2019 European League Against Rheumatism/American College of Rheumatology Classification Criteria for Systemic Lupus Erythematosus
Abstract
Objective
To develop new classification criteria for systemic lupus erythematosus (SLE) jointly supported by the European League Against Rheumatism (EULAR) and the American College of Rheumatology (ACR).
Methods
This international initiative had four phases. 1) Evaluation of antinuclear antibody (ANA) as an entry criterion through systematic review and meta-regression of the literature and criteria generation through an international Delphi exercise, an early patient cohort, and a patient survey. 2) Criteria reduction by Delphi and nominal group technique exercises. 3) Criteria definition and weighting based on criterion performance and on results of a multi-criteria decision analysis. 4) Refinement of weights and threshold scores in a new derivation cohort of 1,001 subjects and validation compared with previous criteria in a new validation cohort of 1,270 subjects.
Results
The 2019 EULAR/ACR classification criteria for SLE include positive ANA at least once as obligatory entry criterion; followed by additive weighted criteria grouped in 7 clinical (constitutional, hematologic, neuropsychiatric, mucocutaneous, serosal, musculoskeletal, renal) and 3 immunologic (antiphospholipid antibodies, complement proteins, SLE-specific antibodies) domains, and weighted from 2 to 10. Patients accumulating ≥10 points are classified. In the validation cohort, the new criteria had a sensitivity of 96.1% and specificity of 93.4%, compared with 82.8% sensitivity and 93.4% specificity of the ACR 1997 and 96.7% sensitivity and 83.7% specificity of the Systemic Lupus International Collaborating Clinics 2012 criteria.
Conclusion
These new classification criteria were developed using rigorous methodology with multidisciplinary and international input, and have excellent sensitivity and specificity. Use of ANA entry criterion, hierarchically clustered, and weighted criteria reflects current thinking about SLE and provides an improved foundation for SLE research.
This criteria set has been approved by the European League Against Rheumatism (EULAR) Executive Committee and the American College of Rheumatology (ACR) Board of Directors. This signifies that the criteria set has been quantitatively validated using patient data, and it has undergone validation based on an independent data set. All EULAR/ACR-approved criteria sets are expected to undergo intermittent updates.
The ACR is an independent, professional, medical and scientific society that does not guarantee, warrant, or endorse any commercial product or service.
Introduction
Systemic lupus erythematosus (SLE) is a complex autoimmune disease with variable clinical features 1, 2. SLE manifestations are associated with multiple autoantibodies, ensuing immune complex formation and deposition, and other immune processes 2, 3. This complex clinical presentation and pathogenesis makes SLE a difficult disease to grasp and define. Classification criteria are essential for the identification of relatively homogeneous groups of patients for inclusion in research studies and trials 4, 5. The 1982 revised American College of Rheumatology (ACR) SLE classification criteria 6 and their 1997 revision 7 have been used worldwide. Since then, our understanding of the disease has advanced. Additional specific skin manifestations were described, some clinical symptoms were better understood, and immunologic tests, such as diminished levels of serum complement components C3 and C4 or testing for anti–β2-glycoprotein I antibodies, entered routine clinical practice. Better understanding of organ system involvement, such as mucocutaneous abnormalities, led to questions about whether some of the independently counted criteria were in fact manifestations of the same phenomenon 8.
The 2012 Systemic Lupus International Collaborating Clinics (SLICC) classification criteria addressed many of these issues 9. Mucocutaneous and neuropsychiatric manifestations were added, as were hypocomplementemia and new antiphospholipid antibody tests; and criteria definitions were refined. The SLICC criteria emphasized that SLE is primarily an autoantibody disease, requiring at least one immunologic criterion to be present, and categorized histology-proven nephritis compatible with SLE as sufficient for classification, if antinuclear antibodies (ANAs) or antibodies to double-stranded DNA (dsDNA) were present. While achieving their goal of increasing sensitivity, the SLICC criteria have lower specificity than the 1997 ACR criteria 9, 10.
Existing SLE classification criteria perform better in patients with longstanding disease than in new-onset SLE 11, and there is an increasing recognition and demand that subjects with early SLE should be included in clinical studies and trials. We therefore attempted to enrich our sample populations for early SLE in several phases of the project.
In parallel with improved understanding of SLE, the field of classification criteria development has also seen advances 4, 12-14. In order to minimize investigator bias, it is now recommended that the cohorts in which the criteria are tested are from independent centers 4. Other methodologic recommendations include a balanced use of both expert-based and data-driven methods, and inclusion of the patient perspective 13, 14. The approach chosen for these 2019 European League Against Rheumatism (EULAR)/ACR SLE classification criteria was specifically designed to maintain this balance and to uphold rigorous methodology.
Methods
Methodologic overview
Using a methodologic approach based on measurement science the criteria were developed in four phases 10: 1) criteria generation, 2) criteria reduction, 3) criteria definition and weighting, and 4) refinement and validation (Figure 1). The whole initiative was overseen by a 12-member steering committee (MA, KHC, DID, MM, RR-G, JSS, DW, DTB, DLK, DJ, TD, and SRJ) nominated by EULAR and the ACR in equal numbers, based on SLE and/or methodologic experience and previous involvement in international projects.
