A new, multi-target stool DNA methylation assay accurately detected and identified the tissue of origin of multiple gastrointestinal (GI) cancers. The research will be presented at the 2023 American Society of Clinical Oncology (ASCO) Breakthrough Meeting, in Yokohama, Japan.

According to the authors, GI cancers account for nearly one-third of the world’s cancer burden. While there are some non-invasive tests for the early detection of colorectal cancer, there are no reliable non-invasive tests for the early detection of other GI cancers.

“Stool is a promising sample for GI cancer detection because it contains the host’s exfoliated cells and circulating-free DNA derived from GI cancer cells. Our study aims to develop a non-invasive, multi-target stool DNA methylation test for the early detection and localisation of GI cancers,” said lead study author Li-Yue Sun, MD, Second Department of Oncology, Guangdong Second Provincial General Hospital.

The study, conducted in China, prospectively enrolled 124 patients (68 males and 56 females) diagnosed with GI cancer (14 patients had stage II cancers; 27 had stage III cancers; and 81 patients had stage IV cancers) who had not received treatment and 92 healthy patients who underwent GI cancer-related examinations every 3 years. The median age of the study participants was 60 years for the patients with GI cancers and 57 years for healthy patients.

Each participant’s stool sample was analysed for DNA, which was then put through methylation analysis. In the DNA of tumours, the pattern of methylation looks different because of changes to the healthy cells. This pattern acts as a signature that can indicate the presence of cancer and where the cancer is located.

Using the methylation analysis, the researchers were able to accurately identify the presence and location of cancer in 79% of the patients already diagnosed with GI cancers. The researchers then combined the methylation analysis with machine learning using multiple logistic regression models and increased the ability to accurately detect and determine the location of the GI cancers.

Overall, 79% (98 of 124) of patients already diagnosed with GI cancer in the study tested positive, and 96% (88 of 92) of healthy patients tested negative.

For each GI cancer type, positivity rates were:

• 71% (24 of 34 patients) for colorectal cancer;
• 83% (19 of 23 patients) for gastric cancer;
• 75% (18 of 24 patients) for oesophageal cancer;
• 81% (17 of 21 patients) for pancreatic cancer; and
• 91% (20 of 22 patients) for ampullary cancer.

Researchers created multiple logistic regression models using machine learning. By combining the DNA methylation analysis with machine learning, the positivity rates in the patients with GI cancers increased to:  

• 88% (30 of 34 patients) for colorectal cancer;
• 91% (21 of 23 patients) for gastric cancer;
• 88% (21 of 24 patients) for oesophageal cancer;
• 90% (19 of 21 patients) for pancreatic cancer; and
• 95% (21 of 22 patients) for ampullary cancer.

Among the 4 healthy patients that tested positive for GI cancers, 3 were diagnosed with advanced adenomas, which have been found to be a precursor to developing adenocarcinoma (a type of cancer that starts in mucus-producing cells found in many organs).

“The results of this study demonstrate the promising utility of machine learning applied to genomics for the detection of cancer in the future. These findings will need to be confirmed in a larger, prospective, multi-centre study,” said ASCO Expert Peter Paul Yu, MD, FACP, FASCO. 

In the future, the researchers hope to conduct prospective studies across multiple centres to investigate the role a non-invasive, multi-cancer detection screening test such as this can have in cancer surveillance and prognosis.  

This study was funded by the Guangzhou Science and Technology Plan Project, the Doctoral Workstation Foundation of Guangdong Second Provincial General Hospital, and the Science Foundation of Guangdong Second Provincial General Hospital.

Source: ASCO