Abstract LB297: Analytical validation of a tissue-free, multi-cancer, post-diagnosis cancer research test that uses cell-free DNA methylation profiling

Abstract

Introduction: Cancer detection blood tests have shown clinical utility after cancer diagnosis, but many evaluate only single cancers and require tumor tissue, thereby limiting their utility. We developed a versatile, tissue-free (ie, no tumor tissue required), multi-cancer detection test (“Post-Diagnosis Cancer Research Solution”) based on methylation sequencing of cfDNA from blood. This technology solution can be used to evaluate cfDNA applications in cancer research, including treatment evaluation, recurrence monitoring, and prognostic guidance. We report the analytical validation of this Post-Diagnosis Cancer Research Solution, characterizing sensitivity, specificity, precision, and input range. Methods: cfDNA samples from cancer and non-cancer donors were analyzed. Analytical sensitivity (LoD95, limit of detection with ≥95% probability) was determined as a function of methyl variant allele fraction (MVAF), a measure of circulating tumor allele fraction. A total of 6 ng of cfDNA was used (cancer cfDNA titrated into a background of non-cancer cfDNA). LoD95 was defined as the lowest observed MVAF with ≥95% cancer signal detection across tested replicates or was estimated using probit regression for eligible samples. Analytical specificity was the rate of non-cancer classification among samples from 128 non-cancer donors. Precision (defined as concordance with the expected cancer/non-cancer result) was evaluated for 15 cancer donors tested near sample LoD95 and 8 non-cancer donors, with ≥18 replicates/donor. Reliability of classification was evaluated as a function of cfDNA input mass (0.25-100 ng total cfDNA, 14 cancer and 2 non-cancer donors). Results: A total of 12 different solid cancer types from 22 individuals with cancer were assessed in the LoD95 analysis. Median LoD95 at 6 ng total cfDNA input was 0.023% (10th percentile 0.0037%; 90th percentile 0.04%) MVAF. LoD95 estimates from in silico titration analyses of >200 clinical samples across a subset of the 12 cancer types were consistent with experimental LoD95 values. Analytical specificity was determined to be 98.47% (95% CI: 94.60-99.58%). Median precision across individuals was 100% (10th percentile 87%; 90th percentile 100%). Classification performance was accurate and consistent across a wide range of cfDNA input mass (100% correct cancer/non-cancer classification from 1.5-100 ng). Conclusions: Results demonstrate that this multi-cancer Post-Diagnosis Cancer Research Solution has high analytical sensitivity, specificity, and precision, with reliable performance across a broad cfDNA input range. This technology does not require a tumor sample and provides a cancer signal estimate in terms of MVAF. These features may enable understanding of cfDNA dynamics in a wide variety of cancers for research studies. Citation Format: Mohini Desai, Svetlana Rakhmanova Shchegrov, Shoujie Chai, Yifan Zhou, Tracy Nguyen, Yirang Cho, Collin Melton, Eric Scott, Manami Roychowdhury-Saha, Pei-Yun Chang, Rita Shaknovich, Byoungsok Jung. Analytical validation of a tissue-free, multi-cancer, post-diagnosis cancer research test that uses cell-free DNA methylation profiling [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 2 (Clinical Trials and Late-Breaking Research); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(8_Suppl):Abstract nr LB297.

A Novel Tissue-Free Method to Estimate Tumor-Derived Cell-Free DNA Quantity Using Tumor Methylation Patterns

“…Here, TMeF linearity upon dilution (to support assessment of TMeF accuracy at low ctDNA levels) was determined by in silico analysis. Although TMeF linearity was consistent with previously reported results from in vitro dilution assays that used TMeF to assess the limit of detection of a post-diagnosis cancer detection test [59], future validation of TMeF with in vitro cfDNA dilutions will be performed to more fully assess its linearity. It should also be noted that there are epigenetic differences within the cancer label groupings used in this study.…”

“…As a sensitive tissue-free measure of ctDNA abundance, TMeF could act as a prognostic marker in the MRD setting. In addition to clinical applications, TMeF is a potential metric for analytical validation studies, due to its ease of implementation and accuracy, specifically to assess the abundance of an analyte in sensitivity (limit of detection) and specificity studies [ 59 ]. Studies investigating the use of TMeF in these clinical applications are underway.…”

Using cfDNA and ctDNA as Oncologic Markers: A Path to Clinical Validation

“…Almost every tumor type is characterized by progressive CpGisland-specific hypermethylation and global CpG hypomethylation [94]. The GRAIL test uses whole genome bisulfite sequencing (WGBS) targeted at over 100,000 methylation regions combined with machine learning to detect cancer and predict tissue of origin (TOO) localization detecting ctDNA down to 0.023% (~1/5000 copies) [95,96]. The OverC test uses an altered WGBS technique called enhanced linear splinter amplification sequencing (ELSAseq), which mitigates the damage to DNA caused by bisulfite treatment and is able to detect ctDNA down to 0.02% (1/5000) [97,98].…”

“…Sensitivity was lower for cancers in stage 1 being 18% rising to 93% for cancers in stage 4. Specificity for both sets was greater than 99% [96]. Towards further clinical validation, a study of 4077 participants, with 2823 patients with cancer, and 1254 control patients, were tested with the refined Galleri WGBS assay and will continue to follow-up for 5 years.…”