Adenoma Detection Rate and Clinical Characteristics Influence Advanced Neoplasia Risk After Colorectal Polypectomy

Large language models for extracting histopathologic diagnoses of colorectal cancer and dysplasia from electronic health records

Background

Accurate data resources are essential for impactful medical research, but available structured datasets are often incomplete or inaccurate. Recent advances in open-weight large language models (LLMs) enable more accurate data extraction from unstructured text in electronic health records (EHRs) but have not yet been thoroughly validated for challenging diagnoses such as inflammatory bowel disease (IBD)-related neoplasia.

Objective

Create a validated approach using LLMs for identifying histopathologic diagnoses in pathology reports from the nationwide Veterans Health Administration database, including patients with genotype data within the Million Veteran Program (MVP) biobank.

Design

Our approach utilizes simple 'yes/no' question prompts for following phenotypes of interest: any colorectal dysplasia, high-grade dysplasia and/or colorectal adenocarcinoma (HGD/CRC), and invasive CRC. We validated the method on diagnostic tasks by applying prompts to reports from patients with IBD (and validated separately in non-IBD) and calculated F-1 scores as a balanced accuracy measure.

Results

In patients with IBD in MVP, we achieved F1-scores of 96.1% (95% CI 92.5-99.4%) for identifying dysplasia, 93.7% (88.2-98.4%) for identifying HGD/CRC, and 98% (96.3-99.4%) for identifying CRC. In patients without IBD in MVP, we achieved F1-scores of 99.2% (98.2-100%) for identifying any colorectal dysplasia, 96.5% (93.0-99.2%) for identifying HGD/CRC, and 95% (92.8-97.2%) for identifying CRC using LLM Gemma-2.

Conclusion

LLMs provided excellent accuracy in extracting the diagnoses of interest from EHRs. Our validated methods generalized to unstructured pathology notes, even withstanding challenges of resource-limited computing environments. This may therefore be a promising approach for other clinical phenotypes given the minimal human-led development required.

Adenoma Detection Rates Calculated Using All Examinations Are Associated With Lower Risk for Postcolonoscopy Colorectal Cancer: Data From the New Hampshire Colonoscopy Registry

INTRODUCTION

We used New Hampshire Colonoscopy Registry data to examine the association between postcolonoscopy colorectal cancer (PCCRC) risk and an adenoma detection rate (ADR) which was calculated using examinations with all indications, as compared with ADR restricted to only screening examinations.

METHODS

Our cohort study included New Hampshire Colonoscopy Registry patients with an index examination and at least 1 follow-up event, either a colonoscopy or a CRC diagnosis. Our outcome, PCCRC, was any CRC diagnosed ≥6 months after an index examination. The exposure variable was endoscopist-specific all-examination ADR (ADR-A), calculated for all indications, divided into quintiles. We also compared the ADR-A with a screening ADR (ADR-S). Cox regression was used to model the hazard of PCCRC on ADR, controlling for age, sex, and other covariates.

RESULTS

In 32,535 patients, a lower hazard for PCCRC (n = 178) was observed for ADR-A's ≥ 23%, as compared with ADR-A's <23% (reference) (23% to <29%: hazard ratio (HR) = 0.56, 95% CI: 0.36-0.87; 29% to <34%: HR = 0.60, 95% CI: 0.38-0.94; 34% to <44%: HR = 0.43, 95% CI: 0.29-0.65; and ≥44%: HR = 0.32, 95% CI: 0.16-0.63). The highest quartile of ADR-A (42%+) (HR = 0.41, 95% CI: 0.23-0.75) had a similar protection from PCCRC as the highest quartile of ADR-S (35%+) (HR = 0.38, 95% CI: 0.21-0.70). We observed 95% CIs for ADR's were 28% narrower (median = 0.72; interquartile range (IQR): 0.10) for endoscopists when using ADR-A vs ADR-S.

DISCUSSION

Our data demonstrating lower PCCRC risk in examinations performed by endoscopists with higher ADR's calculated with all examinations help to validate ADR-A as a quality measure. ADR-A may also increase precision of the calculated ADR. Endoscopists should strive for a higher ADR-A with 44% as an aspirational target.

