JCO Article Insights: ctDNA in DLBCL - Ready for Prime Time?

In this episode of JCO Article Insights, host Dr.  Ash Gurumurthi summarizes JCO articles, "Phased Variant–Supported Circulating Tumor DNA as a Prognostic Biomarker After First-Line Treatment in Large B-Cell Lymphoma: Findings From the DIRECT Study" and " Prospective Validation of Circulating Tumor DNA Measurable Residual Disease After First-Line Therapy in Large B-Cell Lymphoma"

TRANSCRIPT

Ash Gurumurthi: Hi and welcome to JCO Article Insights. I'm your host, Ash Gurumurthi, and today we will be discussing two articles, both published in the Journal of Clinical Oncology, on the real-world utility of circulating tumor DNA (ctDNA) MRD in newly diagnosed large B-cell lymphoma.

The first study is the article "Phased-Variant-Supported Circulating Tumor DNA as a Prognostic Biomarker After First-Line Treatment in Large B-Cell Lymphoma: Findings From the DIRECT Study" by Dr. Joanna Krupka and colleagues in the United Kingdom. For the sake of convenience, I'll refer to this as the DIRECT study. The second study is "The Prospective Validation of Circulating Tumor DNA Measurable Residual Disease After First-Line Therapy in Large B-Cell Lymphoma" by Dr. Steven Wang and colleagues in the Netherlands, referred to as the HOVON 902 study.

By way of background, I wanted to talk about MRD in hematolymphoid malignancies. Nodal diseases have lacked a robust biomarker for end-of-treatment response. They have relied historically on PET scans interpreted using the semiquantitative Deauville 5-point scale, which has a high negative predictive value but a limited positive predictive value. The poor positive predictive value for survival results in extended follow-up with serial imaging for risk stratification with unnecessary and invasive biopsies.

There have been recent revolutionary advancements in ctDNA MRD in B-cell lymphoma. The use of ctDNA in lymphoma began with CAPP-seq, which tracked single nucleotide variants that were tumor specific but was limited by excessive background sequencing noise with false negatives. To overcome this, Dr. Kurtz and colleagues developed the proprietary PhasED-seq assay. This tracks well-recognized phased mutations on the same DNA strand in cis configuration within hypermutated regions that are unique to B-cell lymphoma. Using this method, they pushed their limit of detection at 95%, the so-called LOD95, to 0.7 parts per million under optimal circumstances with 120 nanograms of input cell-free DNA from plasma.

Based on the use of the PhasED-seq assay in trials of newly diagnosed large B-cell lymphoma with the use of investigational agents, the NCCN currently recommends consideration of ctDNA MRD assay with a detection limit of less than 1 part per million if biopsy is not feasible for a positive end-of-treatment PET. However, I believe this threshold needs reconsideration given it is based on an ideal assay LOD95 under optimal circumstances rather than sample-specific LOD95. Real-world validation of the role of end-of-treatment ctDNA and appropriate thresholds for sample-specific LOD95 were lacking until the publication of these two studies.

The DIRECT and the HOVON 902 studies were multicenter, prospective trials using real-world cohorts of newly diagnosed large B-cell lymphoma treated with standard anthracycline immunochemotherapy, ie, R-CHOP chemotherapy. They validated end-of-treatment ctDNA MRD response measured on a phased-variant platform and found them to be strongly prognostic for relapse and survival. This was independent of PET imaging or baseline clinical prognostication like the International Prognostication Index, the IPI. They also demonstrated a threshold with an LOD95 of approximately 1 in 100,000 is necessary for clinical utility.

Both trials recruited over a similar period between 2020 to 2023, with the DIRECT study conducted within the National Health Service in the United Kingdom and the HOVON 902 as a national study in the Netherlands. For survival analysis, only patients who reached the landmark event of end of treatment with an available ctDNA MRD sample without progressive disease or death at that time point were included. These studies evaluated similar-sized cohorts with 134 patients for HOVON 902 and 151 patients for the DIRECT study. As expected, their baseline demographics are reflective of a real-world population of newly diagnosed cases with large B-cell lymphoma.

Although both used comparable statistical methodologies with time-to-event analysis, the primary outcomes vary, making headline comparisons quite challenging. The DIRECT study utilized the time to tumor progression, censoring death unrelated to disease. This was done to isolate the molecular impact of detectable ctDNA at the end of treatment. In contrast, the HOVON 902 study used progression-free survival, which counts all-cause mortality as an event. This naturally results in lower event-free rates for PFS compared to TTP in the DIRECT study.

The trials differed in their choice of phased-variant platforms, with the DIRECT study developing an independent, fully open-source phased-variant ctDNA assay. This has been released on GitHub. In contrast, the HOVON 902 study utilized PhasED-seq by Foresight Diagnostics, which is currently the only proprietary and commercially available phased-variant assay for lymphoid malignancies.

