Article Text

Download PDFPDF

Ultrasensitive serum interferon-α quantification during SLE remission identifies patients at risk for relapse
  1. Alexis Mathian1,
  2. Suzanne Mouries-Martin2,
  3. Karim Dorgham3,
  4. Hervé Devilliers4,
  5. Hans Yssel3,
  6. Laura Garrido Castillo3,
  7. Fleur Cohen-Aubart1,
  8. Julien Haroche1,
  9. Miguel Hié1,
  10. Marc Pineton de Chambrun1,
  11. Makoto Miyara3,
  12. Micheline Pha1,
  13. Flore Rozenberg5,
  14. Guy Gorochov3,
  15. Zahir Amoura1
  1. 1 Sorbonne Université, Assistance Publique–Hôpitaux de Paris, Groupement Hospitalier Pitié–Salpêtrière, French National Referral Center for Systemic Lupus Erythematosus, Antiphospholipid Antibody Syndrome and Other Autoimmune Disorders, Service de Médecine Interne 2, Institut E3M, Inserm UMRS, Centre d’Immunologie et des Maladies Infectieuses (CIMI-Paris), Paris, France
  2. 2 Centre Hospitalier Universitaire de Dijon, Hôpital François-Mitterrand, service de médecine interne et maladies systémiques (médecine interne 2), Dijon, France
  3. 3 Sorbonne Université, Inserm UMRS, Centre d’Immunologie et des Maladies Infectieuses (CIMI-Paris), Assistance Publique–Hôpitaux de Paris, Groupement Hospitalier Pitié–Salpêtrière, Département d’Immunologie, Paris, France
  4. 4 Centre Hospitalier Universitaire de Dijon, Hôpital François-Mitterrand, service de médecine interne et maladies systémiques (médecine interne 2) et Centre d’Investigation Clinique, Inserm CIC 1432, Dijon, France
  5. 5 Université de Paris, Assistance Publique–Hôpitaux de Paris, Hôpital Cochin, Service de Virologie, Paris, France
  1. Correspondence to Dr Alexis Mathian, Internal Medicine, Pitié-Salpêtrière Hospital, Paris 75013, France; alexis.mathian{at}


Objectives Maintenance of remission has become central in the management of systemic lupus erythematosus (SLE). The importance of interferon-alpha (IFN-α) in the pathogenesis of SLE notwithstanding, its expression in remission has been poorly studied as yet. To study its expression in remission and its prognostic value in the prediction of a disease relapse, serum IFN-α levels were determined using an ultrasensitive single-molecule array digital immunoassay which enables the measurement of cytokines at physiological concentrations.

Methods A total of 254 SLE patients in remission, according to the Definition of Remission in SLE classification, were included in the study. Serum IFN-α concentrations were determined at baseline and patients were followed up for 1 year. Lupus flares were defined according to the Safety of Estrogens in Lupus Erythematosus: National Assessment version of the Systemic Lupus Erythematosus Disease Activity Index Flare Index, whereas the Kaplan-Meier analysis and Cox regression analysis were used to estimate the time to relapse and to identify baseline factors associated with time to relapse, respectively.

Results Of all patients in remission, 26% displayed abnormally high IFN-α serum levels that were associated with the presence of antibodies specific for ribonucleoprotein (RNP), double stranded (ds)DNA and Ro/SSA60, as well as young age. Importantly, elevated-baseline IFN-α serum levels and remission duration were associated in an independent fashion, with shorter time to relapse, while low serum levels of complement component 3 and anti-dsDNA Abs were not.

Conclusion Direct serum IFN-α assessment with highly sensitive digital immunoassay permits clinicians to identify a subgroup of SLE patients, clinically in remission, but at higher risk of relapse.

  • systemic lupus erythematosus
  • interferon-alpha
  • remission
  • low disease activity
  • biomarker
  • relapse
  • flare

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Key messages

What is already known about this subject?

  • Serum interferon-alpha (IFN-α) levels have been poorly studied in systemic lupus erythematosus (SLE) patients in remission so far.

  • The single-molecule array assay is an ultrasensitive assay that quantifies directly IFN-α at attomolar concentrations.

What does this study add?

  • A quarter of SLE patients in remission display elevated serum IFN-α levels.

  • Elevated serum IFN-α levels constitute an independent predictive biomarker of lupus flares.

How might this impact on clinical practice or future devemopments?

  • Including serum IFN-α in the routine laboratory assessments in patients in remission will assist clinicians in identifying SLE patients at higher risk of relapse.


