Clinical and radiological prediction models for random skin biopsy positivity in suspected intravascular large B-cell lymphoma: development and internal validation

Random skin biopsy (RSB) is essential for diagnosing intravascular large B-cell lymphoma (IVLBCL), but previous clinical prediction models have shown limited performance. This study developed and internally validated predictive models for RSB positivity using data from a single center with 181 Japanese patients who underwent RSB and positron emission tomography-computed tomography (PET-CT) between 2006 and 2024. Three models were evaluated: clinical (lactate dehydrogenase [LDH] and hypoxia), PET-CT (maximum standardized uptake values [SUVmax] of the lymph nodes, lungs, and adrenal glands), and combined (LDH and SUVmax of the lymph nodes, lungs, and nose). The combined model showed superior discriminative power compared to the clinical model. All models demonstrated good calibration. Decision curve analysis revealed higher net benefits for PET-CT and combined models. These findings suggest incorporating PET-CT improves patient selection for RSB in the diagnosis of IVLBCL, and external validation in larger cohorts is needed.