Citations

Abstract PURPOSE. Increased mammographic percent density (MPD) is a strong independent risk factor for developing breast cancer. Previous studies have shown that tissue sound speed, derived from ultrasound tomography, is a surrogate biomarker of MPD. We examined associations of sound speed and MPD with breast cancer risk in a case-control study. METHOD AND MATERIALS. We evaluated breast cancer risk associated with sound speed and MPD in a case-control study involving 59 participants with recent breast cancer diagnoses (cases, aged 30-70 years) and 150 participants with no history of breast cancer (controls), who were matched to cases on age, race, and menopausal status. The cases and controls were imaged with both ultrasound tomography (UST) and mammography. In cases, breast density was measured pre-treatment in the contralateral breast to avoid potential influences of tumor-related changes on MPD or sound speed. In controls, a randomly selected breast was imaged. The ultrasound tomography images were used to estimate the volume averaged sound speed of the breast, and the Cumulus software package was applied to mammograms to determine MPD. Odds Ratios (ORs) and 95% Confidence Intervals (CIs), adjusted for matching factors, were calculated for the relation of quartiles of MPD and sound speed with breast cancer risk. OR differences were tested using a bootstrap approach. RESULTS. MPD was associated with elevated breast cancer risk compared to controls, consistent with previous studies, although the trend did not reach statistical significance (OR per quartile=1.28, 95%CI: 0.95, 1.73; ptrend=0.10). In contrast, elevated sound speed was significantly associated with increased breast cancer risk in a dose-response fashion (OR per quartile=1.79, 95%CI: 1.30, 2.48; ptrend=0.0004). The OR-trend for sound speed was statistically significantly different from that observed for MPD (p=0.01). DISCUSSION. Our case-control study showed that increasing quartiles of whole breast sound speed were consistently and more strongly associated with increasing breast cancer risk than quartiles of MPD. These results show promise for UST’s role in breast cancer risk stratification. CONCLUSION. Elevated breast density strongly increases breast cancer risk. UST has the potential to provide a more accurate, non-ionizing method for assessing breast density and its associated breast cancer risk. Citation Format: Mark Sak, Neb Duric, Ruth Pfeiffer, Mark Sherman, Peter Littrup, Michael Simon, David Gorski, Teri Albrecht, Haythem Ali, Rachel Brem, Rachel Brem, Sharon Fan, Gretchen Gierach. Tissue sound speed is more strongly associated with breast cancer risk than mammographic percent density: A comparative case-control 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 P3-08-28.

Abstract PURPOSE. Studies have shown that a decrease in mammographic density (MD) or lowering of background parenchymal enhancement (BPE) on MRI after initiation of tamoxifen therapy predicts a favorable response in the preventive or adjuvant settings. However, performing serial mammograms poses radiation concerns, while serial MRIs carry high cost as well as risk of multiple Gadolinium doses. Previous studies have shown that tissue sound speed, derived from whole breast ultrasound tomography measurements, is a surrogate biomarker of MD. Ultrasound is ideal for performing serial measurements because it is fast and poses almost no risks. The purpose of this study was to evaluate repeated measures of the sound speed biomarker at 3, 6 and 12-months following tamoxifen initiation. METHOD AND MATERIALS. We performed a case-control comparison involving 74 participants referred by a health professional to undergo tamoxifen therapy (cases) and 150 matched participants with no history of breast cancer (controls). The cases were scanned with ultrasound tomography at baseline (i.e. before start of tamoxifen therapy), and then at 3, 6 and 12 months after tamoxifen initiation. Controls were scanned at baseline and 12 months. In the case group, sound speed was measured pre-treatment in the contralateral breast to avoid potential influences of tumor-related changes on density. In the control group, a single randomized breast was scanned. A pairwise t-test was used to assess differences in sound speed over time and between cases and controls. RESULTS. There was a steady decline in sound speed over the 12-month period for women undergoing tamoxifen therapy (mean(SD): -3.0(8.2) m/s; P=0.001). Furthermore, significant sound speed reductions were observed as early as 4-6 months after tamoxifen initiation (mean(SD): -2.1(6.8) m/s; P=0.008). In contrast, the controls demonstrated no significant change in sound speed over a 12-month period, and the difference between case-control groups was statistically significant (P=0.0009). DISCUSSION. Breast sound speed decreases rapidly after tamoxifen initiation while it remains steady for the control group. These results suggest that sound speed may be a biomarker of response to tamoxifen. Further studies are needed to assess whether this observation can predict clinical response. CONCLUSION. Ultrasound tomography may have utility in monitoring breast sound speed change as a potential biomarker of clinical tamoxifen response. Citation Format: Neb Duric, Mark Sak, Mark Sherman, Ruth Pfeiffer, Sharon Fan, Michael Simon, David Gorski, Teri Albrecht, Haythem Ali, Peter Littrup, Rachel Brem, Gretchen Gierach. Tissue sound speed: A novel imaging biomarker for measuring tamoxifen response [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 P2-11-11.

