As researchers consider using circulating tumor DNA as an endpoint in clinical trials to evaluate drug efficacy, a collaboration led by Friends of Cancer Research is creating the evidentiary roadmap for the use of ctDNA in regulatory decisions.
As tumor cells, small pieces of DNA are released into the bloodstream, and detectable levels of ctDNA can be used as a novel prognostic indicator of response to a therapeutic regimen, and potentially as an intermediate endpoint that is predictive of survival outcomes.
These research questions are the focus of the multiphase collaboration between Friends, FDA, NCI, clinical researchers, pharmaceutical companies, and diagnostics companies.
The initiative, ctDNA for Monitoring Treatment Response (ctMoniTR) Project, is designed to examine these associations, and provide openly accessible information that would be useful for trialists looking to investigate the utility of ctDNA levels as an endpoint in prospective studies.
The ctMoniTR project seeks to address a specific question: Do changes in ctDNA levels reflect a patient’s response to treatment?
Based on retrospective analyses of data aggregated from clinical trials, the answer is “Yes”—thus far, for non-small cell lung cancer patients treated with PD-1/L1 checkpoint inhibitors, and now, based on new ctMoniTR data, with tyrosine kinase inhibitors.
“When looking at patients who have detectable ctDNA prior to therapy, patients for whom ctDNA remains detectable following therapy, those are the ones that fare worse and don’t seem to be responding as well in terms of long-term outcomes,” Jeff Allen, president and CEO of Friends of Cancer Research, said to The Cancer Letter. “But for patients for whom ctDNA is no longer detectable, those show improved outcomes. Interestingly, that change in ctDNA associated with better outcomes was measured within the first 10 weeks following treatment.”
Friends will be convening a meeting July 11 at Washington Marriott Georgetown in Washington, DC, to present key initial results and discuss the implications of these findings for clinicians and regulators. Richard Pazdur, director of the FDA Oncology Center of Excellence, will be presenting the lunch keynote address.
The registration form is available on the Friends website.
“We will be showing next week that at these early time points, the change of ctDNA to be no longer detectable within the first 10 weeks is associated with improved long-term benefit (PFS and OS). We also looked at whether data from the first RECIST evaluation was an indicator of outcomes—and the ctDNA change was a better indicator of long-term benefit than the first RECIST evaluation was,” Allen said. “When looking at imaging results at a similar early timepoint, grouping those as responders and non-responders by radiographic imaging wasn’t actually an indicator of long-term outcomes. But the change in ctDNA was. It really could be a very useful tool.
“I think the meeting will be a great opportunity to really dive deep into the results. There will also be data from a Baseline ctDNA Project as well as the ctMoniTR Step 2 project that haven’t been presented before,” Allen said. “We’ll have various experts that were involved in and guiding these studies that will be describing the results and thinking about what the next steps should be, including participants from the FDA.”
If these findings are replicated and validated in prospective clinical trials across different disease and therapeutic settings, it is likely that FDA would consider approving drugs with ctDNA as an early endpoint, which could allow a potentially efficacious product to enter the market before confirmatory survival data is obtained.
“ctDNA could be used in early phase clinical trials to aid in signal finding of drug activity and to potentially aid sponsors in their drug development plans,” the agency wrote in a draft guidance published May 2022. “FDA encourages sponsors to develop evidence regarding the usefulness of ctDNA response in addition to or supporting pathologic complete response information after neoadjuvant therapy.”
A series of retrospective studies
The ctMoniTR collaboration, now in its fifth year, is in the “Step 2” phase of the project, which aims to study the association between decreases in ctDNA levels with improved survival outcomes across multiple therapeutic areas and classes of drugs:
- Module 1: Advanced NSCLC with TKI
- Module 2: aNSCLC with anti-PD(L)1 and/or chemotherapy
- Module 3: Solid tumors, including melanoma, breast, head and neck, ovarian, and colorectal, with anti-PD(L)1 or TKI
- Cross-module analysis: Combine data from all modules
The Step 2 Module 1 findings, which were derived from a retrospective aggregate analysis of eight clinical trials of patients with aNSCLC treated with a TKI (i.e., anti-EGFR, ALK, RET, or MET; n=1590), were presented as a poster at the annual meeting of the American Society of Clinical Oncology in early June.
“Patients who had detectable ctDNA levels prior to therapy and then following treatment, if those detectable levels of ctDNA were no longer detectable—so, the ctDNA levels were depleted by therapy—those patients had better outcomes in both progression-free survival and overall survival,” Allen said.
The authors of the TKI study also concluded that ctDNA samples collected within 10 weeks following initial treatment can be used to assess response to treatment and are an indicator of long-term benefit.
In Module 2, the project will include multiple studies with comparator arms evaluating chemotherapy vs. immunotherapy and will be useful to determine whether differences can be predicted based on treatment type.
“We have an impressive statistical operation that is really the foundation of this entire project,” Allen said. “Cancer Research And Biostatistics, or CRAB, serves as the independent analysis center for the project. The NMD Group, led by Névine Zariffa, is a key statistical advisor that we have, and our statistical working group comprised of experts from the data partners and FDA meets biweekly and has been critical for designing the protocol, setting the statistical analysis plan, and evaluating ongoing analyses. This has been a true partnership, and we’re very appreciative of all that have been involved.”
