Citations

1512 Background: Tobacco smoking profoundly impacts LC risk; however, data are limited as to what extent SC prior to diagnosis impacts LC overall survival (OS) and lung cancer specific survival (LCSS). LC screening offers a possible teachable moment, but there is uncertainty of SC benefits after a lifetime of smoking. We use the ILCCO database to answer if SC prior to LC dx is associated with better OS and LCSS, considering time since smoking cessation (TSSC). Methods: Using individual data, analysis was performed on 17 ILCCO studies with available TSSC to estimate survival using univariable analysis and models of stage-adjusted and cumulative smoking-adjusted multivariable analysis. Adjusted Hazard Ratios (aHR) from Cox models, cubic spline smooth curves and Kaplan-Meier curves were created. Sensitivity analysis was performed for TSSC and LCSS on 13 studies. Results: Of 34649 patients, 14322 (41%) were current smokers 14273 (41%) ex-smokers and 6054 (18%) never smokers at diagnosis. We confirmed that ex-smokers (aHR 0.88 CI 0.86-0.91) and never smokers (aHR 0.76 CI 0.73-0.8) improved OS compared to current smokers. Amongst ex-smokers, < 2y TSSC (aHR 0.88 CI 0.82-0.94), 2-5y TSSC (aHR 0.83 CI 0.77-0.90) and > 5y TSSC (aHR 0.8 CI 0.76-0.84) had improved OS compared to CS. Sensitivity analysis showed a trend towards improved LCSS survival for < 2y TSSC (aHR 0.95 CI 0.86-1.05) and 2-5y TSSC (aHR 0.93 CI 0.83-1.04), whereas > 5y TSSC significantly improved LCSS by 15% (aHR 0.85 CI 0.78-0.92). To mimic the LC screening participants, in analysis of > 30 pack-years individuals, associations were strikingly strong: < 2y TSSC had improved OS by 14% (aHR 0.86 CI 0.80-0.93); 2-5y TSSC by 17% (aHR 0.83 CI 0.76-0.90); and > 5 TSSC by 22% (aHR 0.78 CI 0.74-0.83), compared to current smokers; for < 30 packs-years, a trend towards better OS was observed for < 2y TSSC (aHR 0.95 CI 0.92-1.02) and 2-5y TSSC (aHR 0.86 CI 0.74-1.01), whereas > 5y TSSC improved OS by 23% (aHR 0.77 CI 0.72-0.82). Conclusions: Among ex-smokers, the risk of overall death was reduced by 12% on < 2y TSSC, 17% on 2-5y TSSC and 20% > 5y TSSC, whereas for LCSS, the benefit was significant only for > 5y TSCC, compared to current smokers at time of diagnosis. Here we demonstrate that convincing screening participants to quit smoking at any point of their trajectory, even just prior to dx such as < 2y TSSC, improved OS, and LCSS benefit was present beyond 5y of quitting. These relationships are independent of pack-years, age, across all stages and other prognostic variables.

AIMS To report the major characteristics and clinical outcomes of COVID-19 patients treated with ACE inhibitors and ARBs and compare the different effects of the two drugs for outcomes of COVID-19 patients. METHODS This is a retrospective, two-center case series of 198 consecutive COVID-19 patients with a history of hypertension. RESULTS Among 198 patients, 58 (29.3%) and 16 (8.1%) were on were on ARB and ACEI, respectively. Patients who were on ARB or ACEI/ARB had a significantly lower rate of severe illness and ARDS when compared with patients treated with ACEI alone or not receiving RAAS blocker (P＜0.05). The Kaplan-Meier survival curve showed that patients with ARB in their antihypertensive regimen had a trend towards a higher survival rate when compared with individuals without ARB (adjusted hazard ratio, 0.27; 95% CI, 0.07-1.02; P=0.054). The occurrence rates of severe illness, ARDS, and death were similar in the two groups regardless of receiving ACEI. The Cox-regression analysis to compared ACEI vs. ARB groups showed a significantly lower mortality rate in the ARB group (adjusted hazard ratio, 0.03; 95% CI, 0.00-0.58; P=0.02). CONCLUSIONS Our data may provide some evidence of using ARB, but not ACEI, was associated with a reduced rate of severe illness and ARDS, indicating their potential protective impact in COVID-19. Further large sample sizes and multiethnic populations are warranted to confirm our findings.