INTRODUCTION The cerebral cortex underlies our complex cognitive capabilities. Variations in human cortical surface area and thickness are associated with neurological, psychological, and behavioral traits and can be measured in vivo by magnetic resonance imaging (MRI). Studies in model organisms have identified genes that influence cortical structure, but little is known about common genetic variants that affect human cortical structure. RATIONALE To identify genetic variants associated with human cortical structure at both global and regional levels, we conducted a genome-wide association meta-analysis of brain MRI data from 51,665 individuals across 60 cohorts. We analyzed the surface area and average thickness of the whole cortex and 34 cortical regions with known functional specializations. RESULTS We identified 369 nominally genome-wide significant loci ( P < 5 × 10 −8 ) associated with cortical structure in a discovery sample of 33,992 participants of European ancestry. Of the 360 loci for which replication data were available, 241 loci influencing surface area and 66 influencing thickness remained significant after replication, with 237 loci passing multiple testing correction ( P < 8.3 × 10 −10 ; 187 influencing surface area and 50 influencing thickness). Common genetic variants explained 34% (SE = 3%) of the variation in total surface area and 26% (SE = 2%) in average thickness; surface area and thickness showed a negative genetic correlation ( r G = −0.32, SE = 0.05, P = 6.5 × 10 −12 ), which suggests that genetic influences have opposing effects on surface area and thickness. Bioinformatic analyses showed that total surface area is influenced by genetic variants that alter gene regulatory activity in neural progenitor cells during fetal development. By contrast, average thickness is influenced by active regulatory elements in adult brain samples, which may reflect processes that occur after mid-fetal development, such as myelination, branching, or pruning. When considered together, these results support the radial unit hypothesis that different developmental mechanisms promote surface area expansion and increases in thickness. To identify specific genetic influences on individual cortical regions, we controlled for global measures (total surface area or average thickness) in the regional analyses. After multiple testing correction, we identified 175 loci that influence regional surface area and 46 that influence regional thickness. Loci that affect regional surface area cluster near genes involved in the Wnt signaling pathway, which is known to influence areal identity. We observed significant positive genetic correlations and evidence of bidirectional causation of total surface area with both general cognitive functioning and educational attainment. We found additional positive genetic correlations between total surface area and Parkinson’s disease but did not find evidence of causation. Negative genetic correlations were evident between total surface area and insomnia, attention deficit hyperactivity disorder, depressive symptoms, major depressive disorder, and neuroticism. CONCLUSION This large-scale collaborative work enhances our understanding of the genetic architecture of the human cerebral cortex and its regional patterning. The highly polygenic architecture of the cortex suggests that distinct genes are involved in the development of specific cortical areas. Moreover, we find evidence that brain structure is a key phenotype along the causal pathway that leads from genetic variation to differences in general cognitive function. Identifying genetic influences on human cortical structure. ( A ) Measurement of cortical surface area and thickness from MRI. ( B ) Genomic locations of common genetic variants that influence global and regional cortical structure. ( C ) Our results support the radial unit hypothesis that the expansion of cortical surface area is driven by proliferating neural progenitor cells. ( D ) Cortical surface area shows genetic correlation with psychiatric and cognitive traits. Error bars indicate SE. IMAGE CREDITS: (A) K. COURTNEY; (C) M. R. GLASS

Abstract Objectives To adapt and evaluate a deep learning language model for answering why-questions based on patient-specific clinical text. Materials and Methods Bidirectional encoder representations from transformers (BERT) models were trained with varying data sources to perform SQuAD 2.0 style why-question answering (why-QA) on clinical notes. The evaluation focused on: (1) comparing the merits from different training data and (2) error analysis. Results The best model achieved an accuracy of 0.707 (or 0.760 by partial match). Training toward customization for the clinical language helped increase 6% in accuracy. Discussion The error analysis suggested that the model did not really perform deep reasoning and that clinical why-QA might warrant more sophisticated solutions. Conclusion The BERT model achieved moderate accuracy in clinical why-QA and should benefit from the rapidly evolving technology. Despite the identified limitations, it could serve as a competent proxy for question-driven clinical information extraction.

