Research Focus

Dr. Wilkens has a DrPH from the University of North Carolina in Biostatistics and has worked in health research for over 30 years including over 25 years in cancer research. Much of this effort has focused on prevention in the domains of epidemiology and intervention research. A primary focus for Dr. Wilkens at UHCC has been in the quantification of cancer incidence and mortality rates for the multiethnic populations of Hawaiʻi and the US affiliated Pacific, and studying the underlying causes of differences in cancer risks between ethnic groups. Genetic as well as lifestyle factors, including diet, physical activity, smoking and neighborhood environment, have been considered. Dr. Wilkens has published over 300 publications from these efforts.

Dr. Wilkens was one of the original biostatisticians and is currently a Multiple Principal Investigator for the Multiethnic Cohort (MEC) study (U01 CA164973). The MEC was initiated in 1993 and has followed over 215,000 individuals from 5 major ethnic groups in Hawaiʻi and southern California for cancer and other diseases. Dr. Wilkens has vast experience with study design, data structure and data analysis for the MEC and its substudies. Dr. Wilkens leads the Data Management and Analysis Core for the P01 MEC Adiposity Phenotype study (P01 CA168530) that examines the role of body composition in cancer incidence.

Dr. Wilkens has worked in the broader Pacific. Research in the NCI-funded U54 UHCC – University of Guam Partnership (U54 CA143727) has focused on the quantification of cancer risk among ethnic groups in Guam and the role of modifiable factors, such as screening, diet and betel nut use, in cancer risk. The USDA-funded Children's Health Living Program (CHL) in the Pacific (NIFA 2011-68001-30335) quantified the prevalence of obesity in children in the US-affiliated Pacific and investigated the effect of a community-based intervention on obesity prevention.

Dr. Wilkens' primary methodological interest is in accurate measurement of exposure variables. Many of the exposures of interest to cancer are difficult to measure, such as diet and physical activity. Where possible identification of more accurate long-term biomarkers of exposure is important, such as a recently validated marker of heterocyclic aromatic amines in hair. However, there is often no preferred marker, but rather many approaches to measuring these constructs, each with unique strengths and limitations. For instance, diet can be measured through the self-reported instruments of food frequency questionnaires and dietary recalls, and through biomarkers. Biomarkers are considered less biased but generally only measure short-term exposure, and most markers do not solely reflect exposure from diet. Self-reported measures are subject to measurement error, which biases the disease association and reduces power. A better understanding of the underlying associations with disease is possible by appropriately combining data from various sources. A related topic is data integration. As high-dimensional data in various omics domains becomes increasingly available, a question arises as to how to integrate important markers from these different domains to best understand how biological processes influence disease. Newer statistical models that incorporate large sets of independent variables through constraints such as elastic net, as well as allowing for the explicit modeling of data structure, such as through a latent variable approach, hold promise to optimizing the use of data from sources with different measurement properties and scales.