Applying the 2019 EULAR/ACR lupus criteria to patients from an established cohort: a Latin American perspective

Novel multiclass classification machine learning approach for the early-stage classification of systemic autoimmune rheumatic diseases

OBJECTIVE

Systemic autoimmune rheumatic diseases (SARDs) encompass a diverse group of complex conditions with overlapping clinical features, making accurate diagnosis challenging. This study aims to develop a multiclass machine learning (ML) model for early-stage SARDs classification using accessible laboratory indicators.

METHODS

A total of 925 SARDs patients were included, categorised into SLE, Sjögren's syndrome (SS) and inflammatory myositis (IM). Clinical characteristics and laboratory markers were collected and nine key indicators, including anti-dsDNA, anti-SS-A60, anti-Sm/nRNP, antichromatin, anti-dsDNA (indirect immunofluorescence assay), haemoglobin (Hb), platelet, neutrophil percentage and cytoplasmic patterns (AC-19, AC-20), were selected for model building. Various ML algorithms were used to construct a tripartite classification ML model.

RESULTS

Patients were divided into two cohorts, cohort 1 was used to construct a tripartite classification model. Among models assessed, the random forest (RF) model demonstrated superior performance in distinguishing SLE, IM and SS (with area under curve=0.953, 0.903 and 0.836; accuracy= 0.892, 0.869 and 0.857; sensitivity= 0.890, 0.868 and 0.795; specificity= 0.910, 0.836 and 0.748; positive predictive value=0.922, 0.727 and 0.663; and negative predictive value= 0.854, 0.915 and 0.879). The RF model excelled in classifying SLE (precision=0.930, recall=0.985, F1 score=0.957). For IM and SS, RF model outcomes were (precision=0.793, 0.950; recall=0.920, 0.679; F1 score=0.852, 0.792). Cohort 2 served as an external validation set, achieving an overall accuracy of 87.3%. Individual classification performances for SLE, SS and IM were excellent, with precision, recall and F1 scores specified. SHAP analysis highlighted significant contributions from antibody profiles.

CONCLUSION

This pioneering multiclass ML model, using basic laboratory indicators, enhances clinical feasibility and demonstrates promising potential for SARDs classification. The collaboration of clinical expertise and ML offers a nuanced approach to SARDs classification, with potential for enhanced patient care.

Reframing health disparities in SLE: A critical reassessment of racial and ethnic differences in lupus disease outcomes

Health disparities in the prevalence and outcomes of systemic lupus erythematosus (SLE) are well documented across racial and ethnic groups. Similar to other chronic diseases, differences in disease severity among individuals with SLE are likely influenced by both genetic predisposition and multiple social determinants of health. However, research in SLE that jointly examines the genetic and environmental contributions to the disease course is limited, resulting in an incomplete understanding of the biologic and social mechanisms that underly health disparities. While research on health disparities can reveal inequalities and inform resource allocation to improve outcomes, research that relies on racial and ethnic categories to describe diverse groups of people can pose challenges. Additionally, results from research comparing outcomes across socially constructed groups without considering other contributing factors can be misleading. We herein comprehensively examine existing literature on health disparities in SLE, including both clinical studies that examine the relationship between self-reported race and ethnicity and disease outcomes and studies that explore the relationships between genomics and lupus outcomes. Having surveyed this body of research, we propose a framework for research examining health disparities in SLE, including ways to mitigate bias in future studies.

Clinical patterns of disease: From early systemic lupus erythematosus to late-onset disease

Systemic lupus erythematosus (SLE) is a complex disease with an insidious clinical presentation. In up to half of the cases, SLE onset is characterized by clinical and serological manifestations that, although specific, are insufficient to fulfill the classification criteria. This condition, called incomplete SLE, could be as challenging as the definite and classifiable SLE and requires to be treated according to the severity of clinical manifestations. In addition, an early SLE diagnosis and therapeutic intervention can positively influence the disease outcome, including remission rate and damage accrual. After diagnosis, the disease course is relapsing-remitting for most patients. Time in remission and cumulative glucocorticoid exposure are the most important factors for prognosis. Therefore, timely identification of SLE clinical patterns may help tailor the therapeutic intervention to the disease course. Late-onset SLE is rare but more often associated with delayed diagnosis and a higher incidence of comorbidities, including Sjogren's syndrome. This review focuses on the SLE disease course, providing actionable strategies for early diagnosis, an overview of the possible clinical patterns of SLE, and the clinical variation associated with the different age-at-onset SLE groups.