Background: Disease progression prediction based on neuroimaging biomarkers is vital in Alzheimer’s disease (AD) research. Convolutional neural networks (CNN) have been proved to be powerful for various computer vision research by refining reliable and high-level feature maps from image patches. Objective: A key challenge in applying CNN to neuroimaging research is the limited labeled samples with high dimensional features. Another challenge is how to improve the prediction accuracy by joint analysis of multiple data sources (i.e., multiple time points or multiple biomarkers). To address these two challenges, we propose a novel multi-task learning framework based on CNN. Methods: First, we pre-trained CNN on the ImageNet dataset and transferred the knowledge from the pre-trained model to neuroimaging representation. We used this deep model as feature extractor to generate high-level feature maps of different tasks. Then a novel unsupervised learning method, termed Multi-task Stochastic Coordinate Coding (MSCC), was proposed for learning sparse features of multi-task feature maps by using shared and individual dictionaries. Finally, Lasso regression was performed on these multi-task sparse features to predict AD progression measured by the Mini-Mental State Examination (MMSE) and the Alzheimer’s Disease Assessment Scale cognitive subscale (ADAS-Cog). Results: We applied this novel CNN-MSCC system on the Alzheimer’s Disease Neuroimaging Initiative dataset to predict future MMSE/ADAS-Cog scales. We found our method achieved superior performances compared with seven other methods. Conclusion: Our work may add new insights into data augmentation and multi-task deep model research and facilitate the adoption of deep models in neuroimaging research.

Background: The FIGHT (Functional Impact of GLP-1 [glucagon-like peptide-1] for Heart Failure Treatment) trial randomized 300 patients with heart failure with reduced ejection fraction (HFrEF) and a recent hospitalization for heart failure to liraglutide versus placebo. While there was no difference in the primary outcome (rank score of time to death, time to rehospitalization for heart failure, and change in NT-proBNP [N-terminal pro-B-type natriuretic peptide]), there was a significant increase in cystatin C among patients randomized to liraglutide raising concern of adverse renal outcomes. We performed a post hoc analysis of FIGHT to investigate whether liraglutide was associated with worsening renal function (WRF). Methods: The relationship between randomization to liraglutide and WRF was evaluated using logistic regression models. Two hundred seventy-four patients (91%) had complete data to assess for WRF defined as: increase in SCr ≥0.3 mg/dL, or ≥25% decrease in estimated glomerular filtration rate, or an increase in cystatin C ≥0.3 mg/L from baseline to 180-days. Results: Patients with WRF (n=113, 41%), compared with those without, were older, had more comorbidities, and lower utilization of guideline-directed medical treatment. Logistic regression models showed that age and baseline cystatin C levels were associated with WRF. In adjusted models, liraglutide was not associated with excess risk of WRF compared with placebo (odds ratio, 1.02 [95% CI, 0.62–1.67]). There was also no difference in the rank score when WRF was added as a fourth-tier outcome. Conclusions: Liraglutide was not associated with WRF among patients with HFrEF and a recent hospitalization for heart failure. These data support the relative renal safety profile of liraglutide among patients with HFrEF. Registration: URL: http://www.clinicaltrials.gov . Unique identifier: NCT01800968.

Abstract Background: HLA associations for the risk of developing MM have been described in various non-Caucasian ethnic groups but have not revealed consistent results. Racial and genetic differences in cytokine polymorphisms linked to extended haplotypes; association of autoimmune disorders with HLA types; the proximity of the HLA Class I and II genes to TNF and other Class III genes responsible for complement and cytokines suggest that HLA polymorphisms may be associated with the risk of developing MM. Aim: We compared HLA-A,-B and -DRB1 2-digit allele and haplotype frequencies in patients with MM to the normal population in order to define HLA specificities associated with the risk of developing MM. Methods: A cohort of 1803 US Caucasian patients with MM were identified based on availability of HLA-A, B, DRB1 DNA-based typing or serologic typing converted to 2-digit alleles. These patients had either received a matched sibling allogeneic transplant for MM reported to the CIBMTR (N=174) or had HLA typing performed for an unrelated donor search conducted through the US National Marrow Donor Program (NMDP) (N=1629). HLA-A, B and DRB1 allele frequencies among patients were compared to A-B-DRB1 frequencies in HLA typed healthy US Caucasian controls (N= 433838) from the NMDP database. The most frequent alleles among patients and controls as well as HLA haplotype frequencies generated using the EM (expectation-maximization) algorithm were compared using Chi-Square analysis with Bonferroni correction for the number of comparisons. Results: Patients with MM had a median age of 49 (range 3–78) years and M/F ratio of 1.86 (1172/631). The controls had a median age of 44 (range 18–60) years and M/F ratio of 0.735 (183785/250053). Comparison of HLA-A, -B and –DRB1 single-locus