Characteristics of Postpolypectomy Colorectal Cancer Events and Deaths

INTRODUCTION

Postpolypectomy colorectal cancers (PPCRCs) are diagnosed after a cancer-negative colonoscopy with polypectomy. Analyzing PPCRC characteristics informs prevention and early detection strategies. We investigated interval types and etiologies of PPCRCs using World Endoscopy Organization guidelines.

METHODS

PPCRCs were identified in a retrospective cohort of US Veterans who underwent colonoscopy with polypectomy from 1999 to 2016. We classified PPCRCs into interval, noninterval type A, and noninterval type B, defined as cancers diagnosed before, at, and after next recommended surveillance examination, respectively. A root cause analysis was conducted to determine the most plausible etiology.

RESULTS

We identified 396 PPCRC events and 90 PPCRC deaths over a median follow-up of 3.9 and 4.2 years, respectively. Among PPCRC events, 55% (95% confidence interval [CI] 50%-60%) were interval, 12% (95% CI 9%-15%) noninterval type A, and 33% (95% CI 29%-38%) noninterval type B. Interval cancers were more likely to be diagnosed at stage 4 than noninterval cancers (16% interval vs 2.1% noninterval type A, 8.3% noninterval type B, P = 0.003). Most interval cancers were due to possible missed lesions with adequate examinations (60%, 95% CI 53%-66%), whereas most noninterval cancers were likely new CRCs (type A: 48%, 95% CI 34%-62%; type B: 84%, 95% CI 77%-90%). Similar results were found for PPCRC deaths.

DISCUSSION

Most PPCRC events and deaths were diagnosed before the next recommended examination, largely because of procedural factors, underscoring the need to optimize quality of baseline colonoscopy and polypectomy. Many PPCRCs were diagnosed after recommended examination, suggesting the need to improve patient adherence to recommended surveillance intervals.

Systematic review: risk prediction models for metachronous advanced colorectal neoplasia after polypectomy

BACKGROUND AND AIM

Colorectal cancer (CRC) is the fourth leading cause of cancer death globally. CRC surveillance is a common indication for colonoscopy, representing a considerable burden for endoscopy services. Accurate identification of high-risk patients who would benefit from more intensive surveillance, as well as low-risk patients suitable for less frequent follow-up, could improve the effectiveness of surveillance protocols and resource use. Our aim was to identify and critically appraise published risk models for the occurrence of metachronous advanced colorectal neoplasia (ACN), defined here as CRC or advanced adenomas detected during surveillance colonoscopy.

METHODS

We searched PubMed and EMBASE for primary research studies reporting the development and/or validation of multivariable models that predict metachronous ACN risk. Screening of studies for inclusion, data extraction, and risk of bias assessment were conducted by two researchers independently.

RESULTS

We identified nine studies describing nine risk models. Six models were internally validated and two were externally validated. No model underwent both internal and external validation. Good model discrimination (concordance index > 0.7) was reported for two models during internal validation and for one model during external validation. Calibration was acceptable when assessed (n = 4). Methodological limitations and a high risk of bias were observed for all studies.

CONCLUSIONS

Several published models predicting metachronous ACN risk showed some promise. However, adherence to methodological standards was limited, and only two models were externally validated. Head-to-head comparisons of existing models using populations independent from model development cohorts should be prioritized to identify models suitable for use in clinical practice.

Predicting Risk of Colorectal Cancer After Adenoma Removal in a Large Community-Based Setting

INTRODUCTION

Colonoscopy surveillance guidelines categorize individuals as high or low risk for future colorectal cancer (CRC) based primarily on their prior polyp characteristics, but this approach is imprecise, and consideration of other risk factors may improve postpolypectomy risk stratification.

METHODS

Among patients who underwent a baseline colonoscopy with removal of a conventional adenoma in 2004-2016, we compared the performance for postpolypectomy CRC risk prediction (through 2020) of a comprehensive model featuring patient age, diabetes diagnosis, and baseline colonoscopy indication and prior polyp findings (i.e., adenoma with advanced histology, polyp size ≥10 mm, and sessile serrated adenoma or traditional serrated adenoma) with a polyp model featuring only polyp findings. Models were developed using Cox regression. Performance was assessed using area under the receiver operating characteristic curve (AUC) and calibration by the Hosmer-Lemeshow goodness-of-fit test.