Complement as a Biomarker for Systemic Lupus Erythematosus

Systemic lupus erythematosus (SLE) is a disease of immune complex deposition; therefore, complement plays a vital role in the pathogenesis of SLE. In general, complement levels in blood and complement deposition in histological tests are used for the management of SLE. Thus, the evaluation of complement status can be useful in the diagnosis of SLE, assessment of disease activity, and prediction of treatment response and prognosis. In addition, novel complement biomarkers, such as split products and cell-bound complement activation products, are considered to be more sensitive than traditional complement markers, such as serum C3 and C4 levels and total complement activity (CH50), which become more widely used. In this review, we report the complement testing in the management of SLE over the last decade and summarize their utility.

Performances of different classification criteria for systemic lupus erythematosus in a single-center cohort from Turkey

Objective

Sensitivity and specificity of SLE classification criteria may vary in different populations and clinical settings. In this study, we aimed to compare the performances of three criteria sets/rules (1997, 2012, and 2019) in a large cohort of patients and relevant diseased controls.

Methods

The medical records of consecutive SLE patients and diseased controls were reviewed for clinical and laboratory features relevant to all sets of criteria. Criteria sets/rules were analyzed based on sensitivity, positive predictive value, specificity, and negative predictive value, using clinical diagnosis with at least 6 months of follow-up as the gold standard. A subgroup analysis was performed in ANA positive patients.

Results

A total of 393 SLE patients and 308 non-SLE diseased controls were included. Sensitivity was 78.4% for 1997 criteria and was more than 90% for both 2012 (91.9%) and 2019 (94.4%) criteria. Specificity was the highest (95.1%) for 1997 ACR criteria, 91.5% for 2012 SLICC criteria and 91.2% for 2019 EULAR/ACR criteria. When only ANA positive patients were analyzed, sensitivity of each criteria increased by 1%, 0.8%, and 2.2%, respectively. Specificity of 1997 criteria decreased by 2% and specificity of 2012 and 2019 criteria both decreased to less than 90%.

Conclusion

EULAR/ACR criteria were more sensitive than 1997 criteria and had a comparable performance with SLICC criteria. When only ANA positive patients were analyzed, the presence of false positive results (originated from patients with Sjögren’s disease and antiphospholipid syndrome mainly) decreased the specificity of both SLICC and EULAR/ACR criteria.

Diagnostic accuracy of the American College of Rheumatology-1997, the Systemic Lupus International Collaborating Clinics-2012, and the European League Against Rheumatism-2019 criteria for juvenile systemic lupus erythematosus: A systematic review and network meta-analysis

IMPORTANCE/BACKGROUND

Several diagnostic criteria have been developed to effectively diagnose systemic lupus erythematosus (SLE). Three criteria are most common, namely the American College of Rheumatology (ACR)-1997, the Systemic Lupus International Collaborating Clinics (SLICC)-2012, and the European League Against Rheumatism (EULAR/ACR)-2019. Whether they also apply to juvenile SLE is unclear.

OBJECTIVE

To examine the diagnostic accuracy of ACR-1997, SLICC-2012, and EULAR/ACR-2019 for juvenile SLE.

DATA SOURCES

A comprehensive search of PubMed, Cochrane, and Embase was conducted up to 26 March 2022.