allele frequencies between the myeloma patients and controls (those in the top 10 in either group) revealed similarity except for HLA-B*07 and B*44. B*44 was observed less frequently among the patients (a protective effect) frequency (f): 0.126 vs 0.145 (p= 0.005, Odds Ratio (OR): 0.848). On the contrary, HLA-B*07 was more common (a predisposing effect) among patients (OR: 1.135, p=0.047). Among the top 20 two locus (HLA-A-B and HLA-B-DRB1) haplotype frequencies in either group, B*44-DRB1*13 and B*07-DRB1*13 were significant: OR: 0.506 and 1.714 (p=0.003 and 0.007) respectively. A-B-DRB1 haplotype comparisons showed significance for: A*02-B*44-DRB1*04 (OR: 0.686, p=0.021), A*02-B*44-DRB1*13 (OR: 0.445, p=0.065) and A*02-B*07-DRB1*04 (OR: 1.968, p=0.029) confirming the single and B-DRB1 combination comparison results. Significantly different HLA-A, -B and -DRB1 single-locus allele or haplotype frequencies between patients and controls are shown in table 1. Conclusion: This is the largest epidemiologic study of HLA allele frequencies and the first study on HLA haplotype frequencies in MM. We observed a protective effect on the risk of developing MM with B*44 and a predisposing effect with B*07. Similar predisposing and protective haplotype associations were observed with A*02-B*07-DRB1*04 and A*02-B*44-DRB1*04 respectively. A revised analysis using a gender-matched control group is in progress. Further study including major histocompatibility region SNP mapping and candidate gene testing will help elucidate the genetic basis underlying this observation. Table 1. Allele/Haplotype Frequency Controls Frequency MM Odds Ratio (OR) P value (corrected) OR = odds of observing the allele/haplotype in patients compared to controls A*02-B*44-DRB1*13 0.00620 0.00277 0.445 0.065 B*44-DRB1*13 0.01359 0.00686 0.507 0.003 A*02-B*44-DRB1*04 0.02605 0.01813 0.686 0.021 B*44 0.14552 0.12618 0.848 0.005 B*07 0.13313 0.14836 1.134 0.047 B*07-DRB1*13 0.00814 0.01397 1.714 0.007 A*02-B*07-DRB1*04 0.00367 0.00721 1.968 0.029

Abstract Background: Pathogenic variants detected in multi-gene cancer predisposition panels are increasingly used to counsel women regarding their risk for breast cancer. However, the clinical implications of moderate penetrance genes (e.g. CHEK2, ATM) remain unclear. Breast MRI is indicated for women with a lifetime risk of breast cancer of &gt;20%, and polygenic risk scores (PRS) based on common variants discovered in recent genome-wide association studies (GWAS) may identify women with pathogenic variants in CHEK2 and ATM who are above or below this level of risk. PRS may also stratify carriers with substantially elevated lifetime risks of breast cancer. Previous work found that PRS modified breast cancer risk in carriers of pathogenic variant in BRCA1 or BRCA2, but the joint effects of pathogenic variants and PRS have not been studied in samples drawn from the general population. There is also no published study evaluating the effect of PRS among women with pathogenic variants in genes other than BRCA1/2 and CHEK2. Here, we evaluated the joint effects of PRS and pathogenic variants in ten established breast cancer predisposition genes in a population-based case-control sample. Method: We analyzed 43,144 non-Hispanic European-ancestry individuals (20,988 controls and 22,120 cases) drawn from 7 cohorts and 2 population-based case-control studies in the CAnceR RIsk Estimates Related to Susceptibility” (CARRIERS) consortium. Targeted sequencing was performed using a dual bar-coded QIAseq panel of breast cancer predisposition genes. Variant calling was conducted with Haplotype Caller and Vardict. The PRS was calculated based on 106 common variants using effect estimates from the largest published breast cancer GWAS. Genetic status was evaluated for 10 breast cancer predisposition genes: ATM, BARD1, BRCA1, BRCA2, BRIP1, CDH1, CHEK2, NF1, PALB2 and TP53. Lasso Penalized regression was performed to assess the combined effect on overall breast cancer of pathogenic variants and common variants (including main and interaction effects between age, family history, carrier status and PRS). The fraction of women at &gt;2-fold increased risk (which is associated with approximately 20% lifetime risk of breast cancer) was calculated for pathogenic variant carriers and non-carriers of different family history and age groups (&lt;40, 40-50, 50-60, 60-70, &gt;70yrs old). Results: Among non-carriers, one standard deviation change in the PRS was associated with a 1.63x change in odds of breast cancer. The best-fitting risk model did not include any PRS-by-carrier interaction terms, implying the relative risk gradient associated with PRS among women with pathogenic variants is similar to that among non-carriers. For carriers of pathogenic variants in moderate penetrance genes with no family history of breast cancer, the PRS can identify 60.6% of ATM pathogenic variant carriers and 30% of CHEK2 pathogenic variant carriers that have less than 2-fold risk of breast cancer; when carriers have a family history of breast cancer, the PRS can identify 30% of ATM pathogenic variant carriers and 11.8% of CHEK2 pathogenic variant carriers with less than 2-fold risk of breast cancer. Conclusion: In our large, population-based case-control study, the effect of the PRS among pathogenic variant carriers was not significantly different than among non-carriers. However, the PRS may be particularly