Estimating the Hawthorne Effect in Real-World Blood Pressure Control Trials: An Analysis of the BP Home Trial

Results from blood pressure (BP) control interventional trials can inform clinicians and health systems pursuing better BP control, but they must be interpreted cautiously.

Deconstruct the observed drop in systolic BP (SBP) into components attributable to increased adherence to previously prescribed medications, regression to the mean, and initiation of new medications.

Secondary analysis of BP Home, a pragmatic randomized controlled trial.

Patients owning a smartphone who reported uncontrolled BP at their last clinic visit (> 145 mmHg) and a desire to lower their BP by > 10 mmHg.

In BP Home, participants were randomly assigned to receive one of two devices for self-measurement of BP and followed for up to 24 months via electronic health records (EHR).

The primary outcome was EHR-recorded office SBP. We fit SBP trajectories for each participant using linear mixed models, and estimated the contributions of medication adjustments, and increased adherence to pre-existing BP medications due to their enrollment in a research study (i.e., a Hawthorne effect). Regression to the mean was calculated for each participant as the difference between their last measured pre-enrollment SBP and their modeled SBP trajectory at enrollment.

Among participants taking BP medications at enrollment, we estimated an average immediate drop in SBP of -4.2 mmHg (95% CI, -5.1 to -3.3; p < 0.001) at enrollment, explainable by increased medication adherence following study enrollment. Starting a new medication class post-enrollment resulted in an SBP drop of -4.1 mmHg (95% CI, -5.3 to -2.8; p < 0.001). The average expected regression to the mean was -3.9 (95% CI, -4.4 to -3.3; p < 0.001).

A significant portion of BP reductions in trials may stem from increased adherence to pre-existing medications arising from enrollment in a research study and enhanced awareness of their elevated BP.

NCT03796689 (registered on 2019-01-04).