Systemic lupus erythematosus (SLE) is a chronic autoimmune disease of incompletely known aetiology characterised by the presence of anti-nuclear autoantibodies and inflammation in a wide spectrum of organs.1 Despite the improvement of SLE prognosis in the last decades, SLE patients are still at high risk of disease-related complications and premature death.2 3 Several authors have shown that remission defined as the elimination of disease activity predicted a better disease outcome with a lower burden of damage and a lower risk of relapse.4 5 Thus, remission achievement and its maintenance have become central in the management of SLE patients.6 7 Many definitions have been used to better characterise remission states, but a single one has recently made consensus, the DORIS (Definitions of Remission in SLE). For defining remission, the DORIS uses a clinical index, such as the clinical SLE disease activity index (SLEDAI)=0 with routine laboratory assessments including anti-double stranded DNA (anti-dsDNA) antibodies (Abs) and complement.7 Serologically active patients, that is, patients with an increase in serum anti-dsDNA Abs and/or a complement consumption, are much more likely to experience subsequent flares, as compared with those who are serologically inactive.7 However, these two markers fail to reliably predict a lupus flare.8 9 Consequently, no consensus has yet been reached by the DORIS working group as to a reliable definition that enables the distinction of patients who are serologically active from those who are serologically inactive.7

Many authors consider the dysregulation of interferons (IFNs), especially interferon-alpha (IFN-α), to be a central cause of the immunological abnormalities observed in SLE.10–16 Transcriptome analysis using microarray technology revealed upregulation of the expression of numerous IFN-stimulated genes (ISGs) in SLE patients’ peripheral blood mononuclear cells, constituting an overall ‘IFN signature’.17 18 Many reports showed patients’ elevated serum-IFN-α levels to be associated with SLE activity and severity, suggesting that monitoring this cytokine might help physicians to better evaluate disease activity,17–35 although, unexpectedly, some clinically inactive patients have persistently elevated serum IFN-α levels.19–23 25 35 Since serum concentrations of IFN-α are usually very low and often not detectable by classic immunoassays, monitoring the expression of ISGs is used to evaluate IFN serological activity. The resulting ‘IFN scores’ are calculated based on the expression of several representative ISGs.29 36 37 However, the low availability and high complexity of transcriptome-microarray technology implies that IFN scores are not standardised and therefore cannot be easily used in clinical practice. At present, IFN-α overexpression has been poorly studied in patients in remission.

The single-molecule array (Simoa) assay, or digital immunoassay, is an ultrasensitive assay based on enumeration of individual enzyme-labelled immunocomplexes of proteins captured on beads in single-molecule arrays. It enables direct IFN-α quantification at attomolar concentrations,35 38–40 corresponding to a 5000-fold—increased sensitivity over classic ELISAs. We and others have shown that, at physiological concentrations, the digital immunoassay is as sensitive as ISG expression as a means to quantify IFN-α levels and simpler to perform and standardise.35 40 We thus conducted a study to determine the magnitude of serum IFN-α concentrations in SLE patients in remission using the DORIS criteria.7 Additionally, in order to improve the definition of remission, we determined the clinical and biological features associated with elevated IFN-α levels in patients experiencing a remission and we investigated whether high serum IFN-α levels at baseline in patients in remission were predictive of a flare in the ensuing year.

Patients, materials and methods

Study design and patients

The longitudinal study reported here was conducted between September 2014 and September 2017 at the National Referral Center for SLE, Paris, France. Serum samples were obtained at day 0 (=baseline) from consecutive patients diagnosed with SLE according to the 1997 American College of Rheumatology criteria for SLE classification, regardless of the activity of the disease.41 Exclusion criteria were: (1) known or suspected infection or malignancy on the day blood was drawn; (2) increased hydroxychloroquine (HCQ), prednisone and/or immunosupressant (IS) during the 4 weeks preceding day 0. SLE clinical characteristics (see online supplementary file), SELENA-SLEDAI,42–44 SLEDAI-2K,45 class of lupus nephritis according to ISN/RPS-2003,46 and therapeutic regimen were recorded on day 0. Lupus flares were defined according to the SELENA-SLEDAI Flare Index.43 44 The term ‘clinical’ SLEDAI (cSLEDAI) refers to symptoms, signs and routine laboratory testing and disregards only the points that can be given for the presence of anti-dsDNA Abs and/or low complement.7 Five exclusive disease activity statuses were defined, according to the DORIS7 and following Wilhelm et al 47 and Ugarte-Gil et al 48 without physician global assessment (PGA) and serum C4 analysis:

  • Complete remission off treatment: cSELENA-SLEDAI=0, no corticosteroids, no IS, no anti-dsDNA Abs and no complement component 3 (C3) decrease.

