Gutin, Alexander

10512 Background: Polygenic risk scores (PRSs) combine information from single-nucleotide polymorphisms (SNPs) across the genome to explain substantial genetic breast cancer (BC) susceptibility. Previous studies have demonstrated that a multiple-ancestry PRS (MA-385) based on 56 ancestry-informative and 329 BC-associated SNPs is accurate for diverse populations and ranks among the most important known factors affecting the risk of BC development. However, medical guidelines call for more evidence before endorsing the clinical use of PRS, including studies to evaluate possible interactions of SNPs with environmental and hormonal risk factors. Here, we use longitudinal outcomes from the UK Biobank (UKB) to explore interactions of MA-385 and the five most informative individual BC-associated SNPs with the widely used Tyrer-Cuzick (TC) risk model and individual TC risk factors. Methods: The study cohort included 197,509 female UKB participants with no history of cancer at the time of study enrollment. We used Cox proportional hazards models to test associations of MA-385, individual BC SNPs, TC, and individual TC risk factors with BC outcomes. Effect modification of MA-385 and individual SNPs by clinical risk factors was evaluated by including interaction terms in the models. All models were adjusted for age at UKB enrollment and 10 principal components. Hazard ratios (HRs) and 95% confidence intervals (CIs) are reported per standard deviation, and significance tests were performed using two-sided p-values based on likelihood-ratio test statistics. Results: After a median follow-up of 11.8 years, 7,419 (3.8%) participants were diagnosed with BC.In a model including both MA-385 and TC, MA-385 was a significantly better predictor of BC development (HR = 1.56; 95% CI 1.53-1.60, p = 3.2 x 10 -329 ) over the TC model (HR = 1.21; 95% CI 1.18-1.23; p = 5.4 x 10 -69 ). We found no evidence of interaction between MA-385 and TC ( p = 0.31) , nor between individual SNPs and TC. After Bonferroni adjustment for multiple testing (102 tests), we found no evidence of interaction between MA-385 or individual SNPs with clinical TC factors. The strongest evidence for interaction was found between a BC SNP on chromosome 2 (rs13387042) and age at UKB enrollment (unadjusted p = 0.002; Bonferroni adjusted p = 0.23). Conclusions: In a longitudinal analysis of UKB, MA-385 was a highly significant predictor of BC risk and substantially improved prediction over TC. In contrast, interactions of MA-385, and individual BC SNPs, with TC and individual clinical factors in the TC model were not statistically significant. Clinical TC factors have minimal, if any, impact on the strength of the association between MA-385 and BC. These results may help to alleviate concerns expressed in guidelines that SNPs interact with environmental or hormonal risk factors.

10505 Background: Polygenic risk scores (PRS) have been shown to substantially improve breast cancer (BC) risk prediction when incorporated into traditional risk assessment models. However, PRS developed primarily in women of European ancestry have shown poor performance when applied to non-European populations. We previously described development and validation of a multiple-ancestry PRS (MA-PRS) that is based on individual ancestral genetic composition. MA-PRS has been incorporated into the Tyrer-Cuzick model for use in clinical practice and has been shown to provide independent value to the model. Here, we describe performance of the MA-PRS in a large cohort of women from diverse ancestral backgrounds. Methods: The study cohort included 203,586 patients who were referred for hereditary cancer testing and negative for pathogenic variants in genes associated with BC. MA-PRS was calculated as previously described based on 149 SNPs (93 BC and 56 ancestry-informative). Association of MA-PRS with invasive BC status (affected vs unaffected) was analyzed using logistic regression adjusted for personal and family cancer history, age, and ancestry. Calibration was assessed by examining the MA-PRS distribution within unaffected patients, and through goodness-of-fit tests. All analyses were conducted within the full cohort and within ancestral subpopulations. Odds ratios (ORs) are reported per standard deviation. Results: One-fifth (43,782/203,586, 21.5%) of patients were diagnosed with invasive BC, and 60,657 (29.8%) of patients reported BC in a first degree relative. The MA-PRS distribution was centered around zero among unaffected patients overall and within each subpopulation. MA-PRS significantly improved BC risk prediction over clinical factors in the full cohort (OR = 1.43, 95% CI:1.41-1.44) and within each ancestral subpopulation (Table). The strongest associations were observed in Ashkenazi Jewish (OR = 1.56, 95% CI:1.39-1.76) and South Asian (OR = 1.49,95% CI: 1.23-1.81) populations. The top decile of the MA-PRS consistently identified patients with two-fold increased risk of developing BC. Goodness-of-fit tests showed that the MA-PRS was well calibrated and predicted BC risk accurately in the tails of the distribution for both European and non-European women. Conclusions: The breast cancer MA-PRS based on genetic ancestral composition is well-calibrated and strongly associated with BC in all tested ancestral populations. Incorporation of PRS adds independent information to risk models and supports equitable access to personalized BC risk management. [Table: see text]

