Abstract 5242: Do polygenic risk scores add to clinical data in predicting pancreatic cancer? a scoping review
Louise Wang,Alyssa Grimshaw,Catherine Mezzacappa,Navid Rahimi Larki,Yu-Xiao Yang,Amy Justice
DOI: https://doi.org/10.1158/1538-7445.am2023-5242
IF: 11.2
2023-04-04
Cancer Research
Abstract:Abstract Background: Individual susceptibility to pancreatic ductal adenocarcinoma (PDAC) is determined by both genetic and clinical factors. Polygenic risk scores (PRS) represent a summation of an individual’s risk associated alleles across their genome but it is unclear whether PRS improve prognostic assessments beyond available clinical data. Specific Aims: To evaluate the characteristics of studies examining PRS discrimination for PDAC before and after accounting for clinical factors. Methods: Following the PRISMA extension for scoping reviews (DOI: osf.io/97hxw), we performed a comprehensive literature search in conjunction with a professional librarian to identify preprint and published research studies evaluating PDAC PRS. Results: Nineteen studies (of 392 reviewed citations) examined associations between a PDAC-specific PRS and PDAC. The majority were conducted among the UK Biobank (n= 8) and pancreatic cancer consortia (n=4). The rest were institutional biobanks or hospital-based studies. Thirteen studies used a case-control design, and 7 adjusted for clinical risk factors (Table). Only 3 studies (15.8%) evaluated the change in discrimination with the addition of PRS to clinical factors vs. clinical factors alone, and 2 of these reported small but statistically significant improvements in discrimination (ΔAUC: 0.030, 0.039). Source populations were younger/healthier (n=9) than those at risk for PDAC, exclusively European (n=14), or drew controls without relevant exposures such as pancreatic diseases (n=2). Conclusions: Most PDAC-specific PRS studies do not account for well-established clinical factors or evaluate changes in discrimination with the addition of PRS to clinical factors. Of the 3 studies that did, only 2 showed a modest improvement in discrimination. For PRS to be clinically useful, they must demonstrate more substantial improvements in discrimination beyond established risk factors. Studies containing PDAC-specific polygenic risk scores and clinical risk factors First Author Journal Year Study Design Aim/Purpose Population Description Population Size Ancestry Definition of Pancreatic Cancer Deriving the Polygenic Risk score Analysis Clinical risk Factors Evalulation of predictive ability of models containing PRS Evaluation of the Additive Benefit of PRS Clinical Context Byrne medRxiv 2021 cohort evaluate the effect of lifestyle and genetic risk on overall cancer risk for 13 separate cancers UK Biobank 195,822 total individuals (451 cancers) European genetic ancestry ICD codes, cancer registries 22 SNPs used for pancreatic cancer Cox proportional hazards regression Townsend Index (socioeconomic index), education, birth location, income, height no no national cohort of individuals in the UK, notably younger and of European ancestry Galeotti BMJ 2021 case/control test association of pancreatic cancer specific PRS, with addition of ABO SNPs, smoking, and diabetes PANcreaticDisease ReseArch (PANDoRA) consortium: controls were without any pancreatic diseases 9409 total individuals (3619 cases, 5790 controls) European genetic ancestry confirmed diagnosis of pancreatic cancer SNPs at GWAS significance or near significance (p<10−7); also used SNPs necessary to infer ABO blood groups logistic regression all patients: country of origin, subset: smoking, Type 2 diabetes yes, evaluated the AUC for the full prediction model no limited, controls were without pancreatic disease, such as pancreatitis or pancreatic cysts Kachuri Nature Communications 2020 cohort evaluate additive predictive value of adding PRS for 16 separate cancers UK Biobank 413,753 total individuals (493 cancers) self reported European anestry ICD codes from cancer/mortality registries and inpatient hospital encouters (NHGRI)-European Bioinformatics Institute (EBI) Catalog ofpublished GWAS using standard weights, unweighted sum of risk alleles, and inverse variance weights that incorporate the standarderror of the risk effect size Cox proportional hazards regression BMI, smoking status, cigarette pack-years,family history of cancer (prostate, breast, lung or colon/rectum) Yes, calibration (estimate risk) and discrimination (AUC, NRI of cases vs. cancer free individuals) yes, PGOF: 0.171; AUC increase with PRS: 0.030, total AUC with PRS 0.745 used a national cohort of individuals in the UK, notably younger and of European ancestry Nakatochi Plos One 2018 case/control develop a risk model to identify individuals at high risk for pancreatic cancer development in the general Japanese population two separate case- control datasets in Japan controls: matched age/sex, no diagnosis of cancer at time of recuritment from five Japanese hospitals and the epidemiology research program at Aichi Cancer Center 1,328 total individuals (664 cases, 664 controls) genetically determined Japanese ancestry clinical diagnosis or histologically diagnosed pancreatic cancer, including 1.7% endocrine tumors 5 total SNPs significantly associated with pancreatic cancer logistic regression smoking status, family history of pancreatic cancer Yes, AUC in model containing PRS no limited, controls were drawn from non-cancer patients for a epidemiology research program at Aichi Cancer Center and from inpatients at 5 separate hospitals Rothwell CGH 2022 cohort evaluate metabolic syndrome, additional clinical factors across levels of polygenic risk score UK Biobank 366,016 total individuals (478 cancers) mainly European genetic ancestry ICD codes 26 SNPs at genome wide significance level Cox proportional hazards regression total physical activity, height, alcohol use, diet, smoking, highest educational level,college/university degree, use of ibuprofen, hormone replacement therapy, fasting time no, evaluated other clinical factors stratified on PRS tertile levels no used a national cohort of individuals in the UK, notably younger and of European ancestry Salvatore Journal of Biomedical Informatics 2021 case/control evaluate various phenotype risk scores (PheRS) and assess their discriminatory ability, calibration, and accuracy in combination with polygenic risk scores and clinical risk models Michigan Genomics Initiative (MGI) and UK Biobank; controls were matched on age, sex, and length of followup 431,658 total individuals (1088 cases, 430,570 controls) European ancestry ICD codes 18 independent SNPS from the GWAS catalog using 1) LD clumping and 2) P value thresholding logistic regression BMI, alcohol, smoking yes, AUC, Hosmer-Lemeshow goodness of fit, Brier score for covariates only or PRS + covariates); separate testing and validation cohort yes, modest improvement in the AUC after covariates/risk factors with PRS (Δ AUC: 0.017, no 95% CI provided) limited, controls are likely healthier than the at risk population, as they are are either sampled during a preoperative/operative appointment or are drawn from the healthier, younger UK Biobank population Sharma Gastroenterology 2022 case/control test the performance of PRS to discriminate between new onset diabetes and long standing diabetic patients with pancreatic cancer UK Biobank 11,462 total (1042 cases, 10420 controls); age and sex matched cancer free controls European ancestry incident cases of PDAC as measured by ICD codes and self reported 5 PRS calculated from each of the previous GWASes performed Cox proportional hazards regression age at DM diagnosis, DM onset, waist circumference, family history of pancreatic cancer yes, AUC in models with and without PRS yes, AUC increase with PRS: 0.039, total AUC with PRS 0.83 (p value 0.0002) used a national cohort of individuals in the UK, notably younger and of European ancestry Citation Format: Louise Wang, Alyssa Grimshaw, Catherine Mezzacappa, Navid Rahimi Larki, Yu-Xiao Yang, Amy Justice. Do polygenic risk scores add to clinical data in predicting pancreatic cancer? a scoping review. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5242.
oncology