Society: AGA

Background
Colorectal cancer (CRC) remains a leading cause of cancer related mortality worldwide. We utilized cell-free DNA (cfDNA) methylation, fragmentation characteristics of selected cancer-related biomarker regions, tumor-derived signal deduction and a machine learning algorithm to refine a blood test for the early detection of CRC and advanced adenomas (AA). The aim of the study was to assess the diagnostic accuracy of the test for CRC.
Methods
This was a prospective, international (Spain, Ukraine, Germany and USA [part of NCT04792684 study] population), observational cohort study. Plasma samples from 997 patients were collected either prior to a scheduled screening colonoscopy or prior to colonic surgery for primary CRC. cfDNA samples from 170 early stage (I-II), 128 late-stage (III-IV) CRC patients (mean age 66 [44-84], female 48%, distal cancers 60%), 149 AA patients (63 high grade dysplasia; 84 low grade, > 1cm) and 550 age, gender and country of origin matched colonoscopy-checked controls were included. 155 of the control patients had a negative colonoscopy finding (cNEG), 337 had benign findings of diverticulosis, hemorrhoids, previously undiagnosed gastrointestinal diseases and/or hyperplastic polyps (BEN), 58 had non-advanced adenomas (NAA). Samples were analyzed utilizing hybrid-capture based sequencing methodology. Panel of targeted biomarkers was previously identified through tissue- and plasma-based discovery and verification workflow. Individual cfDNA fragments belonging to each biomarker region were scored for cancer-specific methylation and fragmentation signals. Finally, calculated scores were used in prediction model building and testing for establishing panel accuracy.
Results
Prediction model utilizing a panel of methylation and fragmentation scores originating from biomarkers belonging to relevant cancer development and progression related pathways, correctly classified 93% (276/298) of CRC patients and 54% (81/149) AA patients. Sensitivity per cancer stage ranged from 85% (48/56) for stage I, 94% (107/114) stage II, 94% (90/96) stage III and 97% (31/32) stage IV. Fragmentation signals contributed most to early-stage cancers (I-II), while methylation signals were more significant for late stage (III-IV) detection. High grade dysplasia AA sensitivity was 52% (33/63), while low grade >1cm AA sensitivity was 57% (48/84). Specificity of the model was 92% (504/550), with 83% (48/58) NAA, 93% (312/337) BEN and 93% (144/155) cNEG patients correctly identified. Lesion location, gender, age, BMI and country of origin were not significantly (p> 0.05) correlated to prediction outcome.

Conclusions
Use of methylation and fragmentation characteristics of cancer-related cfDNA regions, combined with a machine-learning algorithm is highly accurate for early-stage (I-II) CRCs (91% sensitivity) and AA (54% sensitivity) at 92% specificity).

Introduction: In 2016, the USPSTF recommended aspirin for primary prevention of colorectal cancer. In 2022, they reversed course citing harms among older adults and uncertainty of the preventive mode of action. Thus, understanding the mechanisms which underlie aspirin’s effects in the gastrointestinal tract, including interactions with metabolic pathways associated with CRC risk factors, will augment the development of precision prevention strategies

Methods: We measured aspirin’s influence on host and gut microbe metabolism from the ASPIRED (NCT02394769) randomized clinical trial (RCT) of daily low and standard-dose aspirin intervention for 2-3 months. We integrated untargeted plasma metabolomic profiling (Metabolon Global HD4 platform), producing 966 annotated metabolites, with paired fecal whole shotgun metagenomic sequencing. We applied a mixed-effects modeling approach to participants with paired pre- and post-treatment multi-omic data (placebo, n=57; 81 mg/d, n=57; 325 mg/d, n=50), adjusting for inter-individual variability, and known confounders such as age, sex, BMI, and technical factors such as batch and read depth.

Results: Aspirin use explained a significant proportion of variance in plasma metabolite composition (PERMANOVA, R²=4.5%, P=0.001), driven by the aspirin-related metabolites: salicylate, salicylurate, salicyluric glucuronide, and gentisate (mixed effects, FDR q<0.05). In addition, aspirin significantly decreased plasma kynurenate from baseline (β=-0.14, q=4.5x10-4), a tryptophan metabolite implicated in oncogenesis. The change in plasma kynurenate was inversely associated with change in plasma salicylate (rho=-0.45, P=2.4x10-9). “Aspirin responders'', those achieving a 33.5% reduction in urinary PGE-M from baseline, a biomarker associated with reduced risk of recurrent adenoma, had a greater reduction in plasma kynurenate than non-responders (Wilcoxon, P=0.004). In contrast, changes in the aspirin-related metabolites above were not dependent on responder status, suggesting a specific modulatory role for kynurenate in CRC risk. Aspirin also led to a diminished capacity for gut microbial tryptophan biosynthesis (mixed effects FDR < 0.25), where plasma salicylate was inversely associated with 2 microbial enzymes in the tryptophan biosynthesis pathway (E.C. 4.1.1.48, rho=-0.17, P=0.03; 5.3.1.24, rho=-0.16, P=0.04), suggesting a novel mediating role for the gut microbiome.

Conclusions: In the ASPIRED RCT, aspirin significantly decreased the circulating tryptophan metabolite, kynurenate, and altered gut microbe functional profiles involved in tryptophan biosynthesis. Our findings suggest kynurenate is a novel biomarker and tryptophan metabolism a novel mechanism by which aspirin may reduce CRC risk.

