A Delphi exercise and cluster analysis to aid in the development of potential classification criteria for systemic sclerosis using SSc experts and databases.

OBJECTIVES: Since the 1980 ACR classification criteria for systemic sclerosis (SSc) do not identify 20% with SSc, revised criteria are necessary.

METHODS: Suggested new criteria from the literature were sent in random order to 96 SSc experts. A 3-round Delphi Consensus eliminated criteria. Then cluster analysis reduced items. The Canadian Scleroderma Research Group (CSRG) database was used to determine the prevalence of each item.

RESULTS: Seventy-one of 96 (71%) completed all 3 rounds; 47 items were expanded to 76 in round 2. Thirty items had at least 50% consensus and 18 had >75% agreement to include (a priori cut point). Clustering occurred for 4 categories: proximal to MCP skin involvement, vascular abnormalities, autoantibodies and tissue damage. Proximal to MCPs skin involvement identified 80% of patients. Adding one item from each of the other 3 categories or 1 or more items from 2 of 3 remaining categories increased the proportion of patients classified to 94% in CSRG patients. Categories included (1) Vascular (dilated capillaries, telangiectasia, Raynaud's phenomenon [RP]), (2) Autoantibodies (anticentromere [ACA] or antitopoisomeraseI [Topo1]) and (3) Fibrosis/damage (esophogeal dysmotility dysphagia, sclerodactyly, digital ulcers). In the CSRG, 98% were identified if using proximal skin involvement; or sclerodactyly plus one of: RP, ACA or Topo1.

CONCLUSIONS: This is a first step toward developing new SSc classification criteria. A Delphi exercise alone cannot suffice for item reduction. Also, validation prospectively in SSc patients and diseases that mimic SSc is needed in order to calculate sensitivity and specificity of future criteria.