Interestingly, despite the differences in platforms and the primary end points, the results were remarkably consistent. The DIRECT study found a highly significant difference in the 2-year TTP rate of 96% in those with undetectable ctDNA MRD at the end of treatment compared to 45% in those with detectable ctDNA, with a hazard ratio of 15. Similarly, the HOVON 902 study found a significantly superior 3-year PFS of 85% in those with undetectable ctDNA compared to 17% with detectable ctDNA, with a hazard ratio of 10.

Crucially, both studies found end-of-treatment ctDNA MRD significantly outperformed PET response assessment for long-term PFS. In fact, for the end point of PFS in both trials, the baseline IPI lost all statistical significance in both univariate and multivariable analysis when accounting for ctDNA MRD and PET status at the end of treatment.

While both studies demonstrate the superiority of ctDNA MRD compared to PET in predicting survival, interestingly, the combination of both tests appeared to be complementary in identifying the highest-risk group. The HOVON 902 study identified 13 patients who were double positive, ie, they were positive with end-of-treatment PET and detectable ctDNA MRD. Every single one of these patients progressed over a 3-year period with a dismal overall survival of 17%. The DIRECT study mirrored these findings with the same double-positive group having a 2-year time to progression rate of 23%. Given consistency in identifying the poor outcome of this double-positive population in both studies, this is clearly a group that would benefit from trial-based approaches like consolidation or, alternatively, frequent surveillance for clinical relapse.

On the other hand, the best-performing group was the double negative, ie, those who had achieved PET negative and ctDNA undetectable at the end of treatment. The double-negative group had a 2-year time to progression of 97% in the DIRECT study and a 3-year PFS of 88% in the HOVON 902 trial. This is quite impressive. Based on these findings, we can anticipate that ctDNA may complement rather than wholly replace PET at the end of treatment for response assessment.

Perhaps the most critical finding from both studies challenged current NCCN-recommended ctDNA MRD sensitivity threshold of achieving less than one part per million. While phased-variant assays can theoretically detect this, this is under optimal conditions, specifically 120 nanograms of input cell-free DNA. In both trials, only 3% of samples could achieve this sensitivity, with the vast majority limited to a sample-specific LOD95 of approximately 1 in 100,000 informative reads. The primary constraint was simply limited plasma volume collected, a denominator problem of input cell-free DNA. For example, the HOVON 902 study had a median plasma volume of 5 mL, yielding 20 nanograms of input DNA. The DIRECT study elegantly demonstrated bridging the gap to attain the NCCN standard of LOD95 of less than 1 part per million is practically impossible. This would require greater input DNA, attained through a 20- to 30-milliliter collection of plasma rather than the standard 10 milliliters, and a massive 20- to 40-fold increase in sequencing depth.

With the current real-world sensitivity of roughly 1 in 100,000 in both these studies, the negative predictive value is already nearly at 90%. There is going to be diminishing returns for further analytical sensitivity. This strongly suggests that the NCCN guidelines should be updated to prioritize achievable sample-specific LOD95 rather than assay-specific theoretical limits.

Collectively, these studies validate the real-world utility of ctDNA MRD as an independent predictor of long-term outcomes following first-line therapy of large B-cell lymphoma. Finally, after two decades of the default R-CHOP for all, the field of aggressive large B-cell lymphoma is taking leaps and bounds by integrating ctDNA MRD with the current wave of bispecific and cellular therapies.

I want to now leave you with my five key clinical takeaways from both these studies.

●        Firstly, ctDNA MRD is a more potent independent predictor of outcome than end-of-treatment PET/CT and baseline IPI.

●        Second, ctDNA MRD in first-line large B-cell lymphoma is already reshaping clinical trial space with therapeutic escalation and de-escalation strategies based on ctDNA kinetics during treatment, as well as identifying candidates with persistent ctDNA at the end of treatment for consolidation approaches.

●        Thirdly, this technology is ready for prime time. Whether this is through Foresight's PhasED-seq assays or the open-source method released by the DIRECT group, academic centers can now operationalize this in routine clinical care.

●        Fourth, biology clearly provides a ceiling. Current sensitivity goals of less than one part per million as recommended by the NCCN are limited by the actual amount of cell-free DNA we can extract from a patient's blood, not just the assay's technology. I believe these two studies will inform the NCCN's next revision to move away from theoretical assay limits to a more realistic sample-specific LOD95 of approximately 1 in 100,000.

●        Finally, it appears that the end-of-treatment ctDNA MRD test may be complementary to PET/CT rather than a replacement. Clearly, the best outcomes are seen in double-negative patients, while double-positive results, ie, positive end-of-treatment PET and detectable ctDNA at the end of treatment, identify a group with an extremely high risk of early progression who may need early intervention.

Thank you for listening to JCO Article Insights. Please come back for more interviews and article summaries, and be sure to leave us a rating and review so others can find our show. For more podcasts and episodes from ASCO, please visit asco.org/podcasts.

The purpose of this podcast is to educate and to inform. This is not a substitute for professional medical care and is not intended for use in the diagnosis or treatment of individual conditions.

Guests on this podcast express their own opinions, experience, and conclusions. Guest statements on the podcast do not express the opinions of ASCO. The mention of any product, service, organization, activity, or therapy should not be construed as an ASCO endorsement.