Cancer Biostatistics on Clinical and Preclinical Cancer Research (Brain and HPV/HIV Associated Cancers)
At Barrow Neurological Institute and the Ivy Brain Tumor Center, we have performed phase 0 trials of drug treatments for brain tumors. Our clinical trials are mostly phase 0 or phase 0/2 studies (clinical and preclinical studies):
- A phase 0/2 study of Abemaciclib (CDK4 and 6 inhibitor) in Grade 3 meningioma patients,
- A phase 0/2 studies of Pamiparib in new and recurrent glioblastoma patients,
- A phase 0 trigger trial on Niraparib for newly diagnosed glioblastoma patients.
On-going activities:
- Develop a novel study design to utilize Phase 0 trigger trials to maximize the success in a phase 3 neuro-oncology trial
- Evaluation of phase 0 paradigm in brain tumor studies
- Determination of threshold for pharmacokinetic (PK) and pharmacodynamic (PD) biomarkers.
- Utilization of Phase 2 single-arm trials using two-stage designs with adaptive design
- Statistics on Phase 0 and Phase 0/2 trials
Population-based, Data-driven Research using Public Access Big Data (NIS, SEER, BRFSS, NHANES, TCGA, etc.)
Our laboratory has performed exploratory surveillance studies to identify health disparities by examining longitudinal trajectories or surveillance in health outcomes. My recent paper explored age-specific cervical cancer incidences in the Southern United States (2008-2012, from SEER data) and concluded that increasing incidence rates for older women (65+) need further research to determine whether screening should continue over age 65. Another recent paper explored the longitudinal trajectories of HPV vaccine adaption rates (2008-2016, from NIS-Teen data) and concluded that latent growth modeling and growth mixture models can be an effective tool for capturing disparities in heterogeneity among states.
On-going studies:
- Disparities in treatment on glioblastoma/glioma using SEER/NTCR databases
- Latent class growth modeling to identify hidden patterns in longitudinal survival outcomes
- Statistical genetics and clinical research using machine learning and network analysis
Mendelian Randomization Analysis
We aimed to estimate the effect of long-term exposure to LDL-C on the risk of coronary heart disease. Our laboratory is developing quantitative methods using Mendelian randomization to perform causal inference in observational studies.
Statistical Genetics and Clinical Research using Machine Learning and Network Analysis
Substantial uncertainty exists as to whether combining multiple disease-associated SNPs into a genotype risk score (GRS) can improve the ability to predict the risk of disease in a clinically relevant way. Mixed results have been obtained in studies of complex disease prediction through the combination of multiple disease-associated single nucleotide polymorphisms into a count-based GRS. Our research evaluated genetic risk score methods for predicting dichotomous outcomes as an alternative approach for identifying a combined multiplicative effect of interactions.
Statistical Modeling for Longitudinal Biomarkers and Diagnostic Rules on Early Detection within a Fully Bayesian Framework
A longitudinal biomarker can be monitored over time for changes that may be associated with changes in disease status. We have developed statistical models that led to the development of better diagnostic rules for early detection of complex diseases with higher sensitivity and specificity.
A model with a change point that is the onset of disease within fully Bayesian framework was constructed. This diagnostic rule was developed based on the posterior probability rule, which uses a whole series of biomarker measurements, not a single point like threshold rules. We will continue to enhance statistical methods and diagnostic rules for longitudinal biomarkers. We are also interested in developing latent class growth modeling under a fully Bayesian framework.
