shown to increase the risk of a CV event in adults with SLE and has been associated with subclinical CVD as measured by coronary calcium scores [6, 7, 41–46]. Since the length of disease is correlated with CV risk, there is justifiable concern that children, whose disease has an earlier onset, will be at an increased risk as they age.

The difference in vascular phenotype between ethnicities is similar to that found in previous studies which showed that the presence of carotid plaque was higher in African American women with SLE compared with Caucasian women in contrast to studies of non-SLE subjects [47]. Likewise, Ghosh et al. observed increased fibromuscular dysplasia (FMD) and cIMT in 60 Indian patients with SLE who showed impaired FMD and abnormal IMT [48]. Our group has a greater proportion of African American patients than other paediatric studies [22–27], and this may have affected the results.

The authors acknowledge a number of limitations of this study: the numbers are small, the patient cohort is heterogeneous and the controls were historical. Nonetheless, this is a previously undescribed cohort taken from the largest referral centre for JSLE in the UK and we used a rigorous study protocol which will enable long-term follow up to determine if the increase in surrogate markers for CVD will result in clinical outcomes.

In summary, this study describes early structural changes in the blood vessels of children and adolescents with JSLE, which may be related to ethnicity and renal involvement and which are associated with CVD in adults. The patients in our study cohort had minimal modifiable risk factors and were physically active, limiting the potential for lifestyle advice. Patients should be made aware of this potentially increased risk as they grow up, and efforts should be made to encourage healthy lifestyle choices in order to decrease their CV risk from traditional, modifiable risk factors. Follow-up data from this cohort are needed to determine if the early compensatory changes resulting in normal PWV readings are overwhelmed with time and if the increase in pre-clinical markers of CVD are followed by hard CV endpoints.