  • Clinical remission off treatment: cSELENA-SLEDAI=0, no corticosteroids, no IS. Anti-dsDNA Abs and/or C3 decrease present.

  • Complete remission on treatment: cSELENA-SLEDAI=0, prednisone 1–5 mg/day, IS allowed, no anti-dsDNA Abs and no C3 decrease

  • Clinical remission on treatment: cSELENA-SLEDAI=0, prednisone 1–5 mg/day, IS allowed. Anti-dsDNA Abs and/or C3 decrease present.

  • Not in remission: cSELENA-SLEDAI >0 and/or prednisone >5 mg/day.

HCQ was allowed in all groups. Duration of remission was recorded. Over a consecutive 5-year period of remission, the patient was considered in ‘prolonged remission’, as proposed by Steiman et al,49 and treated in the statistical analysis as a 5-year remission. Alternatively, patients not in remission but fulfilling Lupus Low Disease Activity State (LLDAS)48 50 were included in the ‘LLDAS without remission’ subgroup (see online supplementary file). Sera from age-matched and gender-matched healthy donors (n=68) were collected (Établissement Français du Sang, Île-de-France, Pitié–Salpêtrière Hospital, Paris, France) during the same time period.

Patient follow-up

Patients in remission at day 0 were followed for 1 year (see online supplementary file).

IFN-α digital immunoassay

Serum IFN-α concentrations, expressed in fg/mL, were determined at day 0 with digital immunoassay technology (IFN-α Reagent Kit, Quanterix Simoa, Lexington, Massachusetts, USA), based on a three-step protocol (see online supplementary file) using the HD-1 Analyzer (Quanterix).38 The IFN-α digital ELISA positivity threshold (defining elevated IFN-α) was 136 fg/mL, which is three SD above the mean serum IFN-α concentration calculated from the sera samples from the 68 healthy blood donors.

Statistical analysis

Statistical analyses were performed using GraphPad Prism, V.5.0 software (GraphPad Software, San Diego, California, USA) and SAS V.9.4 software (see online supplementary file).


Patient characteristics and distribution according to disease activity

A total of 407 patients were included in the study, with 254 patients in remission (ie, cSELENA-SLEDAI=0 and prednisone equal or less than 5 mg/day) and 153 not in remission. Patients’ baseline characteristics are described in table 1. From the 254 patients in remission, 86 (33.9%) were in complete remission off treatment, 59 (23.2%) in complete remission on treatment, 47 (18.5%) in clinical remission off treatment and 62 (24.4%) in clinical remission on treatment. Sixty-seven (26.3%) patients were in remission for less than 1 year, 101 (39.8%) for 1–5 years, and 86 (33.9%) for more than 5 years. The median (range) of remission duration was 5 years (0–5) for patients in complete remission off treatment, 1.6 years (0–5) for patients in complete remission on treatment, 4.2 years (0.1–5) for clinical remission off treatment and 1.6 years (0–5) for clinical remission on treatment. Among the 153 patients not in remission, 19 (12.4%) fulfilled the LLDAS criteria, defining the ‘LLDAS without remission’ subgroup.

Table 1

Disease characteristics in SLE patients at baseline

Serum IFN-α in remission and LLDAS

The serum IFN-α levels according to the different states of remission are presented in table 2. A total of 26.0% of patients in remission had still detectable serum IFN-α levels exceeding the positivity threshold. Patients in ‘clinical remission’ had the highest rate of elevated serum IFN-α: 41.9% for ‘clinical remission on treatment’ and 38.3% for ‘clinical remission off treatment’ (vs 16.3% for ‘complete remission off treatment’, p=0.0007 and 0.006, respectively). Among patients in remission, patients in ‘clinical remission on treatment’ had the highest concentration of IFN-α with a median (quartiles) of 109 fg/mL (12–378) (vs 11 fg/mL (0–81) in patients in ‘complete remission off treatment’, p=0.0002). Alternatively, we assessed serum IFN-α levels in different states of remission according to the definition of Zen et al,5 a definition of remission in which the role of corticosteroids is highlighted in comparison to immunosuppressive treatment. As shown in online supplementary file 1, we found similar results: patients in ‘clinical remission on or off corticosteroids’ had the highest rate of abnormal serum IFN-α levels, as compared with those of patients in ‘complete remission’. Thus, a significant number of patients in remission had increased IFN-α levels in their serum. Eight (42.1%) patients of the ‘LLDAS without remission’ subgroup displayed abnormal IFN-α serum concentrations (OR 3.7 (95% CI 1.3 to 11.0) and a mean IFN-α level of 111 fg/mL (0–1647), p=0.02 and p=0.07, respectively, as compared with patients in ‘complete remission off treatment’).