The current project, jointly supported by the EULAR and the ACR, was originally based on two key concepts. One, we hypothesized that the presence of ANA would be better employed as an entry criterion than as a classification criterion 10. Such an approach was thought to reflect underlying SLE pathogenesis, and take into account ANA test characteristics of high sensitivity and limited specificity. Two, we expected individual criteria would not be of equal utility (weight) for the classification of SLE 15, for example, mucosal ulcers versus biopsy-proven lupus nephritis. Accordingly, the validity of using positive ANA as an entry criterion was explicitly addressed in phase I of the current activity 16. Likewise, methodologic strategies to develop weighted criteria were used.
Phase I: Criteria generation
The purpose of phase I was to test ANA as a potential entry criterion and identify candidate criteria that should be considered for SLE classification using both data-based and expert-based methods, including the patient perspective. Phase Ia comprised a systematic literature review of Medline, Embase, and the Cochrane databases with meta-regression to evaluate the operating characteristics of ANA testing for consideration as an entry criterion 16. Phase Ib consisted of a Delphi exercise of international SLE experts from the Americas, Europe, and Asia 17. These experts included rheumatologists, dermatologists, nephrologists, pediatricians, and non-clinical SLE researchers, providing a broad perspective. The Delphi participants were asked to nominate a broad set of items potentially useful in the classification of SLE 17. In rounds 2 and 3, participants rated the items from 1 (not at all appropriate) to 9 (completely appropriate) for classification of SLE. Criteria were retained if they reached a median rating of ≥6.5; that is, at least 50% of the ratings in the high range (7, 8, or 9). Participants were also asked about the importance of ANA and histopathology for classification of SLE. Phase Ic established an international cohort of patients with early SLE or conditions mimicking SLE to identify criteria that may discriminate subjects with early (less than 12 months) disease 18. Phase Id comprised a cross-sectional survey of SLE patients, administered via the quarterly journal of the German SLE patient organization, which asked about symptoms within 1 year before and after the patient's diagnosis of SLE 19. While at a risk of recall bias and not necessarily representative of other regions worldwide, this survey was done to explicitly take a patient standpoint into account.
For phases II and III, additional renowned European and North American SLE experts were nominated by the steering committee and invited to participate.
Phase II: Criteria reduction
Phase IIa
The objective of this phase was to select a set of criteria from phase I that maximized the likelihood of accurate classification of SLE, particularly of early disease. An independent panel of seven of the international SLE experts (RC, NC-C, DDG, BHH, FH, EM, and JS-G) ranked the candidate criteria from phase I. A consensus meeting of 19 international SLE experts (n = 7 nominal group technique [NGT] experts + steering committee + DK [moderator]) using NGT was conducted to reduce the list of criteria 20. Data for each candidate criterion were reviewed and discussed until consensus was achieved. The NGT experts voted on items to be retained.
Phase IIb
NGT participants pointed out that some criteria could be correlated. With the idea of potentially clustering criteria into domains, associations between candidate criteria were evaluated separately in two cohorts, the phase Ic early SLE and the Euro-lupus cohorts 21.
Phase III: Criteria definition and weighting
Phase IIIa
The operating characteristics of the retained candidate criteria were evaluated by literature review. Candidate criteria were hierarchically organized into clinical and immunologic domains, and definitions for the candidate criteria were iteratively refined. SLE patient advocates participated in the review of data and the steering committee discussions 22.
Phase IIIb
One hundred sixty-four case vignettes reflecting broad SLE clinical presentation were sampled from SLE centers across several countries. A panel of six of the international experts not involved in earlier phases of the project (BD, SJ, WJM, GR-I, MS, and MBU) and 11 members of the steering committee assessed and ranked a representative sample of the cases. Subsequently, at a face-to-face meeting, this panel of 17 international SLE experts iteratively compared pairs of criteria, using multicriteria decision analysis facilitated by 1000Minds software 23. The panel unanimously agreed to further reduce the list of criteria. Based on the results, provisional criteria weights were assigned and a provisional threshold score for classification was determined as the lowest score at which the expert panel had achieved consensus on classifying a case vignette as SLE 24.
Phase IV: Refinement and validation
International SLE experts not involved in phase II or phase III panels were asked to contribute cases diagnosed as SLE and controls with conditions mimicking SLE sampled from patients evaluated at their centers. Each center was asked to contribute up to 100 cases and an equal number of controls, preferentially sampling those with early disease, and regardless of their specific clinical or immunologic manifestations. Pseudonymized data on the criteria were collected using a standardized data collection form. Ethics committee approval and informed consent were obtained as per local requirements. The status (“SLE” or not) of each case underwent independent adjudication by three of four SLE experts (GB, BFH, NL, and CT) from different centers. Queries were sent back to the submitting investigator for clarification. Of this cohort, 501 SLE and 500 control subjects were randomly selected to comprise the derivation cohort, while the remaining 696 SLE and 574 control subjects formed the validation cohort.
