Genetic and Circulating Biomarker Data Improve Risk Prediction for Pancreatic Cancer in the General Population

A pancreatic cancer risk prediction model (Prism) developed and validated on large-scale US clinical data

BACKGROUND

Pancreatic Duct Adenocarcinoma (PDAC) screening can enable early-stage disease detection and long-term survival. Current guidelines use inherited predisposition, with about 10% of PDAC cases eligible for screening. Using Electronic Health Record (EHR) data from a multi-institutional federated network, we developed and validated a PDAC RISk Model (Prism) for the general US population to extend early PDAC detection.

METHODS

Neural Network (PrismNN) and Logistic Regression (PrismLR) were developed using EHR data from 55 US Health Care Organisations (HCOs) to predict PDAC risk 6-18 months before diagnosis for patients 40 years or older. Model performance was assessed using Area Under the Curve (AUC) and calibration plots. Models were internal-externally validated by geographic location, race, and time. Simulated model deployment evaluated Standardised Incidence Ratio (SIR) and other metrics.

FINDINGS

With 35,387 PDAC cases, 1,500,081 controls, and 87 features per patient, PrismNN obtained a test AUC of 0.826 (95% CI: 0.824-0.828) (PrismLR: 0.800 (95% CI: 0.798-0.802)). PrismNN's average internal-external validation AUCs were 0.740 for locations, 0.828 for races, and 0.789 (95% CI: 0.762-0.816) for time. At SIR = 5.10 (exceeding the current screening inclusion threshold) in simulated model deployment, PrismNN sensitivity was 35.9% (specificity 95.3%).

INTERPRETATION

Prism models demonstrated good accuracy and generalizability across diverse populations. PrismNN could find 3.5 times more cases at comparable risk than current screening guidelines. The small number of features provided a basis for model interpretation. Integration with the federated network provided data from a large, heterogeneous patient population and a pathway to future clinical deployment.

FUNDING

Prevent Cancer Foundation, TriNetX, Boeing, DARPA, NSF, and Aarno Labs.

An Integrative Pancreatic Cancer Risk Prediction Model in the UK Biobank

Pancreatic cancer (PaCa) is a lethal cancer with an increasing incidence, highlighting the need for early prevention strategies. There is a lack of a comprehensive PaCa predictive model derived from large prospective cohorts. Therefore, we have developed an integrated PaCa risk prediction model for PaCa using data from the UK Biobank, incorporating lifestyle-related, genetic-related, and medical history-related variables for application in healthcare settings. We used a machine learning-based random forest approach and a traditional multivariable logistic regression method to develop a PaCa predictive model for different purposes. Additionally, we employed dynamic nomograms to visualize the probability of PaCa risk in the prediction model. The top five influential features in the random forest model were age, PRS, pancreatitis, DM, and smoking. The significant risk variables in the logistic regression model included male gender (OR = 1.17), age (OR = 1.10), non-O blood type (OR = 1.29), higher polygenic score (PRS) (Q5 vs. Q1, OR = 2.03), smoking (OR = 1.82), alcohol consumption (OR = 1.27), pancreatitis (OR = 3.99), diabetes (DM) (OR = 2.57), and gallbladder-related disease (OR = 2.07). The area under the receiver operating curve (AUC) of the logistic regression model is 0.78. Internal validation and calibration performed well in both models. Our integrative PaCa risk prediction model with the PRS effectively stratifies individuals at future risk of PaCa, aiding targeted prevention efforts and supporting community-based cancer prevention initiatives.

