Statistics from Altmetric.com
The story of epidemiologists and surrogate variables is one of a classic love–hate relationship. Epidemiologists love surrogates because many studies would not be possible without them. When events are rare or the time between an intervention and the outcome is long, surrogate variables are used to fill the gap. Epidemiologists hate surrogate variables because they introduce an additional step in the chain between intervention (or exposure) and outcome, and thus an additional source of error. To introduce surrogates means to increase uncertainty — and, perhaps, an error.
A recent “collection of misleading surrogate end points” reminds us that caution is warranted: examples from this collection include the Cardiac Arrhythmia Suppression Trial in which a group of patients with asymptomatic or mildly symptomatic ventricular arrhythmia after myocardial infarction were treated with encainide or flecainide. Both drugs are used to suppress ventricular arrhythmia (the surrogate end point), but they were actually associated with excess mortality (the relevant outcome). In another case, patients treated with fluoride developed the desired increase in bone mineral density at the lumbar spine (the surrogate end point) but suffered from a higher rate of vertebral fractures than the control group.1 Assessing whether surrogate variables actually measure what they purport to measure (ie, for their validity) is therefore required. Available instruments, however, can only test if a surrogate correctly indicates the direction of an effect — that …
Competing interests None declared.
Provenance and peer review Commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.