Abstract Background: Personalized breast cancer (BC) risk assessment depends on known traditional risk factors, specific germline mutations, and genome-wide polygenic risk scores (PRS). PRS explains a substantial proportion of genetic BC susceptibility. Accuracy of BC risk prediction may be improved by combining a PRS with traditional risk factors. We recently developed and validated a 149-SNP PRS for women of diverse ancestries using ancestry-informative genetic markers and combined this with version 7 of the Tyrer-Cuzick (TC) model to generate a Combined Risk Score (CRS). Here, we describe a pre-specified prospective longitudinal clinical validation of CRS as a predictor of BC risk. Methods: Women in the U.S. who were referred for clinical genetic testing between January 2017 and February 2019 were matched to medical and hospital claims in an anonymized dataset. Women with a pathogenic mutation in a BC-related gene were excluded from analysis. Follow-up began 4 months after testing and extended to the earliest date of BC diagnosis, censoring at the time of BC preventive treatment, or November 1, 2019. Incident BC events were determined by an ICD10 code of C50.* and confirmed by relevant treatment codes. CRS calibration was evaluated by the ratio of observed (O) to expected (E) incident BCs for the full cohort, and for women split into event-based 5-year CRS risk deciles. Cox proportional hazards models were used to evaluate discriminatory accuracy in terms of hazard ratios (HR) with 95% confidence intervals (CI) and p-values from likelihood ratio chi-squared statistics. Kaplan-Meier analysis was used to examine risk for women split into high- or low-risk groups according to a 3% 5-year CRS risk threshold. Results: 130,058 women with 148,349 total patient years met study eligibility criteria and were matched to claims data. Over a median (range) follow-up of 12.1 (4.0-29.5) months, 340 incident BC events were observed. The CRS was well calibrated in the overall cohort with an O/E ratio of 1.11 (95% CI=0.99-1.23) and within deciles of predicted risk (Table). Importantly, in the highest risk decile, the O/E was 0.91 (95% CI=0.63-1.27) with CRS, but 0.67 (95% CI=0.46-0.94) with TC alone, illustrating the superior calibration of CRS. In a Cox model adjusted for age at testing, PRS had an HR per standard deviation (SD) of 1.48 (95% CI=1.33-1.64, p=2.55×10-13); the HR/SD was 1.43 (95% CI=1.29-1.59, p=1.61×10-11) after adjusting for family history. In a bivariate analysis using both CRS and TC to predict time to BC, CRS added significantly to the model after accounting for TC (HR/SD=2.89, 95% CI=2.12-3.94, p=1.20×10-11), whereas TC did not add significant information after accounting for CRS. 15,986 (12.3%) women were above the CRS high-risk threshold, including 123 with events. A total of 10,248 (7.9%) women were reclassified by the CRS model compared to the TC model. Among women who were classified as high-risk by TC, 32.6% were reclassified as low-risk by CRS; among those classified as low-risk by TC, 4.3% were reclassified as high-risk by CRS. The CRS high-risk group experienced events at over three times the rate of the low-risk group (HR=3.75, 95% CI=3.00-4.68, p=6.39×10-27). Conclusion: The CRS was well-calibrated in predicting BC and significantly improved upon a traditional risk factor model. Clinical use of the CRS may lead to improved BC prevention and screening strategies. Table: Absolute risk calibration by 5-year risk decile Incidence is reported per 1,000 women-years. 34 breast cancers were observed per decile. Citation Format: Brent Mabey, Elisha Hughes, Braden Probst, Holly J. Pederson, Timothy Simmons, Brian Morris, Brooke Hullinger, Susan Domchek, Charis Eng, Monique Gary, Jennifer Klemp, Semanti Mukherjee, Vijai Joseph, Kenneth Offit, Olufunmilayo I. Olopade, Sandhya Pruthi, Allison W. Kurian, Mark E. Robson, Pat Whitworth, Susanne Wagner, Jerry Lanchbury, Thomas Slavin, Alexander Gutin. PD14-05 Prospective longitudinal validation of a breast cancer risk prediction model in a cohort of 130,058 women [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr PD14-05.

