Unintentional overestimation of an expected antihypertensive effect in drug and device trials: mechanisms and solutions

Abstract

In clinical practice we pay close attention to choosing an appropriate intervention for patients and performing it safely. We may put less thought into how to measure the effect without bias. Clinical practice involving noisy values such as blood pressure requires intelligent processing to avoid confusing patients, but applying such discretion in clinical trials may inadvertently disrupt quantification of the benefit of the intervention, in unblinded trials. In this article we explore two sources of bias, which label for convenience "big day bias" and "check once more bias", which can lead to unintentional exaggeration of benefits from both drugs and device-based treatment for hypertension. We show in this article why, as denervation trials become increasingly bias resistant, the reported effect size may reduce. If inadvertent bias affects patients who are denervated in the same way as it affects those receiving antihypertensive drugs, the most meticulous denervation trials may show the effect size falling from around 30 mmHg to around 10 mmHg. Some readers will doubt that these biases could be large enough to matter. We therefore include downloadable spreadsheets that any reader can use to explore how powerfully small biases affect the apparent effect sizes. The results may be surprising. A 10-mmHg reduction, without needing to adhere to an extra drug, would still substantially reduce events in the long term, but crucial to such reasoning is the reliable quantification of the blood pressure effect, free of bias.

Keywords: Bias (N05.715.350.150); Clinical trial (V03.200); Hypertension (C14.907.489); Research design (E05.581.500).