Use of ctDNA as a surrogate in early clinical drug development – nearing its end point?

For more than two decades, drug licensing in the US and Europe has permitted the use of surrogate endpoints in the approval of new therapies, largely based on biomarker discovery. Compared to randomised controlled clinical studies assessing final participant outcome, the use of surrogates in early clinical trial designs as a proxy and predictor of survival has undoubtedly allowed for smaller sample sizes, shortened patient follow-up times and helped to reduce financial costs of these studies, While these efforts have accelerated drug discovery efforts and regulatory approvals for some agents such as selpercatinib in patients with NSCLC and RET fusion (Clin Cancer Res. 2023 Sep 15;29(18):3573-3578), and the NTRK inhibitors entrectinib and larotrectinib in solid tumours with NTRK fusions, thus helping expedite patient access to innovative therapies, all is not rosy. The predictability of response criteria remains at the centre of controversies in drug development as discussed in a Special Session at the ESMO Targeted Anticancer Therapies Congress 2024. In fact, numerous reviews and meta-analyses continue to call into question the reliability of classic surrogate endpoints in predicting patient survival or quality of life to a given therapeutic strategy or intervention in the current era of precision oncology.

Led by colleagues at the Knight Cancer Institute, Portland, US, results of a retrospective review of cancer drugs approved by the US  Food and Drug Administration (FDA) from January 1992 to July 2019, showed that 1 in 3 cancer drug approvals used a surrogate endpoint - either response rate or progression-free survival (PFS) - for the first time for a specific cancer type (JAMA Intern Med. 2020;180(6):912-914). The investigators also reported that 61% of FDA approvals had no documented correlation between surrogate endpoint and overall survival (OS), and 16% had a poor correlation.

While such findings flag fundamental questions regarding the use of surrogates as endpoints in clinical trials, many approvals in recent years have been based on early response criteria such as the Response Evaluation Criteria in Solid Tumors (RECIST)-based endpoints including PFS. Longer follow-up requirements, subsequent treatment and cross over in clinical trial design make OS a hard endpoint to meet, and the need to validate and find new surrogates of response remains a critical need in the field.

Promising surrogates to strongly predict clinical outcomes

Treatment response assessment is an evolving field, also spurred by advances in predictive cancer science coupled with the development of increasingly powerful technologies and highly sensitive, less invasive techniques including liquid biopsy.

As history shows, determining the quality of a surrogate marker can prove difficult and controversial. The apparent heterogeneity in methods used to assess surrogacy underpins the importance of developing clear recommendations about approaches and reporting for surrogacy validation. But, as I consider the array of novel candidate surrogates in early clinical drug development, I firmly believe that we are on the cusp of change.

Moving the field forward, we must collectively address myriad aspects and challenges relating to the development and validation stages prior to their application and integration into clinical practice.

A review article led by colleagues at the Sarah Cannon Research Institute, Nashville, US, updates on the utility of dynamic changes in circulating tumour DNA (ctDNA) by liquid biopsy to predict solid tumour response to therapeutic regimens in oncology (Ann Oncol. 2023 Dec 23:S0923-7534(23)05114-1). The authors highlight the limitations of current evaluation of treatment response in solid tumours guided by dynamic changes in tumour bulk measured by imaging. In their proposed plan towards the standardisation of Liquid Biopsy RECIST (LB-RECIST), they address several key issues that will certainly need to be resolved before recommending the introduction of ctDNA response criteria in daily clinical practice. Among many others, these include the standardisation and harmonisation of detection methods as well as reaching consensus on how to define ctDNA response and/or disease progression in earlier settings for precision oncology.

In their expert review, the investigators call for a better understanding of which ctDNA source should be considered preferable over another in a given clinical setting, expanding ctDNA changes as a surrogate for therapeutic response or progression to earlier disease settings, defining timing of response assessment as well as that of comparative imaging studies, addressing potential situations where ctDNA responses might contradict radiological responses through head-to-head comparisons between radiological response criteria and ctDNA response criteria in later studies, as well as other technological, clinical, and regulatory considerations. As concluding remarks, they state that studies aiming to develop and validate ctDNA response criteria are crucial and needed now more than ever. I couldn’t agree more.

The serial sampling of ctDNA

So, are we close to reach an end point in the use of ctDNA as a surrogate endpoint? Not yet, but I do believe that liquid biopsy will continue to prove its worth as a more robust treatment response biomarker and perhaps take centre stage as a complementary tool to potentiate current surrogates in early phase drug development, as well as existing response criteria. However, it is critical to remember that a proper validation is still essential as a next step.

Only by working together with all stakeholders in oncology will we continue to make important progress in developing novel early response biomarkers in innovative clinical trials. These determined efforts will help us to provide our patients with easier, standardised cancer monitoring via minimally invasive approaches, minimising exposure to potentially ineffective therapies and associated side-effects in patients who are not responding to treatment, and potentially permitting an earlier cross-over to alternative therapeutic strategies.

Several early-stage clinical trials have already included ctDNA molecular response data in their designs. The introduction of ctDNA into response assessment in solid tumours, which will most likely be based on a tumour and treatment-dependent context, is now starting to be a reality. The onus is now on us to pave the way for its validation as response criteria, and the integration of this promising approach towards improving outcomes for our patients.