STUDY SELECTION

We included all study designs in which patients had any index tests for ACR-1997, SLICC-2012, or EULAR/ACR-2019; both full-text papers and conference abstracts published in English were used. Exclusion criteria were as follows: (1) case reports; (2) adult subjects; or (3) did not report sufficient information to acquire true positive, false positive, true negative, and false negative values of diagnostic criteria.

DATA EXTRACTION AND SYNTHESIS

Two authors independently screened studies, extracted relevant data, and assessed the risk of bias.

MAIN OUTCOMES AND MEASURES

First, a meta-analysis of the diagnostic accuracy of EULAR/ACR-2019 and a hierarchical summary receiver operating characteristic (HSROC) model was performed to estimate sensitivity and specificity with 95% confidence intervals (CIs). We then carried out a network meta-analysis to compare the performances of these three diagnostic criteria.

RESULTS

In total, 17 relevant studies that included 2339 juvenile SLE patients were eligible to analyze pooled accuracy. In the meta-analysis, 10 studies (1613 cases) reported the diagnostic performance of EULAR/ACR-2019, showing a pooled sensitivity of 0.92 (95% CI, 0.89-0.95), pooled specificity of 0.89 (0.77-0.95), and area under HSROC of 0.96 (0.94-0.97). In the network meta-analysis, the SLICC-2012 (0.94, 0.92-0.96) had the highest sensitivity, followed by EULAR/ACR-2019 (0.93, 0.90-0.95), and ACR-1997 (0.78, 0.72-0.82); the ACR-1997 (0.96, 0.92-0.98) demonstrated the highest specificity. EULAR/ACR-2019 (0.92, 0.87-0.96) and SLICC-2012 (0.92, 0.86-0.96) had the similar specificity.

CONCLUSIONS AND RELEVANCE

We found that the applicability of the new EULAR/ACR-2019 criteria in juvenile SLE is not yet the best diagnostic tool.

TRIAL REGISTRATION

PROSPERO CRD42022321514.

Advances in SLE classification criteria

This year, the American College of Rheumatology (ACR) 1982 classification criteria for systemic lupus erythematosus (SLE) celebrate their 40th anniversary. From this start, the quest for optimal SLE criteria has led to the 1997 ACR update, the 2012 publication of the Systemic Lupus International Collaborating Clinics (SLICC) criteria, and, in 2019, the European League Against Rheumatism (EULAR)/ACR classification criteria. The latter have since been externally validated in more than two dozen studies and have become the gold standard inclusion criterion of SLE clinical trials. This comprehensive review attempts to follow the evolving success story of SLE classification, highlighting relevant decisions and their rationale, and discussing consequences for the way SLE is defined and managed.

Population-based incidence and time to classification of systemic lupus erythematosus by three different classification criteria: a Lupus Midwest Network (LUMEN) study

OBJECTIVE

To estimate the incidence and time-to-classification of SLE by the 1997 ACR (ACR97) criteria, the SLICC criteria, and the European Alliance of Associations for Rheumatology/ACR (EULAR/ACR) criteria.

METHODS

We identified all incident SLE cases from 2000-2018 in the well-defined Olmsted County population. Clinical data included in the ACR97, SLICC and EULAR/ACR criteria were manually abstracted from medical records. All incident cases met at least one of the three classification criteria. Time-to-classification was estimated from the first documented lupus-attributable disease manifestation to the time of criteria fulfilment by each of the three definitions. Annual incidence rates were age or age/sex adjusted to the 2000 US population.

RESULTS

Of 139 incident cases there were 126 cases by the EULAR/ACR criteria, corresponding to an age/sex-adjusted incidence of 4.5 per 100 000 population (95% CI: 3.7, 5.2). The age/sex-incidence was higher than that of the SLICC criteria (113 cases; 4.0 per 100 000 [95% CI: 3.3, 4.7], P = 0.020) and the ACR97 (92 cases; 3.3 per 100 000 [95% CI: 2.6, 3.9], P < 0.001). The median time from first disease manifestation to criteria fulfilment was shorter for the EULAR/ACR criteria (29.4 months) than the ACR97 criteria (47.0 months, P < 0.001) and similar to the SLICC criteria (30.6 months, P = 0.83).