A deep learning algorithm to predict risk of pancreatic cancer from disease trajectories

Pancreatic cancer is an aggressive disease that typically presents late with poor outcomes, indicating a pronounced need for early detection. In this study, we applied artificial intelligence methods to clinical data from 6 million patients (24,000 pancreatic cancer cases) in Denmark (Danish National Patient Registry (DNPR)) and from 3 million patients (3,900 cases) in the United States (US Veterans Affairs (US-VA)). We trained machine learning models on the sequence of disease codes in clinical histories and tested prediction of cancer occurrence within incremental time windows (CancerRiskNet). For cancer occurrence within 36 months, the performance of the best DNPR model has area under the receiver operating characteristic (AUROC) curve = 0.88 and decreases to AUROC (3m) = 0.83 when disease events within 3 months before cancer diagnosis are excluded from training, with an estimated relative risk of 59 for 1,000 highest-risk patients older than age 50 years. Cross-application of the Danish model to US-VA data had lower performance (AUROC = 0.71), and retraining was needed to improve performance (AUROC = 0.78, AUROC (3m) = 0.76). These results improve the ability to design realistic surveillance programs for patients at elevated risk, potentially benefiting lifespan and quality of life by early detection of this aggressive cancer.

Derivation and External Validation of Machine Learning-Based Model for Detection of Pancreatic Cancer

INTRODUCTION

There is currently no widely accepted approach to screening for pancreatic cancer (PC). We aimed to develop and validate a risk prediction model for pancreatic ductal adenocarcinoma (PDAC), the most common form of PC, across 2 health systems using electronic health records.

METHODS

This retrospective cohort study consisted of patients aged 50-84 years having at least 1 clinic-based visit over a 10-year study period at Kaiser Permanente Southern California (model training, internal validation) and the Veterans Affairs (VA, external testing). Random survival forests models were built to identify the most relevant predictors from >500 variables and to predict risk of PDAC within 18 months of cohort entry.

RESULTS

The Kaiser Permanente Southern California cohort consisted of 1.8 million patients (mean age 61.6) with 1,792 PDAC cases. The 18-month incidence rate of PDAC was 0.77 (95% confidence interval 0.73-0.80)/1,000 person-years. The final main model contained age, abdominal pain, weight change, HbA1c, and alanine transaminase change (c-index: mean = 0.77, SD = 0.02; calibration test: P value 0.4, SD 0.3). The final early detection model comprised the same features as those selected by the main model except for abdominal pain (c-index: 0.77 and SD 0.4; calibration test: P value 0.3 and SD 0.3). The VA testing cohort consisted of 2.7 million patients (mean age 66.1) with an 18-month incidence rate of 1.27 (1.23-1.30)/1,000 person-years. The recalibrated main and early detection models based on VA testing data sets achieved a mean c-index of 0.71 (SD 0.002) and 0.68 (SD 0.003), respectively.

DISCUSSION

Using widely available parameters in electronic health records, we developed and externally validated parsimonious machine learning-based models for detection of PC. These models may be suitable for real-time clinical application.

Clinical Prediction Models for Pancreatic Cancer in General and At-Risk Populations: A Systematic Review

INTRODUCTION

Identifying high-risk individuals using a risk prediction model could be a crucial first stage of screening pathways to improve the early detection of pancreatic cancer. A systematic review was conducted to critically evaluate the published primary literature on the development or validation of clinical risk prediction models for pancreatic cancer risk.

METHODS

MEDLINE, Embase, and Web of Science were searched for relevant articles from the inception of each database up to November 2021. Study selection and data extraction were conducted by 2 independent reviewers. The Prediction model Risk Of Bias Assessment Tool (PROBAST) was applied to assess risk of bias.

RESULTS

In total, 33 studies were included, describing 38 risk prediction models. Excluding studies with an overlapping population, this study consist of 15,848,100 participants, of which 58,313 were diagnosed with pancreatic cancer. Eight studies externally validated their model, and 13 performed internal validation. The studies described risk prediction models for pancreatic cancer in the general population (n = 14), patients with diabetes (n = 8), and individuals with gastrointestinal (and other) symptoms (symptoms included abdominal pain, unexplained weight loss, jaundice, and change in bowel habits and indigestion; n = 11). The commonly used clinical risk factors in the model were cigarette smoking (n = 27), age (n = 25), diabetes history (n = 22), chronic pancreatitis (n = 18), and body mass index (n = 14). In the 25 studies that assessed model performance, C-statistics ranged from 0.61 to 0.98. Of the 33 studies included, 6 were rated as being at a low risk of bias based on PROBAST.

