Japan Digestive Disease Week 2025
International Poster Session2 (JDDW)
October 31, 10:18–10:58, Room 15 (Kobe International Exhibition Hall No.1 Building Digital Poster Venue)
IP-8_G
Assessing the GastroFlag Algorithm for Detection of Gastric and Esophageal Cancer in Japan
- 1
- Kansai Occupational Health Association, Acty Health Evaluation Center
- 2
- Kansai Occupational Health Association, Senri LC Health Evaluation Center
- 3
- Medial Early Sign
- 4
- Roche Diagnostics International, United States
- 5
- Roche Diagnostics International, Switzerland
Despite national screening programs, gastric cancer remains a major health issue in Japan. Roche Diagnostics and Medial Early Sign developed GastroFlag, an AI algorithm using blood tests and demographics to identify individuals at higher risk of gastric and esophageal cancers.
This study aims to evaluate the performance of GastroFlag in Japan.
A retrospective case-control study analyzed data from our center (2010-2020). Cases had at least one blood count test within 180 days before diagnosis.The GastroFlag model included age, sex, CBC, BMI, and 13 additional lab tests. Performance metrics were calculated from bootstrapping samples.
The study included 22 esophageal (20 males) and 95 gastric cancer patients (73 males), and 5000 controls (2222 males) aged 40 to 89. At 97% specificity, GastroFlag showed an OR of 34.2 [21.6-54.5] with a sensitivity of 50.7% [40.0-62.8] for gastric cancer detection. For esophageal cancer, OR was 31.1 [9.9-71.1] with sensitivity of 46.2% [23.5-68.8]. For combined detection, OR was 33.7 [22.6-48.9] with sensitivity of 50.5% [41.1-60.2].
GastroFlag can enhance screening by prioritizing individuals for further diagnostic evaluation.
This study aims to evaluate the performance of GastroFlag in Japan.
A retrospective case-control study analyzed data from our center (2010-2020). Cases had at least one blood count test within 180 days before diagnosis.The GastroFlag model included age, sex, CBC, BMI, and 13 additional lab tests. Performance metrics were calculated from bootstrapping samples.
The study included 22 esophageal (20 males) and 95 gastric cancer patients (73 males), and 5000 controls (2222 males) aged 40 to 89. At 97% specificity, GastroFlag showed an OR of 34.2 [21.6-54.5] with a sensitivity of 50.7% [40.0-62.8] for gastric cancer detection. For esophageal cancer, OR was 31.1 [9.9-71.1] with sensitivity of 46.2% [23.5-68.8]. For combined detection, OR was 33.7 [22.6-48.9] with sensitivity of 50.5% [41.1-60.2].
GastroFlag can enhance screening by prioritizing individuals for further diagnostic evaluation.
- Index Term 1: Gastric cancer screening
- Index Term 2: medical algorithm