CONCLUSION

The incidence of SLE was higher by the EULAR/ACR criteria compared with the ACR97 and the SLICC criteria, and the EULAR/ACR criteria classified patients earlier that the ACR97 criteria but similar to the SLICC criteria.

Validation of the 2019 EULAR/ACR classification criteria for systemic lupus erythematosus in ANA-positive Chinese patients

Background

The aim of this study was to validate the 2019 European League Against Rheumatism/American College of Rheumatology (EULAR/ACR) classification criteria for systemic lupus erythematosus (SLE) in antinuclear antibody (ANA)-positive Chinese patients.

Methods

Medical records of all adult patients who attended the rheumatology out-patient clinics between May and September 2019 were reviewed. Patients with ever ANA positive (titre ⩾1:80) were included and evaluated for the fulfilment of the 2019 EULAR/ACR, 2012 Systemic Lupus International Collaborating Clinics (SLICC) and 1997 ACR criteria for SLE classification. The performance of these criteria in predicting a clinical diagnosis of SLE as judged by an independent panel of rheumatologists was studied and compared in different subgroups.

Results

A total of 1533 patients (88.2% women; age at first clinic attendance 45.5 ± 15.6 years) were studied and 562 patients were judged to be clinical SLE. The sensitivity and specificity of the EULAR/ACR (⩾10 points), SLICC and ACR criteria for a clinical diagnosis of SLE was 96.1%, 97.9% and 86.1%; and 85.8%, 86.3% and 94.3%, respectively. Applying the attribution rule to the non-SLE controls, the specificity of the three criteria increased to 95.0%, 92.5% and 98.8%, respectively. The specificity of the EULAR/ACR criteria was higher in male patients (97.9%), those aged >50 years (97.0%) and disease duration of ⩽3 years (97.6%). Using a cut-off of 12 points, the specificity of the EULAR/ACR criteria was further increased (96.6%) while a high sensitivity (95.0%) was maintained.

Conclusion

In Chinese patients with a positive ANA, the EULAR/ACR criteria for clinical SLE perform equally well to the SLICC criteria. Both the EULAR/ACR and SLICC are more sensitive but less specific than the ACR criteria. The specificity of all the three criteria is enhanced by applying the attribution rule to controls. The specificity of the EULAR/ACR criteria is higher in certain patient subgroups or when the cut-off score is raised.

Multicenter lupus register from Argentina, the RELESSAR database: Influence of ethnicity on disease phenotype

OBJECTIVE

The objective is to describe the main characteristics of patients with systemic lupus erythematosus (SLE) in Argentina and to examine the influence of ethnicity on the expression of the disease.

PATIENTS AND METHODS

RELESSAR is a multicentre register carried out by 106 researchers from 67 rheumatologic Argentine centres. It is a cross-sectional study of SLE (1982/1997 ACR) patients. RELESSAR electronic database includes demographic, cumulative SLE manifestations, SELENA-SLEDAI, SLICC-SDI, Katz's severity and Charlson's comorbidity indexes and treatment patterns.

RESULTS

We included 1,610 patients, 91.7% were female with a median age at diagnosis of 28.1 ± 12.8; 96.2% met ≥4 ACR 1982/97 criteria. Frequent manifestations were arthritis (83.5%), malar rash (79.5%), photosensitivity (75.3%), haematological (63.8%) and renal disease (47.4%), antinuclear antibodies (96%), anti-dsDNA (66.5%) and anti-Smith antibodies (29%). The mean Selena-SLEDAI score at last visit was 3.18 (SD 4.3) and mean SDI was 1 (SD 1.3). The accumulated treatments most frequently used were antimalarials (90.4%), corticosteroids (90%), azathioprine (31.8%), intravenous cyclophosphamide (30.2%), mycophenolate mofetil or mycophenolic acid (24.5%), methotrexate (19.3%), belimumab 5.3% and rituximab 5.1%. Refractory lupus was diagnosed in 9.3% of the cases. The main causes of death were lupus activity (25.0%), activity and concomitant infections (25.0%), infections (18.2%), vascular disease (13.6%) and cancer (4.5%). Mortality was associated with higher SLEDAI, Katz, damage indexes and comorbidities. Of the 1610 patients included, 44.6% were Caucasian, 44.5% Mestizo, 8.1% Amerindian and 1.2% Afro-Latin American. Mestizo patients had higher male representation, low socioeconomic status, more inadequate medical coverage, fewer formal years of education and shorter disease duration. Polyadenopathies and Raynaud's phenomenon were more frequent in Caucasians. In the logistic regression analysis higher damage index (OR 1.28, CI 95% 1.02-1.61, p = 0.03) remained associated to mestizo ethnicity.