DISCUSSION

Many clinical risk prediction models for pancreatic cancer had been developed for different target populations. Although low risk-of-bias studies were identified, these require external validation and implementation studies to ensure that these will benefit clinical decision making.

The age-dependent association of risk factors with pancreatic cancer

BACKGROUND

Pancreatic cancer presents as advanced disease in >80% of patients; yet, appropriate ages to consider prevention and early detection strategies are poorly defined. We investigated age-specific associations and attributable risks of pancreatic cancer for established modifiable and non-modifiable risk factors.

PATIENTS AND METHODS

We included 167 483 participants from two prospective US cohort studies with 1190 incident cases of pancreatic cancer during >30 years of follow-up; 5107 pancreatic cancer cases and 8845 control participants of European ancestry from a completed multicenter genome-wide association study (GWAS); and 248 893 pancreatic cancer cases documented in the US Surveillance, Epidemiology, and End Results (SEER) Program. Across different age categories, we investigated cigarette smoking, obesity, diabetes, height, and non-O blood group in the prospective cohorts; weighted polygenic risk score of 22 previously identified single nucleotide polymorphisms in the GWAS; and male sex and black race in the SEER Program.

RESULTS

In the prospective cohorts, all five risk factors were more strongly associated with pancreatic cancer risk among younger participants, with associations attenuated among those aged >70 years. The hazard ratios comparing participants with three to five risk factors with those with no risk factors were 9.24 [95% confidence interval (CI) 4.11-20.77] among those aged ≤60 years, 3.00 (95% CI 1.85-4.86) among those aged 61-70 years, and 1.46 (95% CI 1.10-1.94) among those aged >70 years (Pheterogeneity = 3×10-5). These factors together were related to 65.6%, 49.7%, and 17.2% of incident pancreatic cancers in these age groups, respectively. In the GWAS and the SEER Program, the associations with the polygenic risk score, male sex, and black race were all stronger among younger individuals (Pheterogeneity ≤0.01).

CONCLUSIONS

Established risk factors are more strongly associated with earlier-onset pancreatic cancer, emphasizing the importance of age at initiation for cancer prevention and control programs targeting this highly lethal malignancy.

Germline Aberrations in Pancreatic Cancer: Implications for Clinical Care

Pancreatic ductal adenocarcinoma (PDAC) has an extremely poor prognosis and represents a major public health issue, as both its incidence and mortality are expecting to increase steeply over the next years. Effective screening strategies are lacking, and most patients are diagnosed with unresectable disease precluding the only chance of cure. Therapeutic options for advanced disease are limited, and the treatment paradigm is still based on chemotherapy, with a few rare exceptions to targeted therapies. Germline variants in cancer susceptibility genes-particularly those involved in mechanisms of DNA repair-are emerging as promising targets for PDAC treatment and prevention. Hereditary PDAC is part of the spectrum of several syndromic disorders, and germline testing of PDAC patients has relevant implications for broad cancer prevention. Germline aberrations in BRCA1 and BRCA2 genes are predictive biomarkers of response to poly(adenosine diphosphate-ribose) polymerase (PARP) inhibitor olaparib and platinum-based chemotherapy in PDAC, while mutations in mismatch repair genes identify patients suitable for immune checkpoint inhibitors. This review provides a timely and comprehensive overview of germline aberrations in PDAC and their implications for clinical care. It also discusses the need for optimal approaches to better select patients for PARP inhibitor therapy, novel therapeutic opportunities under clinical investigation, and preclinical models for cancer susceptibility and drug discovery.

Current status of inherited pancreatic cancer

Background

It is estimated that about 10% of pancreatic cancer cases have a genetic background. People with a familial predisposition to pancreatic cancer can be divided into 2 groups. The first is termed hereditary pancreatic cancer, which occurs in individuals with a known hereditary cancer syndrome caused by germline single gene mutations (e.g., BRCA1/2, CDKN2A). The second is considered as familial pancreatic cancer, which is associated with several genetic factors responsible for the more common development of pancreatic cancer in certain families, but the precise single gene mutation has not been found.

