Diagnosis of primary CNS lymphoma through molecular and immune biomarkers in cerebrospinal fluid (PRICELUS): a multicentre, prospective, cohort study

Background Primary CNS lymphoma is typically diagnosed by brain biopsy, a procedure with risks and diagnostic delay. Emerging cerebrospinal fluid (CSF) biomarkers, including the myeloid differentiation primary response 88 (MYD88) L265P mutation in CSF and several chemokines, might improve diagnostic accuracy, but evidence mostly comes from retrospective, non-consecutive cohorts. We aimed to validate the diagnostic utility of CSF biomarkers in patients with suspected primary CNS lymphoma. Methods Patients with suspected primary CNS lymphoma (aged ≥18 years) who underwent CSF analysis were enrolled in five hospitals in the Netherlands between March 1, 2021, and May 1, 2025. Analyses focused on MYD88 and chemokines with reported diagnostic value. A diagnostic panel was constructed with multivariable logistic regression and least absolute shrinkage and selection operator (LASSO) regression. We performed decision tree analysis to evaluate performance. Findings Of 255 patients (median age 61 years [IQR 52–72]; 171 [67%] males and 84 [33%] females), 50 (20%) were diagnosed with CNS lymphoma; 44 had primary CNS lymphoma and six were diagnosed with secondary CNS lymphoma. In univariable analyses, six of seven chemokines showed significant diagnostic value (area under the curve 0·67–0·96). MYD88 was detected in 28 (56%) of 50 patients with CNS lymphoma, with a sensitivity of 56·0% (95% CI 41·3–70·0) and specificity of 98·5% (198 of 201; 95% CI 95·7–99·7). In the multivariable logistic regression model, MYD88, interleukin-10 (IL-10), and C-X-C motif chemokine ligand 13 (CXCL13) remained independently associated with a CNS lymphoma diagnosis. Combining a positive MYD88 result with elevated IL-10 or CXCL13 (or both) yielded a sensitivity of 54·0% (27 of 50; 95% CI 39·3–68·2) with specificity of 100·0% (95% CI 98·2–100·0). A negative result for all three biomarkers provided a negative predictive value of 99·4% (162 of 163). Interpretation In this prospective, multicentre cohort of patients with suspected primary CNS lymphoma, we validated the high diagnostic value of MYD88, IL-10, and CXCL13. Results should be interpreted within the full clinical context and with caution, as the diagnostic algorithm and decision tree were derived within the same cohort. In selected clinical scenarios and with further validation of these results, combined CSF biomarker testing might support non-invasive diagnostic decision making and reduce reliance on brain biopsy. Funding Stichting Annie van Koeverden.