Table 2

Serum IFN-α levels at baseline in SLE patients in remission or not in remission (according to Wilhelm et al 47 modified)

SLE characteristics associated with abnormal serum IFN-α levels in remission

Elevated IFN-α serum levels of patients in remission were significantly associated, in multivariable analysis, with the presence of serum Abs specific for RNP, dsDNA and Ro/SSA60, young age and lower granulocyte counts (table 3). In contrast, disease and remission duration, prednisone intake, IS therapy, HCQ and low C3 serum levels did not show a significant association with IFN-α serum levels.

Table 3

Disease characteristics associated with serum IFN-α levels at baseline in SLE patients in remission

Elevated IFN-α serum levels in SLE patients in remission predict a lupus flare

Of the 254 patients in remission at day 0, 250 were followed for 1 year. Twenty-four (9.6%) patients experienced a flare. The median (range) time of the flare occurrence was 141 (25–349) days. The type and severity of the flares are reported in online supplementary table 3. A total of 37.5% of the flares were severe. The most frequent type of relapses was arthritis (n=15) followed by cutaneous flare (n=7) and serositis (n=4). Unadjusted cox regression showed a significantly higher risk of relapse in patients who displayed at baseline elevated IFN-α (HR 5.5 (95% CI 2.4 to 12.5), p<0.0001) or decreased C3 (HR 3.7 (95% CI 1.6 to 9.1), p=0.003) serum levels, respectively, but not in patients who had a positive Farr assay (HR 1.5 (95% CI 0.7 to 3.4), p=0.3) (figure 1). The highest concentrations of IFN-α at baseline were associated with the greatest frequencies of relapse (online supplementary figure 1). Other factors at baseline associated with the risk of relapse were prednisone intake 1–5 mg/day (HR 3.2 (95% CI 1.4 to 7.5), p=0.007), positive anti-RNP Abs (HR 3.1 (1.4 to 7.0), p=0.006), age <40 years (HR 0.4 (95% CI 0.2 to 0.9), p=0.02) and disease duration <10 years (HR 0.2 (95% CI 0.1 to 0.6), p=0.003). The remission duration was a protective factor for the probability of flare (HR 0.6 (95% CI 0.5 to 0.8) for each consecutive year of remission completed, p=0.0002).

Figure 1

Baseline elevated serum interferon-alpha (IFN-α) and low C3 levels identify patients with elevated risk for future systemic lupus erythematosus flares. Kaplan-Meier curves. Serum IFN-α (positivity threshold of 136 fg/mL), anti-double stranded DNA (anti-dsDNA) antibodies by Farr assay (cut-off value: 9.0 IU/mL) and C3 levels (cut-off value: 0.78 g/L) were assessed at day 0. Kaplan-Meier plots show the percentage of patients who flared in any organ system. Vertical tick marks along each curve represent patients who remained flare-free but did not have a full year of clinical follow-up (censored data). Curves were compared using Log-Rank tests. Crude HRs were calculated using proportional risk cox model. (A) The red dashed line represents the 64 patients with elevated IFN-α serum level at day 0. The continuous blue line represents the 186 patients with negative IFN-α serum level at day 0. (B) The red dashed line represents the 99 patients with positive Farr assay at day 0. The continuous blue line represents the 151 patients with negative Farr assay at day 0. (C) The red dashed line represents the 28 patients with low C3 level at day 0. The continuous blue line represents the 216 patients with normal C3 level at day 0.

Finally, in multivariable analysis, the factors independently associated with the risk of flare were abnormal serum IFN-α levels at baseline (HR 4.0 (95% CI 1.7 to 9.6), p=0.002) and remission duration (HR 0.7 (95% CI 0.5 to 0.9), p=0.02, for each year in remission). Low C3 (HR 2.4 (95% CI 0.9 to 6.2), p=0.07) and prednisone intake (HR 2.4 (95% CI 0.9 to 5.9), p=0.06) were also kept in the model as associated with a higher risk of flare but these associations were not statistically significant. We performed a sensitivity analysis using a lower concentration of IFN-α as the threshold of elevated IFN-α levels and found similar results (data not shown).

