TABA Seminar and Networking Event
Location: 1004 Middlegate Rd, Mississauga, ON
Time: 14:00-16:00 PM EST, snacks and light refreshments provided
Agenda:
Presenter: Dr. Judy-Anne Chapman, PhD, P.Stat.
Title: Adjunctive statistical standardization of adjuvant ER and PgR in Canadian Cancer Trials
Group MA.27 postmenopausal breast cancer trial of exemestane versus anastrozole1
Abstract: American Society of Clinical Oncology/College of American Pathologists guidelines
recommend reporting estrogen receptor (ER) and progesterone receptor (PgR) as
positive with (1%-100%) staining. MA.27 (NCT00066573) was a phase III adjuvant trial
of exemestane versus anastrozole in postmenopausal women with early-stage breast
cancer. IHC ER and PgR. HSCORE and % positivity (%+) were centrally assessed by
machine-image quantitation, and statistically standardized following Box-Cox variance
stabilization transformations. MA.27 primary endpoint was MA.27 distant disease-free
survival (DDFS); secondary endpoint, event-free survival (EFS). Univariate survival was
described with Kaplan-Meier plots, and examined with Wilcoxon (Peto-Prentice) test
statistic. Adjusted Cox multivariable regressions utilized 2-sided Wald tests with nominal significance p<0.05. Of 7576 women accrued, 3048 women’s tumors had machine-
quantitated image analysis results. Higher statistically standardized ER and PgR HSCORE, and %+, were associated with better univariate DDFS and EFS (p<0.001). In multivariable assessments, ER HSCORE and %+ were not significantly associated (p=0.52-0.88) with DDFS in models with PgR, while higher PgR HSCORE and %+ were significantly associated with better DDFS (p=0.001) in models with ER. Adjunctive statistical standardization differentiated quantitated levels of ER and PgR. Patients with higher ER and PgR standardized units had superior DDFS compared with those with HSCOREs and %+ <-1.
Presenter: Dr. Divya Sharma, PhD
Title: Advanced Machine Learning Approaches for Multimodal Data Integration and Applications in Healthcare Research
Abstract: The integration of multimodal data—encompassing clinical assessments, imaging, molecular profiles, and patient-reported outcomes—represents a major challenge in biomedical research due to the diverse scales, formats, and complex interdependencies inherent to these data types. Traditional statistical approaches often falter in addressing such heterogeneity, as they rely on assumptions that cannot fully capture the rich interactions between modalities. In this talk I will discuss about advanced machine learning methodologies tailored for multimodal data integration, with a specific application to osteoarthritis research. By leveraging these techniques, we uncover meaningful patient subgroups and predict disease outcomes, demonstrating the ability to harness complementary information across diverse data sources. This talk will highlight how emerging machine learning methods overcome the limitations of conventional approaches, enabling more precise and data-driven strategies with applications across healthcare.
Presenter: Dr. Yutao Liu, PhD
Title: Bayesian Inference of Finite Population Quantiles for Skewed Survey Data Using Skew-Normal Penalized Spline Regression
Abstract: Skewed data are common in sample surveys. In probability proportional to size sampling, we propose two Bayesian model-based predictive methods for estimating finite population quantiles with skewed sample survey data. We assume the survey outcome to follow a skew-normal distribution given the probability of selection and model the location and scale parameters of the skew-normal distribution as functions of the probability of selection. To allow a flexible association between the survey outcome and the probability of selection, the first method models the location parameter with a penalized spline and the scale parameter with a polynomial function, while the second method models both the location and scale parameters with penalized splines. Using a fully Bayesian approach, we obtain the posterior predictive distributions of the nonsampled units in the population and thus the posterior distributions of the finite population quantiles. We show through simulations that our proposed methods are more efficient and yield shorter credible intervals with better coverage rates than the conventional weighted method in estimating finite population quantiles. We demonstrate the application of our proposed methods using data from the 2013 National Drug Abuse Treatment System Survey.