Refinement
The performance of the draft criteria set was iteratively tested in the derivation cohort. A data-driven threshold for classification was determined by receiver operating characteristics (ROC) analysis and compared with the provisional expert-based consensus threshold. The data of SLE subjects below the threshold (misclassified) were reviewed for groups of patients with unequivocal SLE who still missed classification, and criteria weights adjusted slightly, while preserving the weighting hierarchy (details below in Results, Phase IV section). Sensitivity and specificity were tested against the ACR 1997 and the SLICC 2012 criteria. In addition, ANA as an entry criterion was tested against not having an entry criterion. Finally, the criteria weights were simplified to whole numbers. Refinements to the criteria set were presented to the steering committee and phase III expert panel, and unanimously endorsed.
Validation
The sensitivity and specificity of the final criteria were tested in the validation cohort and compared with previous SLE criteria sets.
Statistical analysis
Descriptive statistics were used to summarize the data. Confidence intervals (CIs) were calculated using the bias-corrected and accelerated bootstrap method (BCa method) with B = 2000 bootstrap samples. The BCa method resamples the input data B times (with replacement) and calculates the required statistics (sensitivity, specificity, AUC). Based on the B bootstraps samples, the bias-correction is applied and the associated 95% CIs for the statistics are estimated. The BCa method has proven to yield very accurate coverage of estimated CIs 25. The number B of bootstrap resamples is recommended to be at least B = 1,000. We have chosen B = 2,000 and additionally checked if B = 5,000 bootstraps changed the estimated confidence bounds, which was not the case. Statistical analyses were performed using R, v.3.4.0 (The R Foundation of Statistical Computing).
Results
Phase I: Criteria generation
Phase Ia: ANA as an entry criterion
A systematic review of Medline, Embase, and the Cochrane database identified 13,080 patients from 64 studies reporting ANA by immunofluorescence on HEp-2 cells. Meta-regression of the operating characteristics of ANA found a sensitivity of 97.8% (95% CI 96.8–98.5%) for ANA of ≥1:80, supporting use of ANA as an entry criterion 16. Since some SLE centers do not have access to HEp-2 ANA, and in view of ongoing work on the standardization of serology and potential future advances in the field, the steering committee and additional autoantibody consultants (MJF and PLM) recommended the provision “or an equivalent positive ANA test. Testing by immunofluorescence on HEp-2 cells or a solid-phase ANA screening immunoassay with at least equivalent performance is highly recommended.”
Phase Ib: Delphi exercise
One hundred forty-seven international SLE experts nominated 145 candidate criteria 17. By rating the appropriateness for SLE classification, the participants in the second and third Delphi rounds reduced the list to 40 candidate criteria (Supplementary Table 1, on the Arthritis & Rheumatology web site at http://onlinelibrary.wiley.com/doi/10.1002/art.40930/abstract).
Phase Ic: International early SLE cohort
The cohort comprised 616 subjects who had been referred for possible SLE with a disease duration of less than 1 year (n = 389 early SLE and n = 227 mimicking diseases) from North America, Europe, Asia, and South America 18. In addition to supporting many of the 40 candidate criteria derived from the Delphi exercise, the comparison between early SLE and non-SLE patients showed that fever occurred more frequently (34.5% versus 13.7%; P < 0.001) in SLE, while SLE patients less commonly suffered from arthralgias (20.3% versus 42.7%; P = 0.001) and fatigue (28.3% versus 37%; P = 0.02).
Phase Id: Patient survey
Three hundred thirty-nine SLE patients (>99% Caucasian, 93% female) responded to the survey 19. More than half of these patients reported mucocutaneous findings in the first year of their disease (Supplementary Table 1, http://onlinelibrary.wiley.com/doi/10.1002/art.40930/abstract), but also fatigue (89%), joint pain (87%), and fever (54%) 19. Given that these items were highlighted both in the early SLE cohort and the patient survey, fever, fatigue, and arthralgias were forwarded to the next phase in addition to the 40 Delphi items. Accordingly, phases Ia–Id resulted in a total of 43 candidate criteria for consideration (Supplementary Table 1).
Phase II: Criteria reduction
Phase IIa
The expert panel NGT exercise reduced the candidate criteria from 43 to 21 26. The panel distinguished potential “entry criteria,” which would be required for classification, from potential “additive criteria.” They endorsed “positive ANA (≥1:80 by HEp-2 immunofluorescence)” as an entry criterion. The 20 remaining additive criteria included: lupus nephritis by renal biopsy, autoantibodies, cytopenias, fever, arthritis, serositis, mucocutaneous and neuropsychiatric manifestations (Supplementary Table 1).
Phase IIb
Associations between the candidate criteria were evaluated in 389 subjects in the early SLE cohort and the 1,000 SLE subjects of the Euro-lupus cohort. Modest statistically significant correlations were limited to the mucocutaneous (r = 0.22–0.30), neurologic (r = 0.22), and immunologic (r = 0.33) domains in the early SLE cohort, and this modest correlation was replicated in the Euro-lupus cohort 21. Given these associations, criteria were clustered within domains, so that only 1 criterion within each domain would be counted.
