A project led by Dr. Ali Jalali, assistant professor of population health sciences, titled “Using All of Us to Enhance the Generalizability of Evidence Derived from Randomized Controlled Trials for Substance Use Disorder Treatment Interventions,” was named a winner of the “Unlocking Insights: Analyzing All of Us Data for Drug Addiction Research” Challenge. Launched by the National Institute on Drug Abuse, this data-science competition invited researchers to leverage data from the National Institutes of Health’s All of Us Research Program to uncover novel patterns and relationships relevant to drug addiction.
The All of Us Research Program comprises one of the largest and most comprehensive health databases in history, with data from more than 870,000 participants across all 50 states in the US. Uniquely, All of Us reflects real-world variability that is often missing from biomedical research, including factors like ancestry, geography, age, and lived experience. This variability makes All of Us particularly valuable for understanding substance use disorders and addiction.
For their study, Dr. Jalali and colleagues analyzed data from a large randomized controlled trial (RCT) alongside the All of Us dataset to estimate how two medications for opioid use disorder (OUD), extended-release naltrexone and sublingual buprenorphine-naloxone, may perform in routine care outside the clinical trial setting. “This work uses a data-fusion technique that allows us to apply statistical conclusions from one study population to another,” explained Dr. Jalali. “The goal is to better understand how findings from a more restricted clinical trial population might translate to a broader and more representative population. If we rely solely on data from RCTs, we may not always capture how treatment effects translate to patient outcomes in other settings.”
The original trial, which included 570 adults with OUD recruited from eight detoxification centers across the US, found that sublingual buprenorphine-naloxone had higher initiation rates and was more cost-effective across multiple outcomes than extended-release naltrexone. However, when Dr. Jalali and colleagues applied those findings to the broader population represented in the All of Us data, estimated treatment benefits were smaller and, in some cases, changed direction. These results suggest that the original trial results may not fully reflect real-world effectiveness when applied to a broader population of individuals with OUD.
Results further indicate that patient characteristics, access to care, and health-related social needs may substantially influence the real-world effectiveness of pharmacological interventions. Since participants in RCTs are more likely to be white, younger, male, and more consistently insured than All of Us patients, RCT estimates alone are less likely to capture the effectiveness of medications for OUD in the broader US population. As such, this study serves as a proof-of-concept demonstrating that data fusion, paired with the breadth of All of Us data, can extend trial-based evidence to external populations initiating medication for OUD treatment.
“When there are populations that are challenging to study using traditional, single-study analytic approaches, we need creative solutions to address that challenge,” said Dr. Jalali. “Data fusion methods provide one avenue of extending, or as I like to refer to it, recycling, RCT data to answer relevant, timely research questions affecting policy and public health.”
Dr. Jalali would like to acknowledge the contributions of the biostatisticians on his team for their expertise and ongoing contributions, including Caroline Andy, research biostatistician III; Colby Lewis V, research biostatistician II; and Rachel Heise, research biostatistician II. He also extends gratitude to his research coordinator, Catherine Rabin, and to Dr. Sean Murphy, professor of population health sciences, for his guidance on the project.
Looking ahead, Dr. Jalali and his team plan to expand this line of research, with this study serving as a first step toward deeper integration of RCT with real-world health care databases in addiction research. “Our aim is to generate evidence that is more useful for large-scale public health decision-making,” said Dr. Jalali. “This proof of concept shows that when important policy decisions are at stake, we need the most accurate and representative evidence possible, which in some cases requires the analysis of multiple data sources.”
- Highlights