RESULTS

Among 95,001 patients randomly divided 70:30 into model development (n = 66,500) and internal validation cohorts (n = 28,501), 495 CRC were subsequently diagnosed; 354 in the development cohort and 141 in the validation cohort. Models demonstrated adequate calibration, and the comprehensive model demonstrated superior predictive performance to the polyp model in the development cohort (AUC 0.71, 95% confidence interval [CI] 0.68-0.74 vs AUC 0.61, 95% CI 0.58-0.64, respectively) and validation cohort (AUC 0.70, 95% CI 0.65-0.75 vs AUC 0.62, 95% CI 0.57-0.67, respectively).

DISCUSSION

A comprehensive CRC risk prediction model featuring patient age, diabetes diagnosis, and baseline colonoscopy indication and polyp findings was more accurate at predicting postpolypectomy CRC diagnosis than a model based on polyp findings alone.

Colon Polyp Surveillance: Separating the Wheat From the Chaff

One goal of colorectal cancer (CRC) screening is to prevent CRC incidence by removing precancerous colonic polyps, which are detected in up to 50% of screening examinations. Yet, the lifetime risk of CRC is 3.9%-4.3%, so it is clear that most of these individuals with polyps would not develop CRC in their lifetime. It is, therefore, a challenge to determine which individuals with polyps will benefit from follow-up, and at what intervals. There is some evidence that individuals with advanced polyps, based on size and histology, benefit from intensive surveillance. However, a large proportion of individuals will have small polyps without advanced histologic features (ie, "nonadvanced"), where the benefits of surveillance are uncertain and controversial. Demand for surveillance will further increase as more polyps are detected due to increased screening uptake, recent United States recommendations to expand screening to younger individuals, and emergence of polyp detection technology. We review the current understanding and clinical implications of the natural history, biology, and outcomes associated with various categories of colon polyps based on size, histology, and number. Our aims are to highlight key knowledge gaps, specifically focusing on certain categories of polyps that may not be associated with future CRC risk, and to provide insights to inform research priorities and potential management strategies. Optimization of CRC prevention programs based on updated knowledge about the future risks associated with various colon polyps is essential to ensure cost-effective screening and surveillance, wise use of resources, and inform efforts to personalize recommendations.

Prevalence of diverse colorectal polyps and risk factors for colorectal carcinoma in situ and neoplastic polyps

BACKGROUND

Most colorectal cancers originate from precancerous polyps. This study aimed to determine the prevalence of colorectal polyps with diverse pathological morphologies and to explore the risk factors for colorectal carcinoma in situ (CCS) and neoplastic polyps.

METHODS

Inpatients admitted from January 2018 to May 2023 were screened through the hospital information system. Polyps were classified according to pathological morphology. The prevalence of polyps was described by frequency and 95% confidence interval. Univariate and multivariate logistic regression analyses were used to explore the risk factors for CCS and neoplastic polyps.

RESULTS

In total, 2329 individuals with 3550 polyps were recruited. Among all patients, 76.99% had neoplastic polyps and 44.31% had advanced adenomas. Tubular adenoma had the highest prevalence at 60.15%, and the prevalence of CCS was 3.86%. Patients with a colorectal polyp diameter ≥ 1.0 cm or number ≥ 3 were 8.07 times or 1.98 times more likely to develop CCS than were those with a diameter < 1.0 cm or number < 3, respectively (OR 8.07, 95%CI 4.48-14.55, p < 0.0001; and OR 1.98, 95%CI 1.27-3.09, p = 0.002). The risk of CCS with schistosome egg deposition was also significantly increased (OR 2.70, 95%CI 1.05-6.98). The higher the levels of carbohydrate antigen (CA) 724 (OR 1.01, 95%CI 1.00-1.02) and CA211 (OR 1.16, 95%CI 1.03-1.32) in patients with colorectal polyps were, the greater the risk of CCS. When colorectal neoplastic polyps were analyzed, we discovered that for each 1-year increase in age, the risk of neoplastic polyps increased by 3% (OR 1.03, 95%CI 1.02-1.04), p < 0.0001. Patients with a polyp diameter ≥ 1.0 cm had a 2.11-fold greater risk of neoplastic polyps compared to diameter < 1.0 cm patients (OR 3.11, 95%CI 2.48-3.92), p < 0.0001. In addition, multiple polyps and CA199 levels are risk factors for neoplastic polyps.