CONCLUSIONS

This study represents the largest number of adult patients with SLE studied in Argentina. Caucasian patients were differentiated by having Raynaud's phenomenon and polyadenopathy more frequently, while patients of Mestizo origin had higher damage indexes.

A glimpse into the future of systemic lupus erythematosus

This viewpoint article on a forecast of clinically meaningful changes in the management of systemic lupus erythematosus (SLE) in the next 10 years is based on a review of the current state of the art. The groundwork has been laid by a robust series of classification criteria and treatment recommendations that have all been published since 2019. Building on this strong foundation, SLE management predictably will take significant steps forward. Assessment for lupus arthritis will presumably include musculoskeletal sonography. Large-scale polyomics studies are likely to unravel more of the central immune mechanisms of the disease. Biomarkers predictive of therapeutic success may enter the field; the type I interferon signature, as a companion for use of anifrolumab, an antibody against the common type I interferon receptor, is one serious candidate. Besides anifrolumab for nonrenal SLE and the new calcineurin inhibitor voclosporin in lupus nephritis, both of which are already approved in the United States and likely to become available in the European Union in 2022, several other approaches are in advanced clinical trials. These include advanced B cell depletion, inhibition of costimulation via CD40 and CD40 ligand (CD40L), and Janus kinase 1 (Jak1) and Tyrosine kinase 2 (Tyk2) inhibition. At the same time, essentially all of our conventional therapeutic armamentarium will continue to be used. The ability of patients to have successful SLE pregnancies, which has become much better in the last decades, should further improve, with approaches including tumor necrosis factor blockade and self-monitoring of fetal heart rates. While we hope that the COVID-19 pandemic will soon be controlled, it has highlighted the risk of severe viral infections in SLE, with increased risk tied to certain therapies. Although there are some data that a cure might be achievable, this likely will remain a challenge beyond 10 years from now.

Systemic lupus erythematosus as a genetic disease

Systemic lupus erythematosus is the prototypical systemic autoimmune disease, as it is characterized both by protean multi-organ system manifestations and by the uniform presence of pathogenic autoantibodies directed against components of the nucleus. Prior to the modern genetic era, the diverse clinical manifestations of SLE suggested to many that SLE patients were unlikely to share a common genetic risk basis. However, modern genetic studies have revealed that SLE usually arises when an environmental exposure occurs in an individual with a collection of genetic risk variants passing a liability threshold. Here, we summarize the current state of the field aimed at: (1) understanding the genetic architecture of this complex disease, (2) synthesizing how this genetic risk architecture impacts cellular and molecular disease pathophysiology, (3) providing illustrative examples that highlight the rich complexity of the pathobiology of this prototypical autoimmune disease and (4) communicating this complex etiopathogenesis to patients.

Performance of the 2019 European League Against Rheumatism/American College of Rheumatology classification criteria for systemic lupus erythematosus in a multiethnic Malaysian cohort

AIM

To evaluate the performance of the 2019 European League Against Rheumatism/American College of Rheumatology (EULAR/ACR) classification criteria for systemic lupus erythematosus (SLE) in a multi-ethnic Malaysian cohort and to compare it against the Systemic Lupus International Collaborating Clinics (SLICC) 2012 and ACR 1997 criteria.

