Tokyo, Japan, April 30, 2026. Craif Inc., a bio-AI startup focused on early cancer detection, presented findings from a collaborative study on lung cancer at the AACR Annual Meeting 2026, held in San Diego, California. The study was conducted in collaboration with Associate Professor Yu Fujita of The Jikei University School of Medicine and Dr. Takashi Nojiri of Higashiosaka City Medical Center. The findings demonstrate that urinary microRNA analysis can detect early-stage lung cancer with high accuracy,without the need for blood draws or hospital visits,and that preoperative urine samples may help predict postoperative recurrence risk.

With the goal of developing a noninvasive testing system for lung cancer, the study focused on microRNAs contained in urinary exosomes. Small RNA sequencing analysis was performed on samples from 278 lung cancer patients and 213 non-cancer controls (491 total), and a diagnostic model for lung cancer identification was constructed using machine learning algorithms. In the independent test set, the model achieved a diagnostic accuracy (AUC) of 0.941 (95% CI: 0.899–0.983), with early-stage cancer (Stage 0/I) sensitivity of 88.2% (95% CI: 73.4–95.3%) and specificity of 87.0% (95% CI: 75.6–93.6%). Age-matching and multivariate analyses confirmed that results were unaffected by background factors including age, sex, BMI, and smoking history.

・Preoperative urine may predict postoperative recurrence risk:

In 76 patients with surgically resected Stage I–II lung cancer, a Cox regression model was developed using preoperative urinary microRNA expression data. yielding a prognostic panel comprising three microRNAs: hsa-miR-181a-5p, hsa-miR-185-5p, and hsa-miR-934. Significant stratification of recurrence-free survival was confirmed between high- and low-risk groups, with a hazard ratio of 8.3 (95% CI: 1.9–37.0) and a 3-year ROC AUC of 0.796. These findings support the potential to assess recurrence risk from preoperative urine samples.

This multicenter case-control study analyzed urine samples collected from four institutions in Japan. Urinary extracellular vesicle (EV)-derived microRNAs from 278 patients with lung cancer and 213 non-cancer controls were analyzed using small RNA sequencing and machine learning to develop and validate an early detection model. A prognostic microRNA panel was also developed using preoperative urinary microRNA expression data, demonstrating the ability to stratify recurrence risk. These findings support a single-assay approach, — the possibility of achieving both early lung cancer detection and recurrence risk assessment through a single urine collection, without blood draws or hospital visits.