CONCLUSION

More than 3/4 of colorectal polyp patients have neoplastic polyps. Patients are more inclined to develop CCS and neoplastic polyps if they have large polyps (> 1.0 cm) or multifocal polyps. The levels of the tumor markers CA724 and CA211 show some potential usefulness for predicting CCS and may be exploited for early identification of high-risk populations.

The Diagnostic Performance of Fecal Immunochemical Tests for Detecting Advanced Neoplasia at Surveillance Colonoscopy

BACKGROUND & AIMS

An increasing burden on health care resources has resulted in a backlog of individuals requiring colonoscopy, with delays in surveillance possibly detrimental for individuals at increased risk of colorectal cancer (CRC). This study investigated the use of a 2-sample fecal immunochemical test (FIT) to establish those most likely to have advanced neoplasia (AN) and in need of prioritized surveillance colonoscopy.

METHODS

This was a prospective study conducted in the tertiary care setting. Participants completed a 2-sample FIT (OC-Sensor, Eiken Chemical Company) within 90 days of surveillance colonoscopy. The sensitivity of FIT for detection of AN (CRC or advanced adenoma) in moderate- and high-risk individuals was determined at fecal hemoglobin thresholds between 2 and 80 μg/g feces.

RESULTS

A total of 766 patients were included (median age, 66.1 years [interquartile range, 58.1-72.9]; 49.9% male), with AN detected in 8.6% (66/766, including 5 CRC). For moderate-risk individuals (with prior history of adenoma or a significant family history of CRC), sensitivity of FIT for AN ranged from 73.5% at 2 μg/g feces, to 10.2% at 80 μg/g feces. For high-risk conditions (confirmed/suspected genetic syndromes or prior CRC), sensitivity of FIT was similar, ranging from 70.6% at the lowest positivity threshold of 2 μg/g feces, to 11.8% at 80 μg/g feces. Independent variables in the whole cohort for association with detection of AN at surveillance colonoscopy were age (odds ratio, 1.03; 95% confidence interval, 1.00-1.06) and FIT hemoglobin result ≥10 μg/g feces (odds ratio, 1.81; 95% confidence interval, 1.04-3.16).

CONCLUSIONS

The use of FIT before surveillance colonoscopy provides clinicians with insights into the risk of AN. This raises the possibility of a method to triage individuals, facilitating the more efficient management of endoscopic resources.

Development and External Validation of a Prediction Model for Colorectal Cancer Among Patients Awaiting Surveillance Colonoscopy Following Polypectomy

Background and Aims

Demand for surveillance colonoscopy can sometimes exceed capacity, such as during and following the coronavirus disease 2019 pandemic, yet no tools exist to prioritize the patients most likely to be diagnosed with colorectal cancer (CRC) among those awaiting surveillance colonoscopy. We developed a multivariable prediction model for CRC at surveillance comparing performance to a model that assigned patients as low or high risk based solely on polyp characteristics (guideline-based model).

Methods

Logistic regression was used for model development among patients receiving surveillance colonoscopy in 2014-2019. Candidate predictors included index colonoscopy indication, findings, and endoscopist adenoma detection rate, and patient and clinical characteristics at surveillance. Patients were randomly divided into model development (n = 36,994) and internal validation cohorts (n = 15,854). External validation was performed on 30,015 patients receiving surveillance colonoscopy in 2020-2022, and the multivariable model was then updated and retested.