METHOD

We conducted a retrospective observational study of 205 patients with a diagnosis of SLE and 100 controls who formed the validation cohort. The sensitivity and specificity of the three classification criteria were evaluated and a further sub-analysis was performed in patients with early disease and among the various ethnicities.

RESULTS

The sensitivities and specificities of the three classification criteria are as follows: EULAR/ACR (90.8%; 94%), SLICC 2012 (96.1%; 94%), and ACR 1997 (82%; 96%). Among patients with early disease, the sensitivity of the SLICC 2012 was higher than that of EULAR/ACR and ACR 1997 (98% vs 94% and 86%); however, the specificity of EULAR/ACR and ACR 1997 were similar (95.2%) and higher than the SLICC 2012 (93.5%). The SLICC 2012 had higher sensitivity than that of the EULAR/ACR among the Malays (94% vs 90%), Chinese (98% vs 90%), and Indians (100% vs 95%). The specificity of the EULAR/ACR and SLICC 2012 were similar in the Malay and Chinese (93.3% each, and 92% vs 94.6%).

CONCLUSION

The EULAR/ACR performed well in our cohort. The EULAR/ACR and SLICC 2012 showed higher sensitivity than the ACR 1997, and the EULAR/ACR showed similar specificity to the ACR 1997 and SLICC 2012 overall, in early disease, and across the different ethnicities.

Comparison of the sensitivity of the EULAR / ACR 2019 and SLICC 2012 classification criteria in a Colombian population with systemic lupus erythematosus

BACKGROUND

/Objective: Systemic lupus erythematosus (SLE) is a multisystem autoimmune disease with a wide range of clinical manifestations. The latest classification criteria, EULAR/ACR 2019, have not been validated in a Latin American population of Amerindian ancestry. The objective of this study is to compare the sensitivity of the EULAR/ACR 2019 and SLICC 2012 classification criteria in a group of SLE patients with the above ancestry.

METHODS

A cross-sectional study was done. Data were obtained from the review of medical records of patients who met the inclusion criteria. The overall sensitivity of the criteria was calculated and compared to each other using the McNemar test.

RESULTS

146 medical records of patients from two referral centers were included. There were no differences in the sensitivity of the EULAR/ACR and SLICC 2012 criteria (84.9% versus 85.6% p = 0.79) nor were differences found when the groups based on disease duration were compared: less than 5 years (91.0% versus 92.5% p = 0.70), between 5 and 10 years (82.8% versus 82.8% p = 1), and 10 years or more (76.7% versus 76.7% p = 1). However, SLICC 2012 criteria was found to better classify patients with a less than 5-year disease duration compared to those with 10-year duration or more (92.5% versus 76.4% p = 0.024).

CONCLUSIONS

There are no statistically significant differences between the EULAR/ACR and SLICC 2012 criteria in the population studied. Nor were differences found when evaluating them by age at diagnosis and duration of the disease except when the group with less than 5 years of duration was compared to those with 10 years or more using the SLICC 2012 criteria.

Phenotypic landscape of systemic lupus erythematosus: An analysis of the Kyoto Lupus Cohort

OBJECTIVES

The present study aimed to clarify comprehensive relationships among the clinical variables of systemic lupus erythematosus (SLE).

METHODS

We retrospectively surveyed 32 clinical variables in 581 patients and conducted comprehensive association studies among SLE clinical phenotypes. A univariate analysis of all possible combinations was performed, and the results of phenotypic correlations were reduced into two dimensions. We also created a regression formula using L1 regularisation (LASSO) to calculate the probability of exhibiting each phenotype.

RESULTS

The univariate analysis identified 26 correlations, including multiple phenotypes with low complement. Some unpredicted correlations were identified, including fever and the anti-Sm antibody (odds ratio; OR = 2.3, p = 1.6 × 10-5) or thrombocytopenia and psychosis (OR = 3.7, p = 3.2 × 10-5). The multivariate analysis accurately estimated the probability of exhibiting each phenotype (area under the curve > 0.7) in 10 out of 20 phenotypes.

