Left ventricular hypertrophy: a surrogate end point or correlate of cardiovascular events in kidney disease?

Prentice [1] offered a statistical definition of a surrogate endpoint as ‘a response variable for which a test of the null hypothesis of no relationship to the treatment groups under comparison is also a valid test of the corresponding null hypothesis based on the true endpoint’. A National Institutes of Health (NIH) working group generated less quantitatively rigorous definitions that contrasted clinical and surrogate endpoints [2]. A ‘clinical endpoint’ is ‘[a] characteristic or variable that reflects how a patient feels, functions, or survives’, and so is related to the surrogate, ‘[a] biomarker that is intended to substitute for a clinical endpoint’. Moreover, ‘[a] surrogate endpoint is expected to predict clinical benefit (or harm or lack of benefit or harm) based on epidemiologic, therapeutic, pathophysiologic, or other scientific evidence’ [2]. While Prentice’s definition of surrogate requires the surrogate to account for the entirety of the effect of an intervention on a clinical endpoint, the NIH definition is more lenient. There must be clear distinction between a surrogate endpoint and a correlate. While an appropriate surrogate necessarily correlates with a clinical endpoint, all correlates are not effective surrogates [3]. For example, fever is a clinical correlate of serious vascular access infection, but is not an adequate surrogate for demonstrating the ability of an intervention to attenuate access-related sepsis. So, all surrogate measures are clinical correlates, but the converse is far from true. Surrogate endpoints are attractive for clinical trials, especially if the surrogate outcome occurs sooner, more often, or is easier to measure. The medical literature is replete with examples of putative surrogates that have failed to adequately predict clinical endpoints [4,5]. For example, reducing ventricular ectopy does not necessarily correlate with improved mortality. In hindsight, these measures were clinical correlates alone. This sobering limitation has been observed even in situations in which a sound chain of logic supports the use of the measure as a surrogate.