Results

One hundred fourteen, 43, and 71 CRCs were detected at surveillance in the 3 cohorts, respectively. Polyp size ≥10 mm, adenoma detection rate <32.5% or missing, patient age, and ever smoked tobacco were significant CRC predictors; this multivariable model outperformed the guideline-based model (internal validation cohort area under the receiver-operating characteristic curve: 0.73, 95% confidence interval (CI): 0.66-0.81 vs 0.52, 95% CI: 0.45-0.60). Performance declined at external validation but recovered with model updating (operating characteristic curve: 0.72 95% CI: 0.66-0.77).

Conclusion

When surveillance colonoscopy demand exceeds capacity, a prediction model featuring common clinical predictors may help prioritize patients at highest risk for CRC among those awaiting surveillance. Also, regular model updates can address model performance drift.

AGA Clinical Practice Update on Risk Stratification for Colorectal Cancer Screening and Post-Polypectomy Surveillance: Expert Review

DESCRIPTION

Since the early 2000s, there has been a rapid decline in colorectal cancer (CRC) mortality, due in large part to screening and removal of precancerous polyps. Despite these improvements, CRC remains the second leading cause of cancer deaths in the United States, with approximately 53,000 deaths projected in 2023. The aim of this American Gastroenterological Association (AGA) Clinical Practice Update Expert Review was to describe how individuals should be risk-stratified for CRC screening and post-polypectomy surveillance and to highlight opportunities for future research to fill gaps in the existing literature.

METHODS

This Expert Review was commissioned and approved by the American Gastroenterological Association (AGA) Institute Clinical Practice Updates Committee (CPUC) and the AGA Governing Board to provide timely guidance on a topic of high clinical importance to the AGA membership, and underwent internal peer review by the CPUC and external peer review through standard procedures of Gastroenterology. These Best Practice Advice statements were drawn from a review of the published literature and from expert opinion. Because systematic reviews were not performed, these Best Practice Advice statements do not carry formal ratings regarding the quality of evidence or strength of the presented considerations. Best Practice Advice Statements BEST PRACTICE ADVICE 1: All individuals with a first-degree relative (defined as a parent, sibling, or child) who was diagnosed with CRC, particularly before the age of 50 years, should be considered at increased risk for CRC. BEST PRACTICE ADVICE 2: All individuals without a personal history of CRC, inflammatory bowel disease, hereditary CRC syndromes, other CRC predisposing conditions, or a family history of CRC should be considered at average risk for CRC. BEST PRACTICE ADVICE 3: Individuals at average risk for CRC should initiate screening at age 45 years and individuals at increased risk for CRC due to having a first-degree relative with CRC should initiate screening 10 years before the age at diagnosis of the youngest affected relative or age 40 years, whichever is earlier. BEST PRACTICE ADVICE 4: Risk stratification for initiation of CRC screening should be based on an individual's age, a known or suspected predisposing hereditary CRC syndrome, and/or a family history of CRC. BEST PRACTICE ADVICE 5: The decision to continue CRC screening in individuals older than 75 years should be individualized, based on an assessment of risks, benefits, screening history, and comorbidities. BEST PRACTICE ADVICE 6: Screening options for individuals at average risk for CRC should include colonoscopy, fecal immunochemical test, flexible sigmoidoscopy plus fecal immunochemical test, multitarget stool DNA fecal immunochemical test, and computed tomography colonography, based on availability and individual preference. BEST PRACTICE ADVICE 7: Colonoscopy should be the screening strategy used for individuals at increased CRC risk. BEST PRACTICE ADVICE 8: The decision to continue post-polypectomy surveillance for individuals older than 75 years should be individualized, based on an assessment of risks, benefits, and comorbidities. BEST PRACTICE ADVICE 9: Risk-stratification tools for CRC screening and post-polypectomy surveillance that emerge from research should be examined for real-world effectiveness and cost-effectiveness in diverse populations (eg, by race, ethnicity, sex, and other sociodemographic factors associated with disparities in CRC outcomes) before widespread implementation.

Colorectal Cancer Polygenic Risk Score Is Associated With Screening Colonoscopy Findings but Not Follow-Up Outcomes

Background and Aims

Colorectal cancer (CRC) polygenic risk scores (PRS) may help personalize CRC prevention strategies. We investigated whether an existing PRS was associated with advanced neoplasia (AN) in a population undergoing screening and follow-up colonoscopy.

