Data Mining Heterogeneity of Treatment Effects on Patients with the Metabolic Syndrome

• Activity Number: 544
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Hua Fang and Jin Wang*+ and Bruce Barton and Honggang Wang and Yunsheng Ma
• Companies: Umass Medical School and UMass Dartmouth and University of Massachusetts Medical School and University of Massachusetts Dartmouth and University of Massachusetts Medical School
• Keywords: data mining ; heterogeneity of treatment effects ; clinical trials ; metabolic syndrome ; dietary quality

Abstract

Heterogeneity of treatment effects (HTE) refers to the fact that patients exposed to a treatment can experience very different outcomes. HTE commonly exists in complex, multi-component treatments. Standard approaches to characterizing treatments use simple "treated" and "non-treated" groups or subgroups based on arbitrarily determined cut-scores. The binary approach overlooks important treatment/exposure information, and sub-grouping generates spurious false-positive findings. We proposed a new methodological approach to HTE by incorporating a novel data mining method to account for individual behavioral variations over the course of a treatment, along with other important factors (e.g., demographics and pre-treatment risks). Specifically, we will apply our new method to a complex dietary-quality dataset, generated from a longitudinal NIH-funded random controlled trial for patients with metabolic syndromes. This method will be compared to approaches such as Bayesian and verified with simulations. This new approach has demonstrated its advantages in application areas such as substance use studies and if successful, will have potential generalizability in clinical trial studies.