Regression Discontinuity Design: Simulation and Application in Two Cardiovascular Trials with Continuous Outcomes

Area of Science:

• Epidemiology
• Biostatistics
• Clinical Trials

Background:

• Regression discontinuity design (RDD) is gaining traction as an alternative to randomized controlled trials (RCTs) for treatment effect evaluation.
• RDD assigns treatment based on a threshold of an assignment variable, adjusting the effect in analysis.

Purpose of the Study:

• To compare treatment effect estimates from RDD with those from RCTs.
• To assess the validity and precision of RDD in real-world validation studies.

Main Methods:

• Performed simulations and a prospective validation study using RCT data as a reference.
• Estimated treatment effects using linear regression (linear terms, restricted cubic spline) and local linear regression.
• Compared RDD estimates against established RCT findings for cardiovascular outcomes and total cholesterol.

Main Results:

• In one study, RDD with restricted cubic spline yielded a blood pressure reduction estimate (-5.9 mmHg) comparable to RCT (-4.0 mmHg), but with wider confidence intervals.
• Local linear regression in RDD revealed different, localized effects.
• RDD estimates for total cholesterol were similar to RCTs but significantly less precise (six times lower).

Conclusions:

• RDD can provide similar treatment effect estimates to RCTs but relies on a crucial assumption of a global treatment effect.
• Researchers must consider the trade-off between potential recruitment advantages and significant precision loss with RDD compared to RCTs.
• Potential bias due to incorrect assumptions is a key concern when employing RDD.