important for estimating breast cancer risk among carriers of pathogenic variants in moderate penetrance genes such as CHEK2 and ATM, enabling a more precise approach for MRI screening strategy and breast cancer risk management. Citation Format: Chi Gao, Eric C. Polley, Chunling Hu, Steven N. Hart, Rohan Gnanaolivu, Kun Y. Lee, Jie Na, Nicholas J. Boddicker, Raed Samara, Paul Auer, Leslie Bernstein, Mia Gaudet, Amy Trentham-Dietz, Song Yao, Christopher Haiman, Janet E. Olson, Susan Neuhausen, Jeffrey N. Weitzel, David E. Goldgar, Susan Domchek, Katherine L. Nathanson, Fergus J. Couch, Peter Kraft, on behalf of the CARRIERS consortium. Polygenic risk scores provide clinically meaningful risk stratification among women carrying moderate penetrance pathogenic variants in breast cancer predisposition genes: Results from the CARRIERS study [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr PD3-02.

Abstract Germline mutations in several cancer predisposition genes included in hereditary multigene testing panels have been associated with increased breast cancer risk. However, estimates of breast cancer risks for each gene have in large part been derived from studies of high-risk populations enriched for family history of breast and ovarian cancer, young age of breast cancer diagnosis, or founder mutations. The breast cancer risks associated with mutations in many of these genes in the general population remain to be clearly defined. As clinical hereditary cancer genetic testing is already provided to many women at increased risk and may soon be offered to all breast cancer patients, population-based risk estimates associated with predisposition gene mutations will be needed for appropriate clinical management of mutation carriers. The “CAnceR RIsk Estimates Related to Susceptibility” (CARRIERS) consortium of 17 cohort-based case-control, population-based case-control, and family studies focused on estimation of population-based breast cancer risks and penetrance analysis for mutations in known and candidate breast cancer predisposition genes. Germline DNA from 39486 breast cancer cases and 35868 matched controls was subjected to dual bar-coded QIAseq multiplex amplification of 1733 target regions in 37 predisposition genes. Products from sets of 768 samples were pooled and sequenced in each lane of a HiSeq 4000 system. Mutations were called by GATK Haplotype Caller and Verdict. High quality sequence data was obtained for 99.3% of target regions. Here we report on results from 34741 population-based breast cancer cases and 32728 matched unaffected controls. The mean age of breast cancer diagnosis for cases was 61.2 years and mean age for controls was 61.1 years. Overall, 20.9% of cases and 14.3% of controls had a family history of breast cancer. Furthermore, 74.3% of cases and 75.8% of controls were non-Hispanic white, whereas 13.3% of cases and 15.3% of controls were African American. Among cases with tumor information available, 81.7% were ER-positive and 7.7% were triple negative breast cancers (TNBC). We assessed mutation frequencies in 24 genes including 12 established breast cancer predisposition genes (ATM, BARD1, BRCA1, BRCA2, CDH1, CHEK2, NF1, PALB2, PTEN, RAD51C, RAD51D, andTP53). The overall mutation frequency was 6.2% in cases and 2.7% in controls. Case-control association analysis after adjusting for study, age, family history of breast cancer, and race/ethnicity showed that BRCA1(OR=7.45; 95%CI:5.24-10.95) and BRCA2(OR=5.31; 95%CI:4.15-6.88) were associated with high risks of breast cancer in the general population. However, PALB2mutations were only associated with moderate risks (OR=3.63; 95%CI:2.57-5.26). ATMmutations conferred attenuated risks (OR=1.83; 95%CI:1.47-2.28) relative to studies of high-risk families or young onset breast cancers. NBNmutations were not associated with increased breast cancer risk. Mutations in BARD1, RAD51C, RAD51Dand XRCC2were specifically significantly associated with moderate risks of ER-negative breast cancer and TNBC, whereas mutations in ATM, CDH1and CHEK2were specifically significantly associated with ER-positive disease. On the basis of these findings age-related absolute breast cancer risks for commonly mutated genes were estimated. Overall, results from the CARRIERS study establish that mutations in predisposition genes are associated with lower risks of breast cancer in the general population than in enriched populations including high-risk families and young onset, bilateral, or multiple primary cases. We anticipate that the results from this study will inform cancer screening and other risk management strategies for women in the general population with mutations in predisposition genes. Citation Format: Fergus J Couch, Chunling Hu, Steven N Hart, Rohan Gnanaolivu, Kun Y Lee, Jie Na, Chi Gao, Nicholas J Boddicker, Bruce Eckloff, Raed Samara, Josh Klebba, Christine B Ambrosone, Hoda Anton-Culver, Paul Auer, Leslie Bernstein, Mia M Gaudet, Christopher Haiman, Esther M John, Susan Neuhausen, Janet E Olson, Tuya Pal, Julie R Palmer, Dale P Sandler, Jack A Taylor, Amy Trentham-Dietz, Celine M Vachon, Clarice Weinberg, Song Yao, Jeffrey N Weitzel, David E Goldgar, Susan M Domchek, Katherine L Nathanson, Peter Kraft, Eric C Polley. Population-based breast cancer risk estimates for predisposition gene mutations: Results from the CARRIERS study [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr PD3-01.