Aim

This review summarizes the current state of knowledge regarding the risk of pancreatic cancer development in hereditary pancreatic cancer and familial pancreatic cancer patients. Furthermore, it gathers the latest recommendations from the three major organizations dealing with the prevention of pancreatic cancer in high-risk groups and explores recent guidelines of scientific societies on screening for pancreatic cancers in individuals at risk for hereditary or familial pancreatic cancer.

Conclusions

In order to improve patients’ outcomes, authors of current guidelines recommend early and intensive screening in patients with pancreatic cancer resulting from genetic background. The screening should be performed in excellence centers. The scope, extent and cost-effectiveness of such interventions requires further studies.

The Role of Circulating Protein and Metabolite Biomarkers in the Development of Pancreatic Ductal Adenocarcinoma (PDAC): A Systematic Review and Meta-analysis

BACKGROUND

Pancreatic ductal adenocarcinoma (PDAC) has a poor prognosis, and this is attributed to it being diagnosed at an advanced stage. Understanding the pathways involved in initial development may improve early detection strategies. This systematic review assessed the association between circulating protein and metabolite biomarkers and PDAC development.

METHODS

A literature search until August 2020 in MEDLINE, EMBASE, and Web of Science was performed. Studies were included if they assessed circulating blood, urine, or salivary biomarkers and their association with PDAC risk. Quality was assessed using the Newcastle-Ottawa scale for cohort studies. Random-effects meta-analyses were used to calculate pooled relative risk.

RESULTS

A total of 65 studies were included. Higher levels of glucose were found to be positively associated with risk of developing PDAC [n = 4 studies; pooled relative risk (RR): 1.61; 95% CI: 1.16-2.22]. Additionally, an inverse association was seen with pyridoxal 5'-phosphate (PLP) levels (n = 4 studies; RR: 0.62; 95% CI: 0.44-0.87). Meta-analyses showed no association between levels of C-peptide, members of the insulin growth factor signaling pathway, C-reactive protein, adiponectin, 25-hydroxyvitamin D, and folate/homocysteine and PDAC risk. Four individual studies also reported a suggestive positive association of branched-chain amino acids with PDAC risk, but due to differences in measures reported, a meta-analysis could not be performed.

CONCLUSIONS

Our pooled analysis demonstrates that higher serum glucose levels and lower levels of PLP are associated with risk of PDAC.

IMPACT

Glucose and PLP levels are associated with PDAC risk. More prospective studies are required to identify biomarkers for early detection.

Hepatocellular Carcinoma Risk Prediction in the NIH-AARP Diet and Health Study Cohort: A Machine Learning Approach

BACKGROUND

Prediction of hepatocellular carcinoma (HCC) development in persons with known risk factors remain a challenge and is an urgent unmet need, considering projected increases in HCC incidence and mortality in the US. We aimed to use machine learning techniques to identify a set of demographic, lifestyle, and health history information that can be used simultaneously for population-level HCC risk prediction.

METHODS

Data from 377,065 participants of the NIH-AARP Diet and Health Study, among whom 647 developed HCC over 16 years of follow-up, were analyzed. The sample was randomly divided into independent training (60%) and validation (40%) sets. We evaluated 123 participant characteristics and tested 15 different machine learning algorithms for robustness in predicting HCC risk. Separately, we evaluated variables selected from multivariable logistic regression for risk prediction.

RESULTS

The random under-sampling boosting (RUSBoost) algorithm performed best during model testing. Fourteen participant characteristics were selected for risk prediction based on differences between cases and controls (Bonferroni-corrected p-values <0.0004) and from the most frequently used variables in the initial two decision trees of the RUSBoost learner trees. A predictive model based on the 14 variables had an AUC of 0.72 (sensitivity=0.68, specificity=0.63) and independent validation AUC of 0.65 (sensitivity=0.68, specificity=0.63). A subset of 9 variables identified through logistic regression also had an AUC of 0.72 (sensitivity=0.67, specificity=0.63) and independent validation AUC of 0.65 (sensitivity=0.70, specificity=0.61).

