The PRICELUS study: CSF biomarkers for PCNSL diagnosis

Results from the multicenter, prospective PRICELUS cohort study evaluating cerebrospinal fluid (CSF) biomarkers for the diagnosis of primary central nervous system lymphoma (PCNSL) in 255 patients with suspected PCNSL were published in The Lancet Haematology by van Rooij et al. The primary endpoint was diagnostic accuracy of CSF biomarkers. 

Key data: In total, 205 patients had non-CNS lymphoma and 50 patients had CNS lymphoma (PCNSL, n = 44; secondary CNS lymphoma, n = 6). In univariable logistic regression, seven of eight biomarkers were associated with CNS lymphoma: MYD88 (p < 0.0001), interleukin-10 (IL-10; p < 0.0001), C-X-C motif chemokine ligand (CXCL)13 (p < 0.0001), programmed cell death protein 1 (PD-1; p < 0.0001), FAS ligand (FASL; p < 0.0001), CXCL10 (p < 0.0001), and CD27 (p = 0.0030). In multivariable analysis, MYD88 (p < 0.0001) and IL-10 (p < 0.0001) maintained independent association, while CXCL13 demonstrated borderline association (p = 0.0810). The initial diagnostic model, including MYD88, IL-10, and CXCL13, achieved an accuracy of 92.5% (cross-validated accuracy, 89.9%), with 94.0% sensitivity and 92.1% specificity. The final decision-tree model, which allowed CNS lymphoma to be diagnosed in patients with MYD88-positivity and either IL-10 >2.0 pg/mL or CXCL13 >90.0 pg/mL, correctly classified 54% of CNS lymphoma cases, with a sensitivity and specificity of 54.0% and 100%, respectively. A negative result for all three biomarkers yielded a negative predictive value (NPV) of 99.4%.

Key learning: In patients with suspected PCNSL, a three-biomarker CSF panel including MYD88, IL-10, and CXCL13 provided diagnostic value, potentially enabling less invasive diagnosis; however, the diagnostic algorithm requires external validation.