Abstract Background: Invasive lobular carcinoma (ILC) of the breast accounts for approximately 10-15% of all histologic subtypes of breast cancer. However, apart from germline mutations in CDH1, the contributions of mutations in other cancer predisposition genes to ILC are largely unknown. Methods: The study population included 3,893 women with ILC of the breast who underwent germline multigene panel testing of cancer predisposition genes at Ambry Genetics from March 2012 to December 2016. The prevalence of predisposition gene mutations in women with ILC was assessed relative to 38,004 women with invasive ductal carcinoma (IDC) who underwent germline multigene panel testing at Ambry Genetics during the same timeframe. Associations between mutations in each gene and risk of ILC were evaluated using reference controls. Results: Of the 3,893 women with ILC included in this study, 70.7% were non-Hispanic white, 65.1% had a family history of breast cancer, and 96.4% had estrogen receptor (ER) positive breast cancer. The overall frequency of germline mutations in cancer predisposition genes (ATM, BARD1, BRCA1, BRCA2, BRIP1, CDH1, CDKN2A, CHEK2, MLH1, MRE11A, MSH2, MSH6, NBN, NF1, PALB2, PTEN, RAD51C, RAD51D and TP53) was 7.8% for women with ILC and 10.7% for women with IDC. Mutations in CDH1 were significantly more frequent (Odds Ratio (OR)=12.1, p&lt;0.001), whereas mutations in BRCA1 (OR=0.1; p&lt;0.001) and PALB2 (OR=0.4; p&lt;0.001)were less frequent in women with ILC, compared to women with IDC. When restricting the analysis to ER positive cases only, significant differences in frequencies of ATM (OR=0.6, p=0.03), BRCA1 (OR=0.3, p&lt;0.001), CDH1 (OR=7.7, p&lt;0.001) and PALB2 (OR=0.3, p&lt;0.001) were observed in ILC compared to IDC. By comparison with gnomAD reference controls, ATM, BRCA2, CDH1, CHEK2, PALB2 and PTEN mutations were significantly associated with an increased risk of lLC and IDC (OR&gt;2), NBN mutations were associated with ILC (OR=3.4; p&lt;0.001) but not IDC, and BRCA1 mutations were not associated with increased risk of ILC. Among 274 women with ILC over the age of 50 without history of any other cancer or a family history of breast or ovarian cancer, 14 (5.1%) were found to carry a mutation in one of the genes evaluated. Conclusions: In a large cohort of patients with ILC, the frequencies of germline mutations in cancer predisposition genes were similar between ILC and IDC except for ATM, BRCA1, CDH1 and PALB2. Six genes, in addition to CDH1, were found to be germline predisposition genes for ILC. A substantial proportion of patients older than 50 with ILC without a family history of breast or ovarian cancer or a personal history of other cancers were found to be mutation carriers. These findings improve our understanding of germline genetic drivers of ILC and suggest that multigene genetic testing should be considered for all ILC patients. Citation Format: Siddhartha Yadav, Holly LaDuca, Eric C. Polley, Chunling Hu, Steven N. Hart, Jie Na, Rohan Gnanaolivu, Brandon Smith, David E Goldgar, Tina Pesaran, Jill S. Dolinsky, Fergus J. Couch. Germline mutations in cancer predisposition genes in patients with invasive lobular carcinoma of the breast [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P6-08-04.