Methods

We evaluated 10-year outcomes in the Cooperative Studies Program #380 screening colonoscopy cohort, which includes a biorepository of selected individuals with baseline AN (defined as CRC or adenoma ≥10 mm or villous histology, or high-grade dysplasia) and matched individuals without AN. A PRS was constructed from 136 prespecified CRC-risk single nucleotide polymorphisms. Multivariate logistic regression was used to evaluate the PRS for associations with AN prevalence at baseline screening colonoscopy or incident AN in participants with at least one follow-up colonoscopy.

Results

The PRS was associated with AN risk at baseline screening colonoscopy (P = .004). Participants in the lowest PRS quintile had more than a 70% decreased risk of AN at baseline (odds ratio 0.29, 95% confidence interval 0.14-0.58; P < .001) compared to participants with a PRS in the middle quintile. Using a PRS cut-off of more than the first quintile to indicate need for colonoscopy as primary screening, the sensitivity for detecting AN at baseline is 91.8%. We did not observe a relationship between the PRS and incident AN during follow-up (P = .28).

Conclusion

A PRS could identify individuals at low risk for prevalent AN. Ongoing work will determine whether this PRS can identify a subset of individuals at sufficiently low risk who could safely delay or be reassured about noninvasive screening. Otherwise, more research is needed to augment these genetic tools to predict incident AN during long-term follow-up.

Development and validation of a risk prediction model for post-polypectomy colorectal cancer in the USA: a prospective cohort study

Background

Effective risk stratification tools for post-polypectomy colorectal cancer (PPCRC) are lacking. We aimed to develop an effective risk stratification tool for the prediction of PPCRC in three large population-based cohorts and to validate the tool in a clinical cohort.

Methods

Leveraging the integrated endoscopic, histopathologic and epidemiologic data in three U.S population-based cohorts of health professional (the Nurses' Health Study (NHS) I, II and Health Professionals Follow-up Study (HPFS)), we developed a risk score to predict incident PPCRC among 26,741 patients with a polypectomy between 1986 and 2017. We validated the PPCRC score in the Mass General Brigham (MGB) Colonoscopy Cohort (Boston, Massachusetts, U.S) of 76,603 patients with a polypectomy between 2007 and 2018. In all four cohorts, we collected detailed data on patients' demographics, endoscopic history, polyp features, and lifestyle factors at polypectomy. The outcome, incidence of PPCRC, was assessed by biennial follow-up questionnaires in the NHS/HPFS cohorts, and through linkage to the Massachusetts Cancer Registry in the MGB cohort. In all four cohorts, individuals who were diagnosed with CRC or died before baseline or within six months after baseline were excluded. We used Cox regression to calculate the hazard ratio (HR), 95% confidence interval (CI) and assessed the discrimination using C-statistics and reclassification using the Net Reclassification Improvement (NRI).

Findings

During a median follow-up of 12.8 years (interquartile range (IQR): 9.3, 16.7) and 5.1 years (IQR: 2.7, 7.8) in the NHS/HPFS and MGB cohorts, we documented 220 and 241 PPCRC cases, respectively. We identified a PPCRC risk score based on 11 predictors. In the validation cohort, the PPCRC risk score showed a strong association with PPCRC risk (HR for high vs. low, 3.55, 95% CI, 2.59-4.88) and demonstrated a C-statistic (95% CI) of 0.75 (0.70-0.79), and was discriminatory even within the low- and high-risk polyp groups (C-statistic, 0.73 and 0.71, respectively) defined by the current colonoscopy surveillance recommendations, leading to a NRI of 45% (95% CI, 36-54%) for patients with PPCRC.

Interpretation

We developed and validated a risk stratification model for PPCRC that may be useful to guide tailored colonoscopy surveillance. Further work is needed to determine the optimal surveillance interval and test the added value of other predictors of PPCRC beyond those included in the current study, along with implementation studies.

Funding

US National Institutes of Health, the American Cancer Society, the South-Eastern Norway Regional Health Authority, the Deutsche Forschungsgemeinschaft.