Additionally, the risks of relapse according to combined SLE biomarkers (low C3, positive Farr assay and elevated IFN-α serum levels) were independently analysed with a proportional risk Cox model (online supplementary table 4). In this model, isolated elevated IFN-α levels were a predictive factor of lupus flare (HR of 5.5 (95% CI 1.7 to 18.1), p=0.005). In contrast, isolated positive Farr assay and isolated low C3 levels had no predictive value of lupus flare.

Potentiality of the digital immunoassay, C3 and Farr assay performances to predict a flare

The time-dependent receiver operating characteristics (ROC) area under the curve (AUC) (figure 2) estimates for the IFN-α digital immunoassay to predict a flare was 0.73, better than that of anti-dsDNA Abs (0.60, p=0.055), and C3 serum levels (0.56, p=0.2). The time-dependent ROC AUCs for the three biomarkers were highest around day 90.

Figure 2

Cumulative time-dependent receiver operating characteristics (ROC) curves. Serum interferon-alpha (IFN-α) digital immunoassay, anti-double stranded DNA (anti-dsDNA) antibodies by Farr assay and C3 levels were assessed at baseline (day 0). Systemic lupus erythematosus (SLE) patients in remission were followed for 1 year. Lupus flares were defined using the Safety of Estrogens in Lupus Erythematosus: National Assessment version of the Systemic Lupus Erythematosus Disease Activity Index Flare Index. Patients who had a flare in any organ system were recorded. The diagnostic performances of the baseline serum-IFN-α digital immunoassay (red line), Farr assay (blue line) and C3 (green line) to predict SLE flare were investigated by computing cumulative time-dependent ROC curves, with lupus flares serving as the gold standard for those analyses. The dashed line represents the upper and the lower 95% CI. The cumulative time-dependent ROC area under the curve estimated to predict a flare are given and were compared based on Uno et al. 71


The significance of monitoring IFN-α serum levels in SLE for the assessment of remission and the risk to develop flares has been poorly documented as yet. In the present study, we show that a significant proportion of patients despite being in remission have high serum IFN-α levels that were found to independently predict the risk of subsequent SLE flares, thereby emphasising the interest of measuring the production of this cytokine to monitor the course of disease.

Earlier publications have reported that between 3% and 32% of SLE patients without active disease present elevated IFN-α serum levels depending on the type of assay used to determine the presence of this cytokine.19–23 25 35 However, because different definitions of disease inactivity were used in these studies, the results cannot directly be compared with our study that takes into account the recently formulated consensual definition of disease remission.5 7 47 Contrary to the results of a previously published report,51 we noted that serum concentrations of IFN-α above the positive threshold values varied between different remission subgroups. Patients in clinical remission, that is, with serological activity, presented more frequently elevated serum levels of IFN-α than patients in complete remission. The presence of serum Abs specific for RNP, dsDNA and Ro/SSA60 were found to be independently associated with the magnitude of serum IFN-α levels. These associations have previously been shown in SLE29–31 52–57 but, to the best of our knowledge, never been reported for patients in remission. Our data are in accordance with those from previous studies that showed a key role of DNA/RNA-associated immune complexes through the activation of TLR7 and TLR9 for the induction of type I IFN production.58 59

