Published in Cancer Science: 301 participants at five Japanese institutions, 100% sensitivity at Stage 0 and 91% at Stage I, 66% specificity in the independent validation cohort

Craif Inc., a bio-AI company working on early cancer detection, reported results from a multicenter study showing that microRNA carried in urinary exosomes, analyzed with machine learning, can identify patients with esophageal squamous cell carcinoma. In an independent validation cohort the model reached an area under the curve (AUC) of 0.85, with sensitivity of 84% and specificity of 66%. The results were published in Cancer Science, the official journal of the Japanese Cancer Association.

Esophageal cancer rarely causes symptoms in its early stages, and most patients are diagnosed only after the disease has progressed. Endoscopy is the established route to early diagnosis, but it is invasive and burdensome enough that people without symptoms cannot easily undergo it repeatedly. The study asked whether urine, which can be collected noninvasively, carries enough signal to identify esophageal cancer patients efficiently when combined with microRNA analysis and AI.

Craif was founded in 2018 and detects biomarkers, including DNA and microRNA, in biofluids such as urine using its NANO IP (NANO Intelligence Platform) analysis platform. This study was run across five institutions in Japan and analyzed samples from 301 participants.

What the study found

• Urinary microRNA was analyzed in 149 esophageal cancer patients and 152 healthy adults across five institutions.
• Recursive Feature Elimination, a machine learning method, produced an algorithm built on 57 microRNAs.
• The model reached an AUC of 0.90 in the training set and 0.85 in the independent validation set, where sensitivity was 84% and specificity was 66%.
• In superficial disease, sensitivity was 100% at Stage 0 and 91% at Stage I.
• Comparing urine before and after treatment, the score fell significantly, which suggests it may reflect the effect of the tumor.

How the study was done

The study was conducted with Tatsuro Murano, who was a staff physician at National Cancer Center Hospital East at the time of the study and is now Lecturer in the Department of Optical Medicine at Institute of Science Tokyo, and Tomonori Yano, Head of the Department of Gastrointestinal Endoscopy at National Cancer Center Hospital East.

Urine samples from esophageal cancer patients and healthy adults were collected jointly at five institutions: National Cancer Center Hospital East, Tokyo Metropolitan Tama Medical Center, Kasuga Clinic (Doyukai Medical Corporation), Tsujinaka Hospital Kashiwanoha (Kokikai Medical Corporation), and Institute of Science Tokyo Hospital. Exosomes were extracted from each sample and microRNA was analyzed comprehensively by next generation sequencing. From those profiles, Recursive Feature Elimination was used to build an algorithm on the 57 microRNAs that contribute most to esophageal cancer detection. The model performed well in an independent validation cohort, with an AUC of 0.85, sensitivity of 84% and specificity of 66%, which supports the possibility of picking up urinary microRNA changes in esophageal cancer patients who have no symptoms.

Scores fell after the tumor was removed

In patients whose tumors were completely resected by endoscopic treatment or surgery, post-treatment urine tended to show microRNA expression patterns closer to those of healthy adults. That points to a further property of the signature, as a biomarker reflecting the presence of esophageal cancer.

Why it matters

Endoscopy is useful for early diagnosis of esophageal cancer, but the invasiveness and the burden of the visit make it hard to undergo repeatedly while still asymptomatic. That leaves a need for tests that hold patient burden down and route the right people efficiently into endoscopy and other detailed examination.

Urine is noninvasive, easy to collect, and suited to repeated evaluation. The urinary microRNA and AI approach shown in this study could become the basis of a low burden first line test that effectively selects the people who need endoscopy. Over time it could contribute to more efficient esophageal cancer testing and wider access to it.

At present this is not covered by Japanese public health insurance, is not recommended for general clinical practice, and is not recommended as part of the national cancer screening program in Japan. 

Publication

Journal: Cancer Science

Title: "Development and validation of a urinary exosomal miRNA diagnostic panel for early detection of esophageal cancer"

Authors: Tatsuro Murano, Hiroki Yamashita, Yuki Kano, Ken Takeuchi, Takayuki Amano, Takanobu Yoshimoto, Mayuko Otomo, Hisashi Fujiwara, Shin Namiki, Hiroki Yamaguchi, Yoriko Ando, Yumi Nishiyama, Mika Mizunuma, Yuki Ichikawa, and Tomonori Yano

Link: https://doi.org/10.1111/cas.70298

About Craif

Craif is a bio-AI company founded in 2018 and working on early cancer detection. It combines AI with NANO IP (NANO Intelligence Platform), its own analysis platform for detecting a range of biomarkers, including DNA and microRNA, in biofluids such as urine. Craif is developing tests intended to support very early detection of cancer, earlier treatment, and a faster return to normal life. Its urine based cancer risk test, miSignal, is on the market in Japan.

Company details

Company: Craif Inc.
Representative Director and CEO: Ryuichi Onose
Founded: May 2018
Business: research and development of next generation testing aimed at early detection and personalized care for disease, primarily in oncology, and provision of miSignal, a urine based cancer risk test
Headquarters: THE PORTAL iidabashi B1F, 8-30 Shinkogawacho, Shinjuku-ku, Tokyo

URL: https://craif.com/