Clinical Trial Emulation Leveraging Genetic Effects: A Proof-of-Concept Application to SPRINT
Santiago Clocchiatti-Tuozzo1, Cyprien Rivier2, Shufan Huo2, Ashkan Shoamanesh3, Hooman Kamel4, Santosh Murthy5, Adam De Havenon2, Lauren Sansing6, Thomas Gill7, Kevin Sheth8, Guido Falcone6
1Yale University, Department of Neurology, 2Yale University, 3McMaster University, 4Weill Cornell Medical College, 5Weill Cornell Medicine, 6Yale School of Medicine, 7Internal Medicine, Yale School of Medicine, 8Yale UniversityDivision of Neuro and Critical Care
Objective:
To present a genomic proof-of-concept emulation of the SPRINT randomized clinical trial (RCT).
Background:
RCTs are the gold standard for evaluating treatment effects, but they are costly and time-consuming. Therefore, selecting the most promising treatments for RCTs is crucial. Genetic variants, randomly distributed during meiosis, create natural experiments for traits they encode. Drugs backed by genomics are twice as likely to gain FDA approval, yet there have not been formal RCT emulations using genomics.
Design/Methods:
We conducted SPRINT-gen, a genomic emulation of SPRINT, using UK Biobank data. SPRINT-gen used the same sample size (n=9,361), mean systolic blood pressure (BP) difference between groups (15 mmHg), analytical approach (Cox regressions), and outcome (composite of stroke, myocardial infarction, coronary artery disease, or cardiovascular death) as SPRINT, which randomized hypertensive, non-diabetic participants into intensive (systolic BP<120 mmHg, n=4,678) and standard (systolic BP<140 mmHg, n=4,683) treatment groups. To emulate this and replicate treatment effects, SPRINT-gen employed a validated polygenic risk score for systolic BP, defining both intensive and standard treatment groups.
Results:
Like SPRINT, SPRINT-gen included 4,678 participants assigned to intensive treatment, 4,683 participants assigned to standard treatment (both mean age 63, 42% female), and a follow-up time of 4.8y. The number of events in the intensive treatment groups in SPRINT and SPRINT-gen was 243 (5.2%) vs. 220 (4.7%), respectively, and 319 (6.8%) vs. 270 (5.8%) in the standard treatment groups, respectively. The effect estimate of intensive vs. standard treatment on the composite outcome was consistent between SPRINT (HR:0.75, 95%CI: 0.64-0.89) and SPRINT-gen (HR:0.76, 95%CI: 0.62–0.92).
Conclusions:
Our genomic-based RCT emulation framework accurately reproduced the primary result of SPRINT. This framework could be valuable for simulating RCTs that target continuous and highly heritable physiological parameters. Genomic trial emulation may significantly improve the success rate of future RCTs.
Disclaimer: Abstracts were not reviewed by Neurology® and do not reflect the views of Neurology® editors or staff.