Phase III: Criteria definition and weighting
Phase IIIa
Based on the literature, definitions of the 20 candidate additive criteria were refined, using a data-driven evaluation of operating characteristics 22, retaining only feasible items with a prevalence of at least 1% according to literature. Literature review led to the consensus decision to evaluate 5 different candidate criteria within the neuropsychiatric domain (delirium, psychosis, seizure, mononeuropathy, cranial neuropathy) and potential separation of acute pericarditis from pleural or pericardial effusions and between diminished C3 or C4 versus diminished C3 and C4. The resulting 23 candidate criteria (Supplementary Table 1, on the Arthritis & Rheumatology web site at http://onlinelibrary.wiley.com/doi/10.1002/art.40930/abstract) were organized into 7 clinical and 3 immunologic domains, with hierarchical clustering 22. Only the highest-ranking item in each domain was to be counted. Instead of devising exclusion definitions for each criterion, the decision was made to attribute any item to SLE only if no more likely explanation was present. For leukopenia and joint involvement, it was decided to formally test alternative definitions in the derivation cohort. Given the importance of testing for antibodies, particularly for anti-dsDNA, for which tests of relatively low specificity are in use, great care was taken to precisely define testing (Table 1).
| Criteria | Definition |
|---|---|
| Antinuclear antibodies (ANA) | ANA at a titer of ≥1:80 on HEp-2 cells or an equivalent positive test at least once. Testing by immunofluorescence on HEp-2 cells or a solid-phase ANA screening immunoassay with at least equivalent performance is highly recommended |
| Fever | Temperature >38.3°C |
| Leukopenia | White blood cell count <4,000/mm³ |
| Thrombocytopenia | Platelet count <100,000/mm³ |
| Autoimmune hemolysis | Evidence of hemolysis, such as reticulocytosis, low haptoglobin, elevated indirect bilirubin, elevated LDH, AND positive Coombs' (direct antiglobulin) test |
| Delirium | Characterized by 1) change in consciousness or level of arousal with reduced ability to focus, 2) symptom development over hours to <2 days, 3) symptom fluctuation throughout the day, 4) either 4a) acute/subacute change in cognition (e.g., memory deficit or disorientation), or 4b) change in behavior, mood, or affect (e.g., restlessness, reversal of sleep/wake cycle) |
| Psychosis | Characterized by 1) delusions and/or hallucinations without insight and 2) absence of delirium |
| Seizure | Primary generalized seizure or partial/focal seizure |
| Non-scarring alopecia | Non-scarring alopecia observed by a clinician† |
| Oral ulcers | Oral ulcers observed by a clinician† |
| Subacute cutaneous OR discoid lupus | Subacute cutaneous lupus erythematosus observed by a clinician:† Annular or papulosquamous (psoriasiform) cutaneous eruption, usually photodistributed If skin biopsy is performed, typical changes must be present (interface vacuolar dermatitis consisting of a perivascular lymphohistiocytic infiltrate, often with dermal mucin noted). ORDiscoid lupus erythematosus observed by a clinician:† Erythematous-violaceous cutaneous lesions with secondary changes of atrophic scarring, dyspigmentation, often follicular hyperkeratosis/plugging (scalp), leading to scarring alopecia on the scalp If skin biopsy is performed, typical changes must be present (interface vacuolar dermatitis consisting of a perivascular and/or periappendageal lymphohistiocytic infiltrate. In the scalp, follicular keratin plugs may be seen. In longstanding lesions, mucin deposition may be noted) |
| Acute cutaneous lupus | Malar rash or generalized maculopapular rash observed by a clinician† If skin biopsy is performed, typical changes must be present (interface vacuolar dermatitis consisting of a perivascular lymphohistiocytic infiltrate, often with dermal mucin noted. Perivascular neutrophilic infiltrate may be present early in the course) |
| Pleural or pericardial effusion | Imaging evidence (such as ultrasound, x-ray, CT scan, MRI) of pleural or pericardial effusion, or both |
| Acute pericarditis | ≥2 of 1) pericardial chest pain (typically sharp, worse with inspiration, improved by leaning forward), 2) pericardial rub, 3)EKG with new widespread ST elevation or PR depression, 4) new or worsened pericardial effusion on imaging (such as ultrasound, x-ray, CT scan, MRI) |
| Joint involvement | EITHER 1) synovitis involving 2 or more joints characterized by swelling or effusion OR 2) tenderness in 2 or more joints and at least 30 minutes of morning stiffness |
| Proteinuria >0.5 g/24 hours | Proteinuria >0.5 g/24 hours by 24-hour urine or equivalent spot urine protein-to-creatinine ratio |
| Class II or V lupus nephritis on renal biopsy according to ISN/RPS 2003 classification | Class II: Mesangial proliferative lupus nephritis: purely mesangial hypercellularity of any degree or mesangial matrix expansion by light microscopy, with mesangial immune deposit. A few isolated subepithelial or subendothelial deposits may be visible by immunofluorescence or electron microscopy, but not by light microscopy Class V: Membranous lupus nephritis: global or segmental subepithelial immune deposits or their morphologic sequelae by light microscopy and by immunofluorescence or electron microscopy, with or without mesangial alterations |
| Class III or IV lupus nephritis on renal biopsy according to ISN/RPS 2003 classification | Class III: Focal lupus nephritis: active or inactive focal, segmental, or global endocapillary or extracapillary glomerulonephritis involving <50% of all glomeruli, typically with focal subendothelial immune deposits, with or without mesangial alterations Class IV: Diffuse lupus nephritis: active or inactive diffuse, segmental, or global endocapillary or extracapillary glomerulonephritis involving ≥50% of all glomeruli, typically with diffuse subendothelial immune deposits, with or without mesangial alterations. This class includes cases with diffuse wire loop deposits but with little or no glomerular proliferation |
| Positive antiphospholipid antibodies | Anticardiolipin antibodies (IgA, IgG, or IgM) at medium or high titer (>40 APL, GPL, or MPL, or >the 99th percentile) or positive anti-β2GPI antibodies (IgA, IgG, or IgM) or positive lupus anticoagulant |
| Low C3 OR low C4 | C3 OR C4 below the lower limit of normal |
| Low C3 AND low C4 | Both C3 AND C4 below their lower limits of normal |
| Anti-dsDNA antibodies OR anti-Sm antibodies | Anti-dsDNA antibodies in an immunoassay with demonstrated ≥90% specificity for SLE against relevant disease controls OR anti-Sm antibodies |
- * SLE = systemic lupus erythematosus; LDH = lactate dehydrogenase; CT = computed tomography; MRI = magnetic resonance imaging; EKG = electrocardiography; ISN = International Society of Nephrology; RPS = Renal Pathology Society; anti-β2GPI = anti–β2-glycoprotein I; anti-dsDNA = anti–double-stranded DNA.