Background/Purpose: Esophageal adenocarcinoma (EAC), a cancer with a devasting 5-year survival rate of <20%, has been rapidly increasing among Western populations during the past 40 years. Barrett’s esophagus (BE), a metaplastic condition that originates in the distal esophagus, is characterized histologically by the transformation of squamous to columnar epithelium with gastric and intestinal features and is the only known precursor lesion for the development of EAC. The large number of subjects diagnosed with BE, and therefore at risk of developing EAC, underscore the importance of identifying biomarkers of disease progression that would benefit surveillance and potentially early treatment.
Methods: We generated epithelial stem cell organoids and organoid-derived monolayers from tissue samples of normal cardia (n=4), non-dysplastic BE (NDBE) (n=3), BE with low-grade dysplasia (LGD) (n=5), BE with high-grade dysplasia (HGD) (n=3), esophageal adenocarcinoma (EAC) (n=4), and gastric adenocarcinoma (GAC) (n=4). RNA was isolated from the organoids at equivalent number of passages for RNA-Seq and comparative transcriptome analysis.
Results: BE organoid stem cells and organoid-derived differentiated epithelium displayed canonical BE markers, including SOX9 and CDX2, regardless of the presence of dysplasia compared with normal gastric cardia-derived organoids and organoid-derived epithelium. Using RNASeq, we compared epithelial stem cell gene expression in NDBE, BE-LGD, BE-HGD, EAC and GAC with normal cardia (Fig.1). The intestinal transcription factor CDX2 as well as HOXA13, HOXB3, HOXC10, and to a lesser extent, TFF3 were strongly upregulated in both NDBE and dysplastic BE (Fig. 1), confirming a classical BE signature. Interestingly, NDBE and dysplastic BE displayed stronger upregulation of stemness and cancer-associated genes than both EAC and GAC (Fig. 1). Strikingly, cancer testis antigen 83 (CT83), a gene not previously linked to BE and not typically expressed in normal tissue outside testis, showed remarkable specificity for BE epithelial stem cells compared to EAC and GAC, suggesting a potential novel biomarker of BE.
Conclusion: The similar transcriptional patterns among NDBE and dysplastic BE calls into question current surveillance protocols and challenges the dogma of basing the risk of developing EAC primarily on histology. Screening transcriptional patterns, regardless of dysplasia status, could help identify BE more likely to progress to EAC. While this study is cross-sectional, it does provide a framework for future work in which the genes or pathways identified here may be manipulated in organoid models in order to determine whether they are essential in driving BE progression to EAC.

Background: While there is an increased risk of pancreatic ductal adenocarcinoma (PDAC) in patients with acute pancreatitis (AP), that risk is highest ≤ 2 years after AP diagnosis, suggesting that AP could be a sign of subclinical PDAC. It is crucial to identify individuals at high risk for subclinical PDAC–associated AP to facilitate early detection and improve survival.

Specific Aims: To estimate the 2-year cumulative incidence of PDAC diagnosis following AP and develop a prediction model using readily available clinical information to predict the presence of subclinical PDAC at the time of AP diagnosis.

Methods: Using a validated AP case finding algorithm (positive predictive value 89.1%), we identified individuals ≥ 40 years old hospitalized for AP from 1/1/1998 – 9/1/2019 within the US Veterans Health Administration. Candidate predictors included age, sex, BMI, tobacco use, alcohol use, chronic pancreatitis, gallstones, hypertriglyceridemia, diabetes, history of pancreatic cyst, history of cancers, ERCP use, and labs [hemoglobin, albumin, total cholesterol, LDL, HDL, triglycerides, AST, ALT, alkaline phosphatase (ALP), bilirubin, and maximal lipase]. We used least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation to identify a parsimonious logistic regression model for predicting the outcome: PDAC diagnosed within 2 years following the AP diagnosis. We evaluated model discrimination by estimating area under the receiver operating curve (AUC) and model calibration by comparing predicted vs. observed probability of outcome.

Results: Out of 51,613 individuals in our cohort, the mean age was 62.4 years old, 95.7% were male, and 50.3% were White (Table). Eight hundred and one individuals were diagnosed with PDAC within 2 years of AP diagnosis (2-year PDAC incidence: 1.6%), a risk 23 times higher than the SEER general population ages 60-64 years old. Almost half of the PDACs (n=326, 40.7%) were discovered after 3 months. In the final prediction model, history of pancreatic cyst was positively associated with PDAC, while common etiologies of pancreatitis were negative associated [history of gallstones and log(triglycerides)]. Positive laboratory predictors on admission included log(ALP), log(ALT), and log(bilirubin). Negative laboratory predictors included log(ALP) prior to admission and log(AST) on admission. The AUC was 0.70 after 10-fold cross validation for internal validation (Figure) with appropriate calibration.

Conclusions: The risk of PDAC diagnosis is high 2 years following acute pancreatitis. We developed a novel prediction model based on readily available clinical information that can identify among AP patients those at high risk for having subclinical PDAC. If externally validated, this model can be applied to all patients presenting with AP to facilitate early diagnosis of PDAC.

Table. Cohort Information and Final Multivariable Prediction Model for PDAC (0 months - 2 years)

Table. Cohort Information and Final Multivariable Prediction Model for PDAC (0 months - 2 years)

ROC Curve

Jane Dobkin

Appearances

A CLINICAL PREDICTION MODEL TO ASSESS RISK OF PANCREATIC CANCER AFTER ACUTE PANCREATITIS DIAGNOSIS

Presenter

Speakers