CONCLUSION

Population-level HCC risk prediction can be performed with a machine learning-based algorithm and could inform strategies for improving HCC risk reduction in at-risk groups.

Identification of Recessively Inherited Genetic Variants Potentially Linked to Pancreatic Cancer Risk

Although 21 pancreatic cancer susceptibility loci have been identified in individuals of European ancestry through genome-wide association studies (GWASs), much of the heritability of pancreatic cancer risk remains unidentified. A recessive genetic model could be a powerful tool for identifying additional risk variants. To discover recessively inherited pancreatic cancer risk loci, we performed a re-analysis of the largest pancreatic cancer GWAS, the Pancreatic Cancer Cohort Consortium (PanScan) and the Pancreatic Cancer Case-Control Consortium (PanC4), including 8,769 cases and 7,055 controls of European ancestry. Six single nucleotide polymorphisms (SNPs) showed associations with pancreatic cancer risk according to a recessive model of inheritance. We replicated these variants in 3,212 cases and 3,470 controls collected from the PANcreatic Disease ReseArch (PANDoRA) consortium. The results of the meta-analyses confirmed that rs4626538 (7q32.2), rs7008921 (8p23.2) and rs147904962 (17q21.31) showed specific recessive effects (p<10-5) compared with the additive effects (p>10-3), although none of the six SNPs reached the conventional threshold for genome-wide significance (p < 5×10-8). Additional bioinformatic analysis explored the functional annotations of the SNPs and indicated a possible relationship between rs36018702 and expression of the BCL2L11 and BUB1 genes, which are known to be involved in pancreatic biology. Our findings, while not conclusive, indicate the importance of considering non-additive genetic models when performing GWAS analysis. The SNPs associated with pancreatic cancer in this study could be used for further meta-analysis for recessive association of SNPs and pancreatic cancer risk and might be a useful addiction to improve the performance of polygenic risk scores.

Inherited Pancreatic Cancer Syndromes and High-Risk Screening

Pancreatic cancer is the third leading cause of cancer death in the United States, with a 5-year survival rate of 9%. Individuals with inherited pancreatic cancer syndromes are at an increased risk for developing pancreatic cancer and may benefit from pancreatic cancer surveillance with the goal to detect and intervene on early-stage cancer or high-risk precursor lesions. Given the screening implications for family members and therapeutic implications for probands, all patients diagnosed with pancreatic cancer are recommended to undergo germline genetic testing.

Clinical Data Prediction Model to Identify Patients With Early-Stage Pancreatic Cancer

PURPOSE

Pancreatic cancer is an aggressive malignancy with patients often experiencing nonspecific symptoms before diagnosis. This study evaluates a machine learning approach to help identify patients with early-stage pancreatic cancer from clinical data within electronic health records (EHRs).

MATERIALS AND METHODS

From the Optum deidentified EHR data set, we identified early-stage (n = 3,322) and late-stage (n = 25,908) pancreatic cancer cases over 40 years of age diagnosed between 2009 and 2017. Patients with early-stage pancreatic cancer were matched to noncancer controls (1:16 match). We constructed a prediction model using eXtreme Gradient Boosting (XGBoost) to identify early-stage patients on the basis of 18,220 features within the EHR including diagnoses, procedures, information within clinical notes, and medications. Model accuracy was assessed with sensitivity, specificity, positive predictive value, and the area under the curve.

RESULTS

The final predictive model included 582 predictive features from the EHR, including 248 (42.5%) physician note elements, 146 (25.0%) procedure codes, 91 (15.6%) diagnosis codes, 89 (15.3%) medications, and 9 (1.5%) demographic features. The final model area under the curve was 0.84. Choosing a model cut point with a sensitivity of 60% and specificity of 90% would enable early detection of 58% late-stage patients with a median of 24 months before their actual diagnosis.

CONCLUSION

Prediction models using EHR data show promise in the early detection of pancreatic cancer. Although widespread use of this approach on an unselected population would produce high rates of false-positive tests, this technique may be rapidly impactful if deployed among high-risk patients or paired with other imaging or biomarker screening tools.