- † This may include physical examination or review of a photograph.
Phase IIIb
The 1.5-day in-person consensus meeting using multicriteria decision analysis involved 74 decisions between pairs of criteria. Criteria weights were calculated by 1000Minds software based on these decisions (Table 2). International Society of Nephrology/Renal Pathology Society class III or IV nephritis consistently attained higher weight than class II or V nephritis, so lupus nephritis by histology was separated into 2 different criteria. Class VI lupus nephritis as an end-stage manifestation was unanimously eliminated. Likewise, the experts unanimously voted to not retain mononeuropathy and cranial neuropathy, which had been included into the set of potential neuropsychiatric items in phase IIIa but turned out to add little to SLE classification. The use of weighted criteria led to a sum score that is a measure of the relative probability of a subject having SLE, with higher scores indicating higher likelihood. Experts reached full consensus on a classification of SLE at a provisional threshold score of >83 of a theoretical maximum of 305 24.
| Domain | Item | Original | Modification | Revised | Simplified |
|---|---|---|---|---|---|
| Constitutional | Fever | 13 | 13 | 2 | |
| Hematologic | Leukopenia | 12 | +7 | 19 | 3 |
| Thrombocytopenia | 26 | 26 | 4 | ||
| Autoimmune hemolysis | 28 | 28 | 4 | ||
| Neuropsychiatric | Delirium | 12 | 12 | 2 | |
| Psychosis | 20 | 20 | 3 | ||
| Seizure | 34 | 34 | 5 | ||
| Mucocutaneous | Alopecia | 13 | 13 | 2 | |
| Oral ulcers | 14 | 14 | 2 | ||
| SCLE/DLE | 29 | 29 | 4 | ||
| ACLE | 38 | 38 | 6 | ||
| Serosal | Effusion | 34 | 34 | 5 | |
| Acute pericarditis | 38 | 38 | 6 | ||
| Musculoskeletal | Joint involvement | 34 | +4 | 38 | 6 |
| Renal | Proteinuria | 27 | 27 | 4 | |
| Class II/V | 55 | 55 | 8 | ||
| Class III/IV | 74 | 74 | 10 | ||
| Antiphospholipid antibodies | Antiphospholipid | 13 | 13 | 2 | |
| Complements | C3 or C4 low | 19 | 19 | 3 | |
| C3 and C4 low | 27 | 27 | 4 | ||
| SLE-specific antibodies | Anti-Sm | 40 | 40 | 6 | |
| Anti-dsDNA | 38 | 38 | 6 |
- * Weights derived from the phase III consensus meeting with multicriteria decisions analysis (original), added points for leukopenia and joint involvement (modification), the resulting weights (revised), and the final simplified weights (simplified). SCLE = subacute cutaneous lupus erythematosus; DLE = discoid lupus erythematosus; ACLE = acute cutaneous lupus erythematosus; SLE = systemic lupus erythematosus; anti-dsDNA = anti–double-stranded DNA.
