Researchers are working to cross translational gaps in cancer imaging biomarker studies, according to a talk given on 5 March at ECR 2026.
Simplified, the three-pronged approach to crossing these gaps considers technical (assay) validation, biological clinical validation, and cost-effectiveness in order to determine reliability before a biomarker can routinely be used in the management of patients with cancer, noted Prof. Fiona Gilbert, pointing to an often cited foundational 2017 paper published in Nature Reviews Clinical Oncology.
Gilbert heads the radiology department at the University of Cambridge in the U.K. During her session, she explained one of the main problems holding back imaging biomarker research.
"The challenges that we found implementing imaging biomarkers in different clinical trials are that you need to have standard operating procedures," Gilbert said, noting that ensuring reproducible image acquisition is one of the first priorities of a robust approach.
Prof. Fiona Gilbert.
In addition, teams must establish a strong quality assurance (QA) program that can be delivered remotely across sites -- and then audit and examine the results of that program. Groups need to agree to an acquisition protocol across sites, she added.
A third key step is training staff, radiographers, and reporting radiologists to ensure consistent results across sites.
Gilbert has been a leading voice in nonrandomized and randomized controlled trials, such as the BRAID trial, which explored breast screening risk-adapted imaging for density, and the Sputnik pulmonary nodule investigation, which compared the cost-effectiveness and accuracy of dynamic contrast-enhanced CT (DCE-CT) with that of F-18 FDG PET-CT. She reflected on lessons learned that are important for imaging biomarker research moving forward.
In the Sputnik study, which took place across 16 U.K. hospitals, PET was delivered by external providers at some sites. While there was agreement on the standard dose patients would receive and quality parameters for standardized acquisition, researchers realized later that there was no consistent central QA looking at acquisition throughout the study. Images were collected at the end, Gilbert said.
"Ensure that your acquisition is as homogeneous as possible and that your reporting protocols are as homogeneous as possible," Gilbert advised, noting that reporting radiologists were on a big learning curve with DCE-CT at that time.
"In terms of safety and, in terms of ensuring that you do perform a good quality trial, then you need to have regular robust, rigorous monitoring from the different sites when you're doing image acquisition and reporting, whatever the biomarker is," Gilbert said.
BRAID compared three new breast screening techniques -- abbreviated MRI, contrast mammography, and whole-breast ultrasound. This study involved 10 U.K. sites.
Reflecting on this study, Gilbert said she learned that performing abbreviated breast MRI was not a simple transition for personnel accustomed to breast MRI. MRI was most challenging for getting consistent image acquisition. Also, recall rates were not monitored throughout the study.
"We obviously recorded them, but we didn't get sites to report them on a regular basis," Gilbert said. "We could have done that." At some sites, contrast mammography produced recall rates of up to 18%, while others had recall rates of 4.3% (within the U.K.-allowed range), she noted. MRI recall rates also varied from 5.2% to 14.7%.
Spend the time, money, and effort to ensure biomarkers are consistent and reproducible, Gilbert said in closing.