Lead-Time Trajectory of CA19-9 as an Anchor Marker for Pancreatic Cancer Early Detection

BACKGROUND & AIMS

There is substantial interest in liquid biopsy approaches for cancer early detection among subjects at risk, using multi-marker panels. CA19-9 is an established circulating biomarker for pancreatic cancer; however, its relevance for pancreatic cancer early detection or for monitoring subjects at risk has not been established.

METHODS

CA19-9 levels were assessed in blinded sera from 175 subjects collected up to 5 years before diagnosis of pancreatic cancer and from 875 matched controls from the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial. For comparison of performance, CA19-9 was assayed in blinded independent sets of samples collected at diagnosis from 129 subjects with resectable pancreatic cancer and 275 controls (100 healthy subjects; 50 with chronic pancreatitis; and 125 with noncancerous pancreatic cysts). The complementary value of 2 additional protein markers, TIMP1 and LRG1, was determined.

RESULTS

In the PLCO cohort, levels of CA19-9 increased exponentially starting at 2 years before diagnosis with sensitivities reaching 60% at 99% specificity within 0 to 6 months before diagnosis for all cases and 50% at 99% specificity for cases diagnosed with early-stage disease. Performance was comparable for distinguishing newly diagnosed cases with resectable pancreatic cancer from healthy controls (64% sensitivity at 99% specificity). Comparison of resectable pancreatic cancer cases to subjects with chronic pancreatitis yielded 46% sensitivity at 99% specificity and for subjects with noncancerous cysts, 30% sensitivity at 99% specificity. For prediagnostic cases below cutoff value for CA19-9, the combination with LRG1 and TIMP1 yielded an increment of 13.2% in sensitivity at 99% specificity (P = .031) in identifying cases diagnosed within 1 year of blood collection.

CONCLUSION

CA19-9 can serve as an anchor marker for pancreatic cancer early detection applications.

Genome-Wide Gene-Diabetes and Gene-Obesity Interaction Scan in 8,255 Cases and 11,900 Controls from PanScan and PanC4 Consortia

BACKGROUND

Obesity and diabetes are major modifiable risk factors for pancreatic cancer. Interactions between genetic variants and diabetes/obesity have not previously been comprehensively investigated in pancreatic cancer at the genome-wide level.

METHODS

We conducted a gene-environment interaction (GxE) analysis including 8,255 cases and 11,900 controls from four pancreatic cancer genome-wide association study (GWAS) datasets (Pancreatic Cancer Cohort Consortium I-III and Pancreatic Cancer Case Control Consortium). Obesity (body mass index ≥30 kg/m2) and diabetes (duration ≥3 years) were the environmental variables of interest. Approximately 870,000 SNPs (minor allele frequency ≥0.005, genotyped in at least one dataset) were analyzed. Case-control (CC), case-only (CO), and joint-effect test methods were used for SNP-level GxE analysis. As a complementary approach, gene-based GxE analysis was also performed. Age, sex, study site, and principal components accounting for population substructure were included as covariates. Meta-analysis was applied to combine individual GWAS summary statistics.

RESULTS

No genome-wide significant interactions (departures from a log-additive odds model) with diabetes or obesity were detected at the SNP level by the CC or CO approaches. The joint-effect test detected numerous genome-wide significant GxE signals in the GWAS main effects top hit regions, but the significance diminished after adjusting for the GWAS top hits. In the gene-based analysis, a significant interaction of diabetes with variants in the FAM63A (family with sequence similarity 63 member A) gene (significance threshold P < 1.25 × 10-6) was observed in the meta-analysis (P GxE = 1.2 ×10-6, P Joint = 4.2 ×10-7).

CONCLUSIONS

This analysis did not find significant GxE interactions at the SNP level but found one significant interaction with diabetes at the gene level. A larger sample size might unveil additional genetic factors via GxE scans.

IMPACT

This study may contribute to discovering the mechanism of diabetes-associated pancreatic cancer.