• systemic lupus erythematosus
• autoimmune diseases
• epidemiology

In their letter, Dr Sacre and colleagues1 describe an interesting retrospective study on 129 patients with pericardial effusion, of whom 17 were diagnosed with systemic lupus erythematosus (SLE). The authors arrive at a reassuring sensitivity of 100% for the new European League Against Rheumatism/American College of Rheumatology (EULAR/ACR) 2019 classification criteria.2 3 However, specificity was clearly lower at 84%, below the specificity of the ACR and the Systemic Lupus International Collaborating Clinics (SLICC) criteria. The latter is a somewhat unexpected result. Based on the experience of the last months, the most common reasons for suboptimal specificity in applying the EULAR/ACR 2019 criteria are incorrect attribution to SLE and reliance on non-specific serological tests. Given the data presented by Dr Sacre and colleagues,1 both may also have played a role in misclassifying 2 of 6 patients with other autoimmune disease and 3 of 26 patients with idiopathic pericarditis as having SLE.

In addition to serositis, EULAR/ACR classification criteria items in non-SLE patients were limited to fever, joint involvement and low C4 in patients with other autoimmune diseases.1 By attribution rule,2–4 these items should not be counted for SLE if due to another autoimmune disease. Similarly, patients without autoimmune disease had fever and proteinuria besides serositis. Again, these should not be counted if in fact attributed to an infection.

Moreover, 2 of 26 patients without an autoimmune disease were reported to have antibodies against double-stranded DNA (dsDNA). This should depend on an assay with at least 95% specificity against relevant disease controls,2 3 usually a Crithidia test or radioimmune assay. Such test would be unlikely to become positive in idiopathic pericarditis. Anti-dsDNA tests of lower specificity, on the other hand, should not be counted.

While we thank Dr Sacre and colleagues1 for their interesting data and while the sensitivity results are reassuring, we would like to remind authors of the importance of following the attribution rule of the EULAR/ACR criteria, that the criteria items should only be counted if there is no more likely alternative explanation.2–4 Since this single rule has replaced the exclusion criteria for individual items, not following the attribution rule will inevitably lead to underestimating specificity when the new SLE classification criteria are applied.