Abstract A027: Mode of presentation and symptoms at diagnosis of early- and late-onset colorectal cancer: a large retrospective AI-powered analysis
Bibliographic record
Abstract
Abstract Introduction Early-onset colorectal cancer (EOCRC) is a distinct clinical entity and a rising global challenge. As presentation and symptom patterns at diagnosis have implications for early detection and prognosis, we aimed to characterize differences between EOCRC and late-onset colorectal cancer (LOCRC) at scale. Methods A retrospective cohort of colorectal cancer (CRC) cases diagnosed 1990-2025 was collected. EOCRC and LOCRC were defined as age <50 and ≥50 years at diagnosis. Presentation mode (symptomatic, emergency, screening, incidental, or unclear) and symptoms at diagnosis were classified using the GPT-4o large language model (LLM) through prompt engineering based on manually curated clinical notes relevant to diagnosis and validated on 500 random samples. Patient-level clinical and demographic data were obtained from institutional datasets. Chi-square and odds ratios were used for comparisons. Results LLM-based classification showed strong agreement with manual review (Cohen’s κ=0.90). Of 41,152 CRC cases, 29% were EOCRC and 71% LOCRC. Stage IV at diagnosis (38% vs 31%, p<0.001) and left-sided primaries (76% vs 66%, p<0.001) were more prevalent in EOCRC compared with LOCRC. EOCRC patients more often presented as symptomatic (72% vs 65%, OR 1.36, 95% CI 1.29-1.44, p<0.001) or emergent (23% vs 20%, OR 1.22, 95% CI 1.15-1.29, p<0.001) and were less often detected by screening (3% vs 12%, OR 0.24, 95% CI 0.21-0.27, p<0.001), whereas no meaningful difference was observed for incidental cases. The symptomatic or emergent presentation of EOCRC was more frequently characterized by abdominal/rectal pain (45% vs 35%, OR 1.49, 95% CI 1.41-1.57), bleeding-related symptoms (49% vs 44%, OR 1.25, 95% CI 1.19-1.32), and changes in bowel habits (30% vs 27%, OR 1.16, 95% CI 1.09-1.23), and less often by anemia (11% vs 16%, OR 0.63, 95% CI 0.58-0.68) or constitutional symptoms (13% vs 16%, OR 0.79, 95% CI 0.73-0.85, p<0.001 for all) compared with LOCRC. The proportion of screening-detected cases rose from 3% to 4% in EOCRC (p=0.007) and 9% to 15% in LOCRC (p<0.001) between 1990-2009 and 2010-2025. After the 2021 national screening guideline update, lowering the starting age from 50 to 45 years, screening-detected cases among 45–49-year-old patients increased from 4% to 13% (p<0.001). Conclusions EOCRC predominantly presents as symptomatic or emergent and at a more advanced stage, underscoring the need for enhanced screening strategies and increased public awareness. Citation Format: Emerik Osterlund, Berta Martin-Cullell, Sebastian Correa Cautino, Mahmoud M. G. Yousef, Songwit Payapwattanawong, Neha Y. Agrawal, Paul M. Roy, Kaysia Ludford, John Paul Y. C. Shen, Xiling Shen, Michael J. Overman, Kanwal P. S. Raghav, Arvind Dasari, Van K. Morris, Christine M. Parseghian, Luisa M. Solis Soto, Michael G. White, Yi-Qian Nancy. You, Victoria Serpas Higbie, Scott Kopetz, Guglielmo Vetere. Mode of presentation and symptoms at diagnosis of early- and late-onset colorectal cancer: a large retrospective AI-powered analysis [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: The Rise in Early-Onset Cancers—Knowledge Gaps and Research Opportunities; 2025 Dec 10-13; Montreal, QC, Canada. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(23_Suppl):Abstract nr A027.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.005 | 0.017 |
| Meta-epidemiology (narrow) | 0.000 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.003 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".