Phase IV: Refinement and validation
Twenty-one centers from the US, Canada, Mexico, Austria, Croatia, France, Germany, Greece, Hungary, Italy, Portugal, Spain, the UK, Turkey, Hong Kong, and Japan submitted a total of 2,339 cases from their cohorts; 1,197 SLE and 1,074 non-SLE diagnoses (Table 3) were verified by 3 adjudicators blinded to the proposed classification criteria system. Due to lack of consensus during adjudication, 68 subjects (2.9%) were excluded from the analysis.
| Derivation cohort | Validation cohort | |||
|---|---|---|---|---|
| SLE | Non-SLE | SLE | Non-SLE | |
| n | 501 | 500 | 696 | 574 |
| Female/male | 447/54 | 421/79 | 608/88 | 490/84 |
| Age, mean ± SD years | 45 ± 14 | 54 ± 16 | 45 ± 14 | 56 ± 16 |
| Disease duration, mean ± SD years | 11 ± 8 | 9 ± 8 | 11 ± 8 | 9 ± 8 |
| Ethnicity | ||||
| Black | 29 | 10 | 56 | 12 |
| East Asian | 36 | 29 | 53 | 34 |
| Hispanic | 59 | 48 | 73 | 51 |
| South/Southeast Asian | 16 | 6 | 21 | 11 |
| White | 355 | 404 | 480 | 461 |
| Other | 6 | 3 | 13 | 5 |
| SLE | 501 | 696 | ||
| Non-SLE | 500 | 574 | ||
| Adult-onset Still's disease | 2 | 11 | ||
| Autoimmune thyroiditis | 6 | 5 | ||
| Behçet's disease | 7 | 9 | ||
| Cancer | 2 | 3 | ||
| Inflammatory myositis | 37 | 27 | ||
| Fibromyalgia | 6 | 3 | ||
| Membranous nephritis | 11 | 14 | ||
| Mixed connective tissue disease | 9 | 15 | ||
| Osteoarthritis | 2 | 0 | ||
| Primary antiphospholipid antibody syndrome | 45 | 48 | ||
| Psoriatic arthritis | 12 | 9 | ||
| Rheumatoid arthritis | 94 | 110 | ||
| Sarcoidosis | 2 | 2 | ||
| Sjögren's syndrome | 112 | 124 | ||
| Spondyloarthritis | 5 | 5 | ||
| Systemic sclerosis | 99 | 120 | ||
| Tuberculosis | 0 | 2 | ||
| Undifferentiated connective tissue disease | 16 | 20 | ||
| Vasculitis | 9 | 13 | ||
| Viral infection | 5 | 5 | ||
| Other | 19 | 29 | ||
- * Inflammatory myositis includes dermatomyositis, polymyositis, and juvenile dermatomyositis. SLE = systemic lupus erythematosus.
Derivation cohort
Of the 2,271 triple-adjudicated cases, 501 SLE and 500 non-SLE cases were randomly assigned to the derivation cohort. The provisional weighting system derived from phase III was tested in the derivation cohort. ROC analysis suggested a data-driven threshold of ≥70 (of a maximum of 305), with a sensitivity of 95.4% and a specificity of 95.2%, which was superior to the consensus-derived provisional threshold of >83 that had high specificity (98.8%), but lower sensitivity (81.6%). Review of subjects below the threshold of 70 identified a subgroup of SLE subjects with joint involvement and/or leukopenia. Thus, weights for leukopenia and joint involvement were each adjusted (Table 2) to reduce misclassification. When alternative definitions for leukopenia and joint involvement were tested, leukopenia defined as a white blood cell count (WBC) <4000/mm3 at least once 9 also had a slightly higher sensitivity + specificity (1.944 versus 1.942) than leukopenia defined as WBC <4,000/mm3 on 2 or more occasions 6, 26. Joint involvement defined as EITHER “synovitis involving 2 or more joints, characterized by swelling or effusion,” OR “tenderness in 2 or more joints and at least 30 minutes of morning stiffness” 9 had a higher combined sensitivity and specificity than arthritis defined simply as synovitis of 2 or more joints (1.944 versus 1.900). When retested, the revised criteria had increased sensitivity, and maintained sensitivity + specificity. Evaluating ANA as an entry criterion, the criteria with the ANA entry criterion had better performance than without (sensitivity + specificity 1.944 versus 1.930). Next, the weights were simplified by division to whole numbers to achieve a threshold of 10 (Table 2). In the derivation cohort, the sensitivity and specificity of the final criteria set (Figure 2) were reaching the performance benchmarks set for this project (Table 4).
| ACR 1997 criteria | SLICC 2012 criteria | EULAR/ACR 2019 criteria | |
|---|---|---|---|
| Derivation | |||
| Sensitivity (95% CI) | 0.85 (0.81–0.88) | 0.97 (0.95–0.98) | 0.98 (0.97–0.99) |
| Specificity (95% CI) | 0.95 (0.93–0.97) | 0.90 (0.87–0.92) | 0.96 (0.95–0.98) |
| Combined (95% CI) | 1.80 (1.76–1.83) | 1.87 (1.84–1.90) | 1.94 (1.92–1.96) |
| Validation | |||
| Sensitivity (95% CI) | 0.83 (0.80–0.85) | 0.97 (0.95–0.98) | 0.96 (0.95–0.98) |
| Specificity (95% CI) | 0.93 (0.91–0.95) | 0.84 (0.80–0.87) | 0.93 (0.91–0.95) |
| Combined (95% CI) | 1.76 (1.73–1.80) | 1.80 (1.77–1.84) | 1.90 (1.87–1.92) |
- * ACR = American College of Rheumatology; SLICC = Systemic Lupus International Collaborating Clinics; EULAR = European League Against Rheumatism; 95% CI = 95% confidence interval.
