Circulating Tumor DNA Measurements to Predict Early Outcome in Diffuse Large B-Cell Lymphomas

Major advances in our understanding and treatment of diffuse large B-cell lymphoma (DLBCL) have come from genomic analysis of this disease. Several recent landmark studies have identified that adding pretreatment genomic information to established clinical prognostic tools in DLBCL such as the International Prognostic Index (IPI) score and interim positron emission tomography (PET) may greatly improve risk stratification.^1-3  However, the connection between the use of this risk stratification to select patients for intensified therapy and improvement of patient outcome is not clear. To this end, a recent study by Dr. David M. Kurtz and colleagues rigorously set out to evaluate whether dynamic and serial measurements of circulating tumor DNA (ctDNA) from plasma of DLBCL patients could serve to predict outcome early during the course of treatment.

In this study, ctDNA from 217 patients with DLBCL or primary mediastinal B-cell lymphoma were assessed in a discovery and in two validation cohorts. The authors sequenced tumor and pretreatment cell-free DNA from plasma in all patients to identify somatic alterations for ctDNA quantitation and disease monitoring. They applied deep sequencing of genes for mutational genotyping and tracking of mutations before and during therapy for predicting event-free survival (EFS) at 24 months as well as overall survival (OS). Specifically, they performed cancer personalized profiling by deep sequencing (CAPP-Seq), a form of targeted deep next-generation sequencing invented by these authors.⁴  Using this method, 99 percent (215 of 217) of patients had at least one tumor-specific alteration detected for monitoring, and 98 percent (212 of 217) had ctDNA prior to therapy. Pretreatment ctDNA was significantly correlated with IPI, total metabolic tumor volume (TMTV), EFS, and OS in patients receiving frontline therapy. Moreover, pretreatment ctDNA was also associated with OS for patients receiving salvage therapy. Finally, in multivariate analysis, pretreatment ctDNA levels remained prognostically important when controlling for IPI, molecular subtype, and TMTV. Overall, these data indicate that ctDNA burden may itself have prognostic value in DLBCL.

One clear benefit of the use of ctDNA for disease monitoring is the ease with which repeated sampling of disease can be performed. Given this, the authors then investigated whether dynamic changes in ctDNA levels might also predict disease response. They densely sampled plasma from the discovery cohort of 14 patients and noted that achieving 1) a two-log (i.e., 100-fold) decrease in ctDNA between pretreatment samples and end of cycle 1 of therapy, or 2) a 2.5-log drop in ctDNA between pretreatment samples and end of cycle 3 of therapy, clearly separated patients who achieved a complete response from others. These same thresholds (defined by the authors as “early molecular response” [EMR] or “major molecular response” [MMR], depending on whether it was assessed at the end of cycle 1 or end of cycle 3, respectively) were confirmed in the two validation cohorts. Moreover, they identified that the line of therapy (frontline vs. salvage) or the type of therapy (R-CHOP [rituximab-cyclophosphamide, doxorubicine, vincristine, and prednisolone] vs. EPOCH-R [etoposide phosphate, prednisone, vincristine sulfate, cyclophosphamide, doxorubicin hydrochloride-rituximab]) did not influence these cut-offs in ctDNA metrics. Finally, EMR and MMR were also predictive of EFS and OS in patients receiving frontline therapy.