External validation of prediction scores of positive random skin biopsy in suspected intravascular large B-cell lymphoma

Background: Intravascular large B-cell lymphoma (IVLBCL), a rare but increasingly recognized aggressive extranodal B-cell lymphoma, often presents with diagnostically challenging manifestations. Three prediction scores—the Kameda, Okayama, and St. Marianna scores—have recently been proposed to guide the use of random skin biopsy (RSB) for diagnosing IVLBCL, but none have undergone external validation. Methods: We retrospectively included 86 consecutive patients aged ≥15 years with suspected IVLBCL who underwent RSB between May 2013 and June 2024. The outcome of interest was a positive RSB for IVLBCL lesions. We evaluated the 2-risk-group classification measures and the 3-risk-group model performance. Results: Fourteen patients (16%) had a positive RSB. Although the St. Marianna score had the best 3-risk-group model performance, both the Okayama and St. Marianna scores achieved 100% sensitivity and negative predictive value in the 2-risk-group classification. The Kameda score rarely stratified patients as high risk, resulting in limited predictive performance. In settings where missed diagnoses are considered more harmful than unnecessary biopsies, the 2-risk-group classification and the 3-risk-group model of each score performed similarly. Conclusion: The St. Marianna score or Okayama scores may help reducing unnecessary biopsies without missing IVLBCL diagnoses in settings similar to our cohort. These results should be validated locally before routine clinical use.