Pre-treatment circulating tumor DNA predicts survival of diffuse large B-cell lymphoma patients and allows genetic and biological evaluation

Abstract

Abstract Introduction. Circulating tumor DNA (ctDNA) was identified as a powerful biomarker for treatment response evaluation and outcome prediction in malignant tumors, including diffuse large B-cell lymphoma (DLBCL). However, more studies are necessary to validate ctDNA utilization in general, and specifically in a real-world setting to facilitate standardization and transition into routine clinical practice. Therefore, we assessed pre-treatment ctDNA in 169 unselected previously untreated DLBCL patients consequently uniformly treated with R-CHOP chemoimmunotherapy at five academic hematology centers in the Czech Republic. Methods. ctDNA was analyzed by CAPP-Seq (CAncer Personalized Profiling by deep Sequencing) and a custom panel of 521 genes. DNA alterations were identified by VarScan2. ctDNA concentration was calculated from average variant allele frequencies and reported as human haploid genome equivalents (hGE) per ml of plasma. Germinal center B-cell like (GCB) and non-GCB subtyping was done by immunohistochemistry. Genetic classes were analyzed by LymphGen 2.0, COSMIC (Catalogue Of Somatic Mutations In Cancer) signatures by SigProfiler. Results. In all patients, the median age at diagnosis was 64 years (range 24-81), clinical stage III–IV in 68%, more than one extranodal involved sites in 40%, PS ECOG 2–4 in 24%, elevated LDH in 60%, International Prognostic Index (IPI) 3–5 in 50%, bulky disease over > 7.5 cm in 40%, and non-GCB DLBCL subtype in 47% of patients. Median follow-up was 2.73 years. The 2-year progression free and overall survivals (PFS and OS) were 73.6% and 84.3%, respectively. Without any threshold for input DNA quality and quantity, ctDNA was detected in 136 out of 169 analyzed patients (80%). Clinical characteristics of patients with and without detected ctDNA did not show any differences. The median plasma ctDNA concentration was 995 hGE/ml. ctDNA concentration was significantly higher in patients with clinical stage III–IV (median 1416.9 vs. 542.7 hGE/ml, p = 2.62e-5), age > 60 years (median 1110.7 vs. 691.2, p = 0.04), PS ECOG 2–4 (median 2130.9 vs. 824.2 hGE/ml, p = 2.62e-5), elevated LDH serum levels (median 1804.1 vs. 513.3 hGE/ml, p = 1.14e-9), bulky disease over > 7.5 cm (median 2234.9 vs. 624.8 hGE/ml, p = 4.58e-6), and in patients with IPI 3-5 (median 1646.7 vs. 631.7 hGE/ml, p = 8.33e-5). To find a pre-treatment ctDNA plasma concentration that would identify high risk patients (based on PFS analysis), patients from the General University Hospital (n = 72) were used as a discovery cohort and patients from other centers (n = 64) as a validation cohort (clinically well-balanced). In the discovery cohort, the threshold was determined at 3000 hGE/ml using Harrell's C-index. High ctDNA (> 3000 hGE/ml, 32% of patients) was associated with inferior PFS (2-year PFS 46% vs. 88%, 5-year PFS 38% vs. 71%, HR 3.99, p = 0.001). In the validation cohort, ctDNA concentration > 3000 hGE/ml (19% of patients) was also associated with inferior PFS (2-year PFS 50% vs. 78%, 5-year PFS 40% vs. 67%, HR 2.9, p = 0.024). It confirmed identified threshold that was consequently used for further analyses. In all patients, high ctDNA level (> 3000 hGE/ml, 26% of patients) was associated with inferior PFS (2-year PFS 48% vs. 83%, 5-year PFS 38% vs. 66%, HR 3.37, p < 0.001), and, importantly, also with inferior OS (2-year OS 74% vs. 88%, 5-year OS 54% vs. 75%, HR 2.56, p = 0.016). Moreover, ctDNA was an independent prognostic factor for PFS in multivariate analysis with IPI score (high ctDNA, HR 2.12, p = 0.029; IPI ordinal, HR 1.42, p = 0.01). High ctDNA patients had more advanced clinical stage, worse PS ECOG, higher LDH levels, higher IPI score (all with p < 0.001), and higher proportion of bulky disease (p = 0.005). The LymphGen assigned a genetic subtype to 44% of cases, reflecting the capability of the LympGen model to classify 50-60% of tumors (if copy number and translocation information is available). COSMIC mutational signatures showed differences between genetic and cell of origin subtypes, e.g., activation induced cytidine deaminase activity specifically in non-GCB DLBCL. Conclusion. Our study confirmed pre-treatment ctDNA concentration as a strong and independent risk factor associated with unfavorable DLBCL survival in a routine clinical setting, allowing parallel genetic and biological evaluation. First two authors contributed equally. Supported by NU21-03-00411, DRO-VFN00064165, LX22NPO5102, and SVV 260637.