Validation
The validation cohort, that is, the full cohort minus the derivation cohort, comprised 1,270 triple-adjudicated subjects (n = 696 SLE, n = 574 controls). The criteria, with positive ANA as an entry criterion, weighted criteria in 7 clinical domains (constitutional, hematologic, neuropsychiatric, mucocutaneous, serosal, musculoskeletal, renal) and 3 immunologic domains (antiphospholipid antibodies, low complements, anti-Sm and anti-dsDNA as SLE-specific antibodies), and a classification threshold score of ≥10 (out of a theoretical maximum of 51) (Figure 2), had a sensitivity of 96.1% and a specificity of 93.4% (Table 4). It demonstrated improved performance compared with the ACR 1997 and SLICC 2012 criteria.
Discussion
New SLE classification criteria were developed with support by both the ACR and EULAR. Through a four-phase, iterative process, we have defined an additive, weighted multicriteria system that produces a measure of the relative probability that an individual can be classified as SLE. The system defines a threshold above which experts would classify cases as SLE for the purpose of research studies. We have carefully defined the criteria to improve reliability and precision, and have grouped the criteria into 10 hierarchical domains. We have validated the criteria against a large number of cases, including many patients with manifestations that resemble SLE but who do not have SLE. This approach, as well as the resulting criteria system, represents a paradigm shift for the classification of SLE.
We have defined positive ANA at any time as required entry criterion. There were three possible ways to deal with ANA testing. The previous criteria sets have treated ANA the same as the much more specific antibodies against Sm and dsDNA, which we considered suboptimal given important differences in sensitivity and specificity. We could have excluded ANA completely in classifying lupus, but we still consider ANA a useful test and concept. We therefore decided to test ANA as an entry criterion, which reflects the use of ANA as a highly sensitive screening test.
Criteria using ANA as entry criterion had better performance. During the phase I Delphi exercise, 58% of SLE experts did not feel comfortable and an additional 19% were uncertain about classifying a patient with SLE in the absence of ever having a positive ANA 17. The systematic literature review and meta-regression of data on 13,080 subjects demonstrated ANA ≥1:80 to have a sensitivity of 98% with a lower limit of the 95% CI at 97% 16. In the phase I early SLE cohort, 99.5% of the 389 SLE patients were ANA positive 18. The frequencies of ANA-positive SLE patients in the derivation and validation cohorts (99.6% and 99.3%, respectively) were in the same range. Since both in the early SLE cohort and in the derivation and validation cohorts, patients were included in many centers worldwide independent of ANA positivity, the latter data provide additional support for ANA as an entry criterion.
Using ANA as entry criterion means the new criteria cannot classify SLE among patients who are persistently ANA negative. While possibly also distinguished by lower cytokine levels 27 and lower efficacy of immunomodulatory treatment 28, such a subgroup of patients exists. Although small, it may vary in size in different populations 16. This patient subset needs to be put high on the scientific agenda for further investigation. Additional characterization of this phenomenon may lead to an alternative entry criterion for this small group of patients. For the moment, we still think it is acceptable to exclude ANA-negative patients from clinical trials.
Molecular classification criteria were also considered during the development of these criteria 29. Many novel biomarkers were nominated, such as increased circulating B lymphocyte stimulator (BLyS), interferon-γ (IFNγ)–induced protein 10 kd (IP-10), monocyte chemoattractant protein 1 (MCP-1), tumor necrosis factor α (TNFα), type I IFN signature, or increased Th17 and plasma cell populations. They were all voted out in the expert Delphi exercise, largely because of limited availability in the clinical setting and/or insufficient evidence 5. However, inclusion of novel biomarkers, beyond autoantibodies, may ultimately further improve the specificity of SLE classification, increase alignment of classification with underlying disease pathogenesis, and improve the performance and information content of clinical trials. Thus, testing of biomarkers against these criteria is an important area for future research.
A new clinical criterion, unexplained fever, turned out to be common and remarkably characteristic for SLE. However, since infections are a major cause of death in SLE, it is of utmost importance to stress that fever, like all other criteria manifestations, should only be counted if no better explanation exists, and that infections have to be suspected first in any patient with (potential) SLE, particularly when C-reactive protein is elevated 30. The concept that all criteria are only to be counted if SLE is thought to be the most likely cause of the manifestation (i.e., no other more likely cause exists) is central to these new EULAR/ACR criteria, and is explicitly stated as an overarching principle. Some criteria, such as delirium, psychosis, and acute pericarditis, were in part redefined based on existing scientific definitions 22. Where alternative definitions were used, the performance of the alternative definitions was comparatively evaluated in the derivation cohort.
The differential weighting of criteria better represents their relative contribution to an individual's classification of SLE. For SLE, renal biopsy with class III or IV lupus nephritis carries the most weight and in the presence of a positive ANA is enough to classify a patient as SLE. This further develops a concept of the SLICC criteria 9 and reflects the current thinking of SLE experts; in the Delphi exercise, 85% would classify SLE on renal pathology alone 17. Renal biopsy with class II or V lupus nephritis still carries a large weight (8 points) but is not by itself sufficient for the classification of SLE.