CONCLUSIONS

The present results show the phenotypic architecture of SLE and represent a model for estimating the probability of exhibiting each phenotype. They also offer insights into the pathology of SLE and estimating the probability of the onset of new phenotypes in clinical practice.

Systemic Lupus Erythematosus Classification and Diagnosis

Since the European League Against Rheumatism/American College of Rheumatology 2019 classification criteria for systemic lupus erythematosus (SLE) were published, they were externally validated by groups worldwide. In particular, the new criteria worked well also in East Asian and pediatric cohorts. Antinuclear antibodies (ANA) as an entry criterion were critically discussed, but the group of ANA-negative patients is small (<5%) worldwide. Specificity of the criteria is dependent on correct attribution only of those criteria that are not better explained by other causes. Although the classification criteria should not be used for diagnosis, many novel aspects inform diagnostic considerations.

Lupus or not? SLE Risk Probability Index (SLERPI): a simple, clinician-friendly machine learning-based model to assist the diagnosis of systemic lupus erythematosus

Objectives

Diagnostic reasoning in systemic lupus erythematosus (SLE) is a complex process reflecting the probability of disease at a given timepoint against competing diagnoses. We applied machine learning in well-characterised patient data sets to develop an algorithm that can aid SLE diagnosis.

Methods

From a discovery cohort of randomly selected 802 adults with SLE or control rheumatologic diseases, clinically selected panels of deconvoluted classification criteria and non-criteria features were analysed. Feature selection and model construction were done with Random Forests and Least Absolute Shrinkage and Selection Operator-logistic regression (LASSO-LR). The best model in 10-fold cross-validation was tested in a validation cohort (512 SLE, 143 disease controls).

Results

A novel LASSO-LR model had the best performance and included 14 variably weighed features with thrombocytopenia/haemolytic anaemia, malar/maculopapular rash, proteinuria, low C3 and C4, antinuclear antibodies (ANA) and immunologic disorder being the strongest SLE predictors. Our model produced SLE risk probabilities (depending on the combination of features) correlating positively with disease severity and organ damage, and allowing the unbiased classification of a validation cohort into diagnostic certainty levels (unlikely, possible, likely, definitive SLE) based on the likelihood of SLE against other diagnoses. Operating the model as binary (lupus/not-lupus), we noted excellent accuracy (94.8%) for identifying SLE, and high sensitivity for early disease (93.8%), nephritis (97.9%), neuropsychiatric (91.8%) and severe lupus requiring immunosuppressives/biologics (96.4%). This was converted into a scoring system, whereby a score >7 has 94.2% accuracy.

Conclusions

We have developed and validated an accurate, clinician-friendly algorithm based on classical disease features for early SLE diagnosis and treatment to improve patient outcomes.

New lupus criteria: a critical view

PURPOSE OF REVIEW

To review the validation of the European League Against Rheumatism (EULAR)/American College of Rheumatology (ACR) 2019 classification criteria for systemic lupus erythematosus (SLE).

RECENT FINDINGS

Positive antinuclear antibodies, which constitute the obligatory entry criterion of the EULAR/ACR criteria, were found in the vast majority of SLE patients worldwide, with 97% (94-100%) of patients antinuclear antibodies positive in studies investigating EULAR/ACR criteria performance. Combined over the publications, EULAR/ACR criteria sensitivity was 92% (range 85-97%). Specificity varied more relevantly, with the publications published after the EULAR/ACR 2019 criteria showing 93% (83-98%) specificity. Of particular relevance is the good performance of the EULAR/ACR criteria seen in pediatric SLE as well as in early SLE.

SUMMARY

The new classification criteria have been investigated in an impressive number of cohorts worldwide, adding to the data from the EULAR/ACR criteria project cohort. It is critical to strictly keep to the attribution rule, that items are only counted if there is no more likely alternative explanation than SLE, the domain structure, where only the highest weighted item in a domain counts, and the limitation to highly specific tests for antibodies to double-stranded DNA.