Mammographic percent density (MPD) is an independent risk factor for developing breast cancer, but its inclusion in clinical risk models provides only modest improvements in individualized risk prediction, and MPD is not typically assessed in younger women because of ionizing radiation concerns. Previous studies have shown that tissue sound speed, derived from whole breast ultrasound tomography (UST), a non-ionizing modality, is a potential surrogate marker of breast density, but prior to this study, sound speed has not been directly linked to breast cancer risk. To that end, we explored the relation of sound speed and MPD with breast cancer risk in a case-control study, including 61 cases with recent breast cancer diagnoses and a comparison group of 165 women, frequency matched to cases on age, race, and menopausal status, and with a recent negative mammogram and no personal history of breast cancer. Multivariable odds ratios (ORs) and 95% confidence intervals (CIs) were estimated for the relation of quartiles of MPD and sound speed with breast cancer risk adjusted for matching factors. Elevated MPD was associated with increased breast cancer risk, although the trend did not reach statistical significance (OR per quartile = 1.27, 95% CI: 0.95, 1.70; ptrend = 0.10). In contrast, elevated sound speed was significantly associated with breast cancer risk in a dose–response fashion (OR per quartile = 1.83, 95% CI: 1.32, 2.54; ptrend = 0.0003). The OR trend for sound speed was statistically significantly different from that observed for MPD (p = 0.005). These findings suggest that whole breast sound speed may be more strongly associated with breast cancer risk than MPD and offer future opportunities for refining the magnitude and precision of risk associations in larger, population-based studies, including women younger than usual screening ages.

The association of progesterone/progesterone metabolites with elevated mammographic breast density (MBD) and delayed age-related terminal duct lobular unit (TDLU) involution, strong breast cancer risk factors, has received limited attention. Using a reliable liquid chromatography-tandem mass-spectrometry assay, we quantified serum progesterone/progesterone metabolites and explored cross-sectional relationships with MBD and TDLU involution among women, ages 40–65, undergoing diagnostic breast biopsy. Quantitative MBD measures were estimated in pre-biopsy digital mammograms. TDLU involution was quantified in diagnostic biopsies. Adjusted partial correlations and trends across MBD/TDLU categories were calculated. Pregnenolone was positively associated with percent MBD-area (MBD-A, rho: 0.30; p-trend = 0.01) among premenopausal luteal phase women. Progesterone tended to be positively associated with percent MBD-A among luteal phase (rho: 0.26; p-trend = 0.07) and postmenopausal (rho: 0.17; p-trend = 0.04) women. Consistent with experimental data, implicating an elevated 5α-pregnanes/3α-dihydroprogesterone (5αP/3αHP) metabolite ratio in breast cancer, higher 5αP/3αHP was associated with elevated percent MBD-A among luteal phase (rho: 0.29; p-trend = 0.08), but not postmenopausal women. This exploratory analysis provided some evidence that endogenous progesterone and progesterone metabolites might be correlated with MBD, a strong breast cancer risk factor, in both pre- and postmenopausal women undergoing breast biopsy. Additional studies are needed to understand the role of progesterone/progesterone metabolites in breast tissue composition and breast cancer risk.

Abstract Purpose: With increasing incidence of renal mass, it is important to make a pretreatment differentiation between benign renal mass and malignant tumor. We aimed to develop a deep learning model that distinguishes benign renal tumors from renal cell carcinoma (RCC) by applying a residual convolutional neural network (ResNet) on routine MR imaging. Experimental Design: Preoperative MR images (T2-weighted and T1-postcontrast sequences) of 1,162 renal lesions definitely diagnosed on pathology or imaging in a multicenter cohort were divided into training, validation, and test sets (70:20:10 split). An ensemble model based on ResNet was built combining clinical variables and T1C and T2WI MR images using a bagging classifier to predict renal tumor pathology. Final model performance was compared with expert interpretation and the most optimized radiomics model. Results: Among the 1,162 renal lesions, 655 were malignant and 507 were benign. Compared with a baseline zero rule algorithm, the ensemble deep learning model had a statistically significant higher test accuracy (0.70 vs. 0.56, P = 0.004). Compared with all experts averaged, the ensemble deep learning model had higher test accuracy (0.70 vs. 0.60, P = 0.053), sensitivity (0.92 vs. 0.80, P = 0.017), and specificity (0.41 vs. 0.35, P = 0.450). Compared with the radiomics model, the ensemble deep learning model had higher test accuracy (0.70 vs. 0.62, P = 0.081), sensitivity (0.92 vs. 0.79, P = 0.012), and specificity (0.41 vs. 0.39, P = 0.770). Conclusions: Deep learning can noninvasively distinguish benign renal tumors from RCC using conventional MR imaging in a multi-institutional dataset with good accuracy, sensitivity, and specificity comparable with experts and radiomics.