Abstract Background: Patients with homologous recombin ation (HR) deficient tumors may benefit from treatment with DNA damaging agents. Markers of genomic instability can be used to identify HR deficiency, including a 3-biomarker Genomic Instability Score (GIS). For patients with ovarian cancer (OC), the FDA-approved GIS threshold for identifying HR deficiency is 42, set as the 5th percentile for BRCA deficient tumors. Recently, a lower 1st percentile cutoff of 33 was explored in OC; this threshold was significantly associated with improved outcome after platinum-based treatment.1,2 Determining an optimal GIS threshold for different types of tumors is crucial, as the GIS distribution may vary between different cancers and even between different cancer subtypes. We propose GIS thresholds for breast cancer separately for triple-negative breast cancer (TNBC) and estrogen receptor-positive (ER+) breast cancer, using the exploratory threshold of 33 for OC as a comparator. Methods: GISs in BRCA deficient tumors were determined for patients newly diagnosed with varying stages of OC, TNBC, or ER+ breast cancer across 5 cohorts (Timms et al,3 TCGA,4 Abkevich et al,5 TBCRC008,6 the OlympiaD trial7). GIS was determined as a combination of loss of heterozygosity, telomeric-allelic imbalance, and large-scale state transitions. BRCA deficiency was defined by loss of function resulting from a pathogenic variant in BRCA1 or BRCA2 or by methylation of the BRCA1 promoter region, with loss of heterozygosity in the affected gene. GIS distributions in different cancer types and subtypes were compared using the Kolmogorov-Smirnov test. A normal distribution was fit to GISs in BRCA deficient ER+ breast tumors. The 1st percentile of the fitted distribution was chosen as the threshold. Results: A total of 561 OC tumors (190 BRCA deficient), 99 TNBC tumors (44 BRCA deficient), and 406 ER+ breast tumors (76 BRCA deficient) were included across the 5 cohorts. When score distributions were evaluated for BRCA deficient tumors, the GIS distribution within ER+ breast cancer was significantly different than for OC (p=9.6x10-5) and TNBC (p=3.2x10-4). This indicates that different GIS thresholds are appropriate for breast cancer subtypes and that the GIS threshold developed for OC is not appropriate for ER+ breast cancer. The 1st percentile of a normal distribution fit in BRCA deficient ER+ breast cancer tumors yields a threshold of 24. Using this threshold, 45.1% (183/406; 75 BRCA deficient, 108 BRCA intact) of ER+ breast tumors were HR deficient. In contrast, the GIS distribution for TNBC was not significantly different than for OC (p=0.77). Using the exploratory threshold of 33, 63.6% (63/99; 44 BRCA deficient, 19 BRCA intact) of TNBC tumors were HR deficient. Conclusions: When compared to OC, the distribution of GIS in BRCA deficient tumors was different for ER+ breast cancer, but not for TNBC. These findings are consistent with the fact that OC and TNBC are known to have similar molecular signatures.8 Exploratory thresholds of 24 for ER+ breast cancer, and 33 for TNBC and OC could be examined to determine if these cutoffs are associated with a benefit from treatment with DNA targeting agents. Clinical validity and utility of these more inclusive 1% thresholds would require demonstration of correlation with clinical outcomes. The threshold difference observed between these cancer subtypes also suggests that cancer or cancer subtype specific thresholds may be needed as evaluations of HR deficiency expands beyond OC to identify candidates for PARP inhibitors. References: 1 Mol Cancer Res. 2018;16(7):1103-11. 2 Cancers. 2021;13(5):946. 3 Br J Cancer. 2012;107(10):1776-82. 4 Nature. 2012;490(7418):61-70. 5 Breast Cancer Res. 2014;16(145):1-9. 6 J Nucl Med. 2015;56(1):31-7. 7 NEJM. 2017;377(17):1700. 8 Int J Mol Sci. 2016;17(5):759. Citation Format: Kirsten M Timms, Lauren Lenz, Chris Neff, Cara Solimeno, Darl Flake, Judy C Boughey, Matthew P Goetz, Andrea Richardson, Anna Maria Storniolo, Alexander Gutin, Roisin M Connolly, Vered Stearns, Jerry S Lanchbury. Identifying homologous recombination deficiency in breast cancer: Genomic instability score thresholds differ in breast cancer subtypes [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P5-13-09.