Lupus or not? SLE Risk Probability Index (SLERPI): a simple, clinician-friendly machine learning-based model to assist the diagnosis of systemic lupus erythematosus

OBJECTIVES

Diagnostic reasoning in systemic lupus erythematosus (SLE) is a complex process reflecting the probability of disease at a given timepoint against competing diagnoses. We applied machine learning in well-characterised patient data sets to develop an algorithm that can aid SLE diagnosis.

METHODS

From a discovery cohort of randomly selected 802 adults with SLE or control rheumatologic diseases, clinically selected panels of deconvoluted classification criteria and non-criteria features were analysed. Feature selection and model construction were done with Random Forests and Least Absolute Shrinkage and Selection Operator-logistic regression (LASSO-LR). The best model in 10-fold cross-validation was tested in a validation cohort (512 SLE, 143 disease controls).

RESULTS

A novel LASSO-LR model had the best performance and included 14 variably weighed features with thrombocytopenia/haemolytic anaemia, malar/maculopapular rash, proteinuria, low C3 and C4, antinuclear antibodies (ANA) and immunologic disorder being the strongest SLE predictors. Our model produced SLE risk probabilities (depending on the combination of features) correlating positively with disease severity and organ damage, and allowing the unbiased classification of a validation cohort into diagnostic certainty levels (unlikely, possible, likely, definitive SLE) based on the likelihood of SLE against other diagnoses. Operating the model as binary (lupus/not-lupus), we noted excellent accuracy (94.8%) for identifying SLE, and high sensitivity for early disease (93.8%), nephritis (97.9%), neuropsychiatric (91.8%) and severe lupus requiring immunosuppressives/biologics (96.4%). This was converted into a scoring system, whereby a score >7 has 94.2% accuracy.

CONCLUSIONS

We have developed and validated an accurate, clinician-friendly algorithm based on classical disease features for early SLE diagnosis and treatment to improve patient outcomes.

The Performances of the ACR 1997, SLICC 2012, and EULAR/ACR 2019 Classification Criteria in Pediatric Systemic Lupus Erythematosus

Objective.

Systemic lupus erythematosus (SLE) is a chronic autoimmune disease. The American College of Rheumatology (ACR) 1997, Systemic Lupus International Collaborating Clinics (SLICC) 2012, and European League Against Rheumatism (EULAR)/ACR 2019 SLE classification criteria are formed based on data mainly from adult patients. We aimed to test the performances of the SLE classification criteria among pediatric patients with SLE.

Methods.

Pediatric patients with SLE (n = 262; 80.9% female) were included from 3 different centers in Turkey. As controls, 174 children (60.9% female) with other diseases who had ANA (antinuclear antibody) test results were included. The gold standard for SLE diagnosis was expert opinion.

Results.

The sensitivities of the ACR 1997, SLICC 2012, and EULAR/ACR 2019 criteria were 68.7%, 95.4%, and 91.6%, respectively. The specificities of the ACR 1997, SLICC 2012, and EULAR/ACR 2019 criteria were 94.8%, 89.7%, and 88.5%, respectively. Eighteen patients with SLE met the SLICC 2012 but not the EULAR/ACR 2019 criteria. Among these, hematologic involvement was prominent (n = 13; 72.2%). Eight patients with SLE fulfilled the EULAR/ACR 2019 but not the SLICC 2012 criteria. Among these, joint involvement was prominent (n = 6; 75%).

Conclusion.

To our knowledge, this is the largest cohort study of pediatric SLE to test the performances of all 3 classification criteria. The SLICC 2012 criteria yielded the best sensitivity, whereas the ACR 1997 criteria had the best specificity. SLICC 2012 criteria performed better than EULAR/ACR 2019 criteria. Separation of different hematological manifestations in the SLICC 2012 criteria might have contributed to the higher performance of this criteria set.