ESMO Asia 2020: Early Detection and Localization of Cancer Using a Blood-Based Assay

Get Permission

The ELSA-seq assay was able to detect diverse cancer types at early stages with high specificity and was able to provide information regarding the tissue of origin, according to findings presented by Gao et al at the ESMO Asia Virtual Congress 2020 (Abstract LBA3).

Qiang Gao, MD, PhD, of the Liver Cancer Institute, Zhongshan Hospital, Fudan University in Shanghai, discussed the underlying rationale for the prospective, multicenter, longitudinal PREDICT study, which aimed to identify multiple cancers noninvasively at early stages. According to Dr. Gao, early detection of cancer offers clinical benefits, particularly for cancer types that do not have effective screening methods.

He further explained that a pilot project, the THUNDER study was designed to develop and validate ELSA-seq, a sensitive targeted methylation sequencing assay that detects epigenetic alterations from circulating cell-free DNA.

He presented findings from the second THUNDER substudy (THUNDER-II) that focused on detecting malignancies in the liver, colon/rectum, esophagus, pancreas, lung, and ovary.


The THUNDER-II substudy was composed of four independent steps: marker discovery, model training, validation, and single-blind test. The investigators combined data generated in-house and from public sources to design a targeted methylation panel.

The substudy analyzed 625 patients with cancer and 483 noncancer control patients who were categorized into a training set made up of 274 patients with cancer and 195 controls, or an independent validation set with 351 patients with cancer and 288 controls.

The patients with cancer and noncancer control patients were generally matched with respect to age, sex, and smoking status.


  • At 99.5% training specificity, the cross-validated sensitivity of the assay was 79.9%.
  • The classifier returned a tissue-of-origin result in 98.6% of cases, and 81.0% of these predictions were correct.

Among patients with cancer, 79.5% had been diagnosed with cancer at early stages (I–III).

Efficacy of the Assay

Regarding the assay, at 99.5% training specificity (95% confidence interval [CI] = 96.7–100), the cross-validated sensitivity was 79.9% (95% CI = 74.6–84.4). These results were consistent in the validation set, which demonstrated 98.3% specificity (95% CI = 95.8–99.4) and 80.6% (76.0–84.6) sensitivity across disease stages and the diverse cancer types.

In terms of tracking the location of the malignancies, the classifier returned a tissue-of-origin result in 98.6% of cases, and 81.0% (95% CI = 77.2–84.3) of these predictions were correct.

According to the authors, results from the THUNDER-II study demonstrated that early cancer signals could be identified by ELSA-seq with high specificity. This method also enabled accurate information on the tissue of origin, which may provide guidance for subsequent diagnostic workup. Further independent validation results of the THUNDER-II will be reported in the near future. 

Based on these findings, the authors suggest potential implementation of this sensitive and robust assay as a multicancer detection test.

Disclosure: This study was funded by Burning Rock Biotech. For full disclosures of the study authors, visit