Data from a UW-CTRI study known as the Wisconsin Smokers’ Health Study 2 has helped colleagues from other methodology centers and UW-CTRI to develop a variety of methodological innovations.
The latest novel approach was recently published in the journal Multivariate Behavioral Research.
In longitudinal studies, exposure to treatment and mediators (factors influenced by treatment) rarely stay constant. For example:
- Participants change how they adhere to medications over time.
- Stress levels and cravings fluctuate weekly, hourly, etc.
- Biological markers evolve over years.
So the research teams set out to account for these types of evolving factors, said UW-CTRI Director of Research Dr. Megan Piper.
“I have worked with Dr. Donna Coffman for years since she was a trainee at Dr. Linda Collins’ methodology center at Penn State,” Piper said.
Coffman, who now is at the University of South Carolina, and Piper are co-authors of the new paper by Dr. Yajnaseni Chakraborti (University of Pennsylvania).
The goal of the analyses was to use WSHS2 data as an example of a data set that has:
- A randomly assigned treatment (medication type).
- A time-varying mediator (adherence to medication use).
- An outcome (withdrawal).
- A time-varying confounder (stress).
“Our collaborators developed the necessary math to be able to analyze these difficult scenarios,” Piper said. “What they found was that treatment type did have an effect on daily withdrawal, and this was due, in part, to daily use of study medication and differed based on environmental stressors.”
They also found that the three medications (varenicline, patch or a combination of nicotine-replacement medications) did not differ in their impact on withdrawal. “This was mediated via medication use, which was similar across treatment conditions and may have been why we didn’t see treatment effects,” Piper said.
Researchers can also pinpoint key timing in a quit attempt when people on specific treatments are likely to be most vulnerable, thanks to time-varying effect models.
“Unfortunately, we have lots of data showing why these medications don’t exert differential effects,” Piper said.
Within its scope, this study serves as a preliminary framework for studying the causal structure of time-varying bio-behavioral processes.
Chakraborti Y, Yucel R, Piper ME, Mennis J, Alberg AJ, Baker TB, Coffman DL. (2026) Time-Varying Path-Specific Direct and Indirect Effects: A Novel Approach to Unpacking Dynamic Behavioral Processes with Application to Smoking Cessation. Multivariate Behavioral Research. 1–19. Online January 20, 2026.