The DORIS group has recently agreed that the subsequent occurrence of flares is among the most appropriate outcome variables for defining the prognostic value of remission and furthermore suggested the inclusion of serological criteria for this definition.7 Results from previous reports on smaller SLE patient cohorts, based on the analysis of expression of ISGs in peripheral blood cells by microarray, failed to demonstrate an association between IFN scores and the longitudinal risk of relapse.30 31 It is to be noted however that type I and type II IFNs largely overlap in the genes that they control, making it difficult to distinguish the signatures of IFN-gamma (IFN-γ) from IFN-α and IFN-beta (IFN-β). Indeed, the results from a modular repertoire analysis have emphasised that IFN signatures in SLE are not restricted to IFN-α, but also involve IFN-β and IFN-γ, thus underscoring the non-specific nature of IFN scores.60 Therefore, in a disease like SLE in which IFN-α appears to be central in the pathogenesis, it is likely that clinical activity better correlates with the values of IFN-α serum levels, directly measured by a highly specific digital ELISA, than with the less specific expression of IGSs. Moreover, the IFN scores are likely to be rather poorly sensitive to changes in IFN-α serum concentrations and therefore of little use in monitoring the risk of relapse of disease. The results of our study suggest that the measurement of additional interferons, including IFN-β, IFN-γ and IFN-λ with the ultrasensitive digital immunoassay, might help to determine whether the expression of other IFN members also correlates with disease activity or specific characteristics of the disease.34 61 62 Nonetheless, our results corroborate those from an earlier study showing the usefulness of monitoring expression levels of certain IFN-regulated chemokines to predict future flares: serum levels of CXCL10, CCL2 and CCL19 chemokines were found to be linked with SLE activity and performed better than other laboratory tests to predict a flare over the following year.63 Yet, these chemokines can be induced by other SLE-associated cytokines beyond IFN-α, such as IFN-γ and their monitoring remains difficult in routine practice. As reported by others, we did not find an association between the presence of anti-dsDNA Abs and the risk of flare. Not surprisingly, the duration of remission was significantly associated with a poor risk of flare in the following year, thus identifying a subgroup of patients in prolonged remission with less risk of relapse and for whom clinical monitoring can probably be lightened.

The sustained presence of IFN-α in serum of SLE patients in remission may have pathological consequences by itself. IFN-α overexpression could be an explanation of chronic fatigue, depression and reduction of sleep secondary to the stimulation of the dopamine metabolism by IFN-α in the central neurological system.64–66 IFN-α plays also a prominent role in endothelial cell damage and up-regulation of the expression of scavenger receptors in monocyte and macrophages, leading to increased lipid uptake and foam cell formation, a process that may be at the origin of the accelerated atherosclerosis observed in SLE.67–69 These data suggest that for patients in clinical remission, return of serum IFN-α to normal values could become one of the objectives of the treatment.

Our study has some limitations. It unfortunately lacks data on patients’ ethnicity, which is known to bias serum IFN-α levels.55 We also used a definition of remission and LLDAS in the absence of PGA that has not been assessed in our cohort. However, other teams have already adapted the definition of remission and LLDAS arguing that in the SLE Response Index, the SLEDAI is the variable with the highest impact on the definition of response and that, even without PGA, a modified definition of remission or LLDAS is still entirely valid.48 70 Finally, serum IFN-α concentrations were not assessed during the longitudinal part of the study and therefore information with respect to intra-individual variability of serum IFN-α levels over time is lacking. This aspect will be subject to further studies.

In conclusion, our data confirm that a large number of SLE patients in remission display elevated serum IFN-α concentrations, especially in the presence of anti-dsDNA and anti-ribonucleoprotein Abs (ie, anti-Ro/SSA 60, anti-RNP), as well as young age. This overexpression is an independent predictive biomarker of lupus flare in the following year. Including serum IFN-α measurements in the routine laboratory assessments in patients in remission could help clinicians to identify a subgroup of SLE patients clinically in remission but who still overexpress IFN-α and are at higher risk of relapse. These data suggest that the return of normalcy of serum IFN-α could become one of the objectives of the treatment. Our results are to be validated in other independent cohorts.


We thank the patients, the healthy donors, the nurses and the Department of Internal Medicine two staff who participated in this study. We thank Professor Jean-Christophe Lega for its advices.



  • Handling editor Josef S Smolen

  • AM, SM-M and KD contributed equally.

  • GG and ZA contributed equally.

  • Correction notice This article has been corrected since it published Online First. Figure one has been corrrected.

  • Contributors SM-M, AM, HD, KD, HY, GG, FR and ZA contributed to the conception and design of the study; SM-M, AM, KD, FC-A, JH, MH, MM, MP, FR and ZA were involved in the acquisition of data; SM-M, AM, HD, KD, HY, MPdC, GG, FR and ZA contributed to the analysis and interpretation of data. All authors contributed to drafting and/or revising the manuscript.

  • Funding Assistance Publique–Hôpitaux de Paris, Institut National de la Santé et de la Recherche Médicale (Inserm), Sorbonne Université, the French Arthritis Foundation, the Société Nationale Française de Médecine Interne and The Fondation pour la Recherche Médicale.

  • Competing interests None declared.

  • Patient consent for publication Informed consent was obtained from all the participants.

  • Ethics approval The local Ethics Committee of the Pitié-Salpêtrière Hospital approved this study.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement Data are available upon reasonable request.