Drug safety is a severe clinical pharmacology and toxicology problem that has caused immense medical and social burdens every year. Regretfully, a reproducible method to assess drug safety systematically and quantitatively is still missing. In this study, we developed an advanced machine learning model for de novo drug safety assessment by solving the multilayer drug‐gene‐adverse drug reaction (ADR) interaction network. For the first time, the drug safety was assessed in a broad landscape of 1,156 distinct ADRs. We also designed a parameter ToxicityScore to quantify the overall drug safety. Moreover, we determined association strength for every 3,807,631 gene‐ADR interactions, which clues mechanistic exploration of ADRs. For convenience, we deployed the model as a web service ADRAlert‐gene at http://www.bio-add.org/ADRAlert/. In summary, this study offers insights into prioritizing safe drug therapy. It helps reduce the attrition rate of new drug discovery by providing a reliable ADR profile in the early preclinical stage.

Abstract Background: The Checklist for Early Recognition and Treatment of Acute Illness and Injury (CERTAIN) study is a before-and-after clinical quality improvement study involving 20 countries. Our aim was to describe the challenges and solutions for study implementation across various intensive care units.Methods: After local institutional review board approval, each study center had a setup period before patient data were collected. Each center was required to accomplish 4 stages: baseline data collection, remote simulation training, local implementation, and postimplementation data collection and maintenance of process of care. We measured the time required to complete each study stage and used the Define, Measure, Analyze, Improve, and Control (DMAIC) framework to identify the challenges encountered during each stage and guide improvement strategies to facilitate implementation of CERTAIN.Results: Between June 2012 and December 2017, 55 centers from 20 countries were recruited to participate. The time to complete each stage at different centers varied significantly. The most common domains in which challenges occurred were 1) leadership support and team building, 2) communication, 3) work environment culture, 4) language barriers, 5) infrastructure and technology, and 6) sustainability. Obstacles in these 6 areas delayed reaching major milestones of the study. Different strategies were deployed to overcome these barriers: engaging leadership for support, using native-speaking clinicians during training sessions, using various communication platforms, providing personal computers to local hospitals, making in-person site visits, offering flexible training time, and supporting local centers’ grant applications. Centers that enrolled during the control phase, compared with those that enrolled earlier, had significantly shorter times and less variation in times for each study stage and for the total study time.Conclusions: By using the DMAIC method, we identified the predominant 6 domains in which challenges occurred during implementation of the global CERTAIN study and developed targeted strategies to overcome the challenges and shorten the duration of the study. The DMAIC method would be useful in improving the efficiency of other large-scale multinational quality improvement projects.

AbstractIntroductionA recent study reported that diffuse left ventricular (LV) fibrosis is a predictor of atrial fibrillation (AF) recurrence following catheter ablation, by measuring postcontrast cardiac T1 (an error prone metric as per the 2017 Society for Cardiovascular Magnetic Resonance consensus statement) using an inversion‐recovery pulse sequence (an error prone method in arrhythmia) in AF ablation candidates. The purpose of this study was to verify the prior study, by measuring extracellular volume (ECV) fraction (an accurate metric) using a saturation‐recovery pulse sequence (accurate method in arrhythmia).Methods and ResultsThis study examined 100 AF patients (mean age = 62 ± 11 years, 69 males and 31 females, 67 paroxysmal [pAF] and 33 persistent [peAF]) who underwent a preablation cardiovascular magnetic resonance (CMR) exam. LV ECV and left atrial (LA) and LV functional parameters were quantified using standard analysis methods. During an average follow‐up period of 457 ± 261 days with 4 ± 3 rhythm checks per patient, 72 patients maintained sinus rhythm. Between those who maintained sinus rhythm (n = 72) and those who reverted to AF (n = 28), the only clinical characteristic that was significantly different was age (60 ± 12 years vs 66 ± 9 years); for CMR metrics, neither mean LV ECV (25.1 ± 3.3% vs 24.7 ± 3.7%), native LV T1 (1093.8 ± 73.5 ms vs 1070.2 ± 115.9 ms), left ventricular ejection fraction (54.1 ± 11.2% vs 55.7 ± 7.1%), nor LA end diastolic volume/body surface area (42.4 ± 14.8 mL/m2 vs 43.4 ± 19.6 mL/m2) were significantly different (P ≥ .23). According to Cox regression tests, none of the clinical and imaging variables predict AF recurrence.ConclusionNeither LV ECV nor other CMR metrics predict recurrence of AF following catheter ablation.