Systemic lupus erythematosus is characterized by a wide variety of autoantibodies, which, if pathogenic, can lead to inflammation, damage, thrombosis or functional defects of essentially all organ systems. The 1982 American College of Rheumatology classification criteria and their 1997 revision have greatly influence the disease concept over the last decades. With the at least 4 out of 11 criteria to be present, they essentially depict the concept of multiple autoantibodies and multiple organs systems. Putting all items on equal weight made the system relatively easy to use and to memorize, but this approach was not entirely intuitive. Dermatologists have criticized that patients could meet the criteria based on fulfilling all 4 mucocutaneous criteria only. Importantly, sensitivity was felt to be suboptimal, reaching 83% in the cohort of the SLE International Collaborating Clinics (SLICC) group, and lower values in very early disease.
The 2012 SLICC SLE classification criteria introduced two new concepts. First, patients have to be positive for autoantibodies, namely for anti-nuclear antibodies (ANA) or antibodies to double-stranded DNA. Second, if patients are autoantibody-positive, classification criteria can be fulfilled by nephritis on histology without any other items. The SLICC groups also introduced large lists of skin and neuropsychiatric symptoms and refined some of the definitions. While many of these ideas are intuitively correct, not all were derived in a pre-defined scientific way. These criteria managed to increase sensitivity to 97% in their own validation data and to a range of 92–99% in other cohorts. At the same time, however, the SLICC set lost specificity, falling to 84% as compared to the 96% of the ACR criteria. This trade-off was presumably influenced by the SLICC group's choice to keep the overall structure the same. Early disease sensitivity was still not optimal.
In a large transatlantic project jointly supported by EULAR and the ACR, we have over the last years tried to develop (even) better SLE classification criteria. The main goals are to have a relatively intuitive set that helps in teaching, increase sensitivity as compared to the ACR criteria, but maintain specificity in the same range, improve the performance in early SLE, involve the larger SLE community as far as possible, and do this in a strictly scientific way. Given that ANA often are the door to SLE in the diagnostic approach, and that ANA have very high sensitivity, but modest specificity for SLE, we explored whether ANA could be used as an entry criterion. Meta-regression, after compiling ANA data of 13,080 SLE patients from a systemic literature search, gave a sensitivity of 97.8% (CI 96.8–98.5%) for a titer of at least 1:80 on HEp2-ANA immunofluorescence. It was therefore decided that HEp2-ANA of ≥1:80 would be used as an entry criterion.
Three different approaches were used to maximize the overview on potential items, namely an SLE expert Delphi exercise involving 123 SLE experts, an international early disease cohort with 389 SLE patients and 227 patients with conditions mimicking SLE, and a patient survey, which 339 SLE patients filled out and mailed anonymously. The resulting 41 items were then reduced in a nominal group technique exercise with 7 international SLE experts, resulting in a list of 20 items plus ANA as an entry criterion. These items were then submitted to multi-criteria decision analysis in a two day conference with 6 international experts together with the steering committee. This approach under the same moderator (Dr. Ray Naden) had already been successfully used for the rheumatoid arthritis and systemic sclerosis criteria.
The present results are 20 weighted items, which can be additively used within a continous scale, and a cut-off for classification. Weights for class III or IV lupus nephritis are much higher than for leukopenia or unexplained fever. This candidate system will now be tested against both the ACR and the SLICC criteria in a large cohort of SLE patients and patients with mimicking conditions. If successful, ANA of ≥1:80 and weighted criteria will lead to better performance, particularly in early disease, and give us a system that is hopefully intuitive enough to convey an idea of the disease.
Disclosure of Interest None declared
Statistics from Altmetric.com
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.