The TKI study builds on the project’s Step 1 findings, which were published August 2022 in JCO Precision Oncology. In Step 1, the investigators observed strong associations between reductions in ctDNA levels from on-treatment liquid biopsies with improved OS (HR, 2.28; p < 0.001) and PFS (HR, 1.76; p < 0.001) in a pooled analysis of five independent clinical trials with NSCLC patients who were treated with immune checkpoint inhibitors.
“Even with the level of variability across different diagnostic assays, and even variability that came with different sampling time points as an example, we were still able to see these strong associations with reduction in ctDNA and improvement in overall survival,” Allen said. “[The Step 1 results] really served as the prototype here showing that 1) these types of data could be aggregated through this unique partnerships, and 2) that ctDNA change may be a biomarker that could be an indicator of response to therapy and has potential to accelerate drug development, and actually be a tool that can help inform regulatory decision making into the future.”
A novel prognostic indicator and endpoint
Beyond drug development, ctDNA has become an important biomarker in routine clinical care for patients with metastatic NSCLC.
In November 2022, results from a large prospective cohort study published in Nature Medicine provided confirmation of the clinical utility of ctDNA in treatment decision-making—demonstrating that patients who were matched to targeted therapy by ctDNA sequencing experienced a nearly 40% reduction in mortality, based on an adjusted multivariate analysis that included disease stage, targeted therapy, tumor volume, and even metabolic tumor volume (The Cancer Letter, Nov. 18, 2022).
Although patients with advanced stages of disease are more likely to shed ctDNA, which is associated with a poor prognosis, there is variability by tumor stage and type.
“In advanced disease, tumor types such as colorectal cancer are positive for ctDNA presence in almost 100% of the cases, whereas this is the case in ~75% of patients with NSCLC or breast cancer,” Ana Vivancos and Josep Tabernero, of the Vall d’Hebron Institute of Oncology, wrote in a companion piece to the Jee et al. paper. “In contrast, only ~30% of patients with metastatic gastrointestinal stromal tumors show detectable ctDNA in plasma.”
These considerations, particularly whether decreasing ctDNA levels on treatment are predictive of response and survival in other cancer types, would need to be explored in greater detail as the ctMoniTR project moves into studying the utility of ctDNA in other solid tumors beyond NSCLC in Module 3.
The upcoming July 11 meeting will also delve into results of the Baseline ctDNA Project that examines pre-treatment ctDNA levels across different major tumor types and stages, as well as explore potential differences between numerous assays.
“The Baseline ctDNA research partnership involves eight different assays and test developers. We’ll be presenting information from five different tumor types,” Allen said. “These data will show that later-stage tumors produce more ctDNA at a more detectable level. And in looking at those late-stage cancers, at least the five cancers that we’ve looked at initially, we’ve seen that there’s been similar detectability and similar ability to measure ctDNA across those different tumor types.
“So, while the initial ctMoniTR data is in lung cancer, the Baseline Project reinforces that ctDNA can be valuable in a number of different tumor types, which really could help us begin to understand its utility across the entire cancer research continuum,” Allen said. “I think some of that will really build on the baseline work to understand the kinetics across numerous different tumor types, how similar or not they actually are. And so, the Baseline project was a useful exercise to begin to understand how different some of these measurements were across late-stage tumor types.”
FDA will be paying particular attention to the limitations of these studies as the agency evaluates the evidence that is generated in the ctMoniTR collaboration.
“I think what regulators will likely want to see moving forward are prospective studies that select the thresholds in advance, in order to not have potentially biased results, to be able to help better understand those outcomes in a prospective manner,” Allen said. “I hope what we’re able to provide by going through the ctMoniTR process is begin to establish some of those methodologies and explore some of those potential thresholds and other study attributes for which a prospective trial could be designed.”
A threshold that needs to be examined in the future, for example, pertains to the magnitude of change in ctDNA level, and at which cutoff points would those changes be considered to be clinically meaningful.
“What we looked at here was a percent change from baseline,” Allen said. “So, we were able to normalize the variability between the different assays. The different assays that were used do have different characteristics, a different limit of detection as an example, or they may have different sensitivity or specificity, so we needed a common metric in order to be able to combine data from multiple trials.
“So, each patient always received the same test, even if each trial did not necessarily utilize the same tests across the multiple different trials in that cohort, but by looking at a percent change of variant allele frequency, we’re able to normalize those values.”
More studies are needed before ctDNA response rates can be reliably used as a regulatory endpoint, but for now, there is sufficient evidence to support the use of ctDNA as a selection criterion.
“By selecting patients who had detectable levels of ctDNA at the outset of a trial, you could construct a clinical trial that would go a little bit faster, at least for obtaining an initial indicator of drug activity,” Allen said. “Ultimately, when that evidence does get brought before the regulators, hopefully they’ve had the opportunity to have access to this information and ask questions of the data along the way, to help inform the questions that they may ask as the prospective trials are being designed.
“I hope that we are providing robust evidence that is beginning to show that ctDNA change is a viable biomarker for evaluating potential treatment efficacy.”