The numerical goal of this project was to keep the specificity similar to the specificity of the ACR 1997 criteria, but increase the sensitivity to the high sensitivity level of the SLICC criteria, if possible. The validation cohort data suggest that this goal has been achieved. From our data, it appears that the SLICC criteria increase in sensitivity was to a significant degree founded in accepting renal histology and adding subacute cutaneous lupus and low complement levels. These three advances are mirrored in the current criteria. Many of the other additional symptoms of the SLICC criteria were of very low frequency. Specificity was increased by weighting of criteria, by the NGT expert panel decision to not allow lymphopenia to go forward, and, importantly, by the decision that no criterion be counted if better explained by another condition.
The new criteria provide a simple, directed, and highly accurate method for classifying SLE. An electronic “app” is in preparation, which will assist in the use of these criteria. However, it is important to stress that classification criteria are not designed for diagnosis or treatment decisions 5. They should never be used to exclude patients who do not fully meet these criteria from receiving appropriate therapies. This is also pertinent to patients with ANA-negative SLE discussed above. Diagnosis of SLE remains the purview of an appropriately trained physician evaluating an individual patient 5.
The new SLE classification system also provides new research opportunities. With much interest in early or latent SLE 31, 32, the additive point system and the relative probability of classification it produces allows for systematic study of individuals who fall below the classification threshold. This will facilitate studies of disease evolution and early intervention. Furthermore, the use of an additive scoring system will allow for studying the idea of “ominousity,” that is, the potential implications of having very high scores on disease severity and subsequent prognosis. This work would need to reconsider the relative contribution of individual criteria (weights) and consider additional criteria that potentially contribute to ominousity.
It is anticipated that other groups will test these criteria, which will constitute important external validation. This will be particularly important for pediatric SLE and those with organ-dominant, for example, skin-dominant, disease, since it is a limitation of this criteria project that the patient cohorts do not represent these subgroups. Similar limitations also pertain to several racial/ethnic groups (for example, African American/Black, Hispanic, and Asian patients) and to men with SLE, each only included in lower numbers (Table 3). It is important to independently test the EULAR/ACR criteria in these subgroups. Leukocyte counts, for example, are more frequently below 4,000/mm3 in African Americans 33, which may have an influence on criteria performance. It is also possible that the academic center patient populations included differ from patients in community practice clinics. Investigators testing the new criteria in different populations are reminded about the critical importance of the correct attribution of each criterion. Criteria can only be counted when not better explained by another condition (see Supplementary Table 2, on the Arthritis & Rheumatology web site at http://onlinelibrary.wiley.com/doi/10.1002/art.40930/abstract). The attribution process requires diligence and clinical experience.
In summary, our multiphase methodologic approach and ensuing classification system using ANA as an entry criterion and weighted, hierarchically clustered criteria constitute a paradigm shift in the classification of SLE. These criteria have excellent performance characteristics and face validity, as the structure and weighting were designed to reflect current thinking about SLE. The inclusion of fever assists with the classification of early SLE. The separation of renal biopsy findings reflects their differential impact on the probability of SLE classification. These criteria have strong operating characteristics, with excellent sensitivity and specificity. This classification system was built using rigorous methodology that was both data-driven and expert-based. With the inclusion of over 200 SLE experts from multiple countries and medical disciplines, methodologists, patient advocates, and over 4,000 subjects, this work is the largest international, collaborative SLE classification effort to date.
Acknowledgments
The authors wish to acknowledge the diligent work of Banita Aggarwal and Keshini Devakandan in data entry, data cleaning, queries to submitting investigators, data cutting, and maintenance of the derivation and validation cohorts; and of Corine Sinnette, MA in the preparatory work for the multicriteria decision analysis exercise.
Author Contributions
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Drs. Aringer and Johnson had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design
Aringer, Costenbader, Daikh, Brinks, Mosca, Ramsey-Goldman, Smolen, Wofsy, Boumpas, Kamen, Jayne, Naden, Dörner, Johnson.
Acquisition of data
Aringer, Costenbader, Daikh, Mosca, Bertsias, Hoyer, Leuchten, Tani, Tedeschi, Touma, Schmajuk, Anic, Assan, Chan, Clarke, Crow, Czirják, Doria, Graninger, Halda-Kiss, Hasni, Izmirly, Jung, Kumánovics, Mariette, Padjen, Pego-Reigosa, Romero-Díaz, Rúa-Figueroa Fernández, Seror, Stummvoll, Tanaka, Tektonidou, Vasconcelos, Vital, Wallace, Yavuz, Meroni, Fritzler, Naden, Dörner, Johnson.
Analysis and interpretation of data
Aringer, Costenbader, Daikh, Brinks, Mosca, Ramsey-Goldman, Smolen, Wofsy, Boumpas, Kamen, Jayne, Cervera, Costedoat-Chalumeau, Diamond, Gladman, Hahn, Hiepe, Jacobsen, Khanna, Lerstrøm, Massarotti, McCune, Ruiz-Irastorza, Sanchez-Guerrero, Schneider, Urowitz, Bertsias, Hoyer, Leuchten, Tani, Tedeschi, Touma, Schmajuk, Meroni, Fritzler, Naden, Dörner, Johnson.
