Intended for healthcare professionals

Rapid response to:


The idolatry of the surrogate

BMJ 2011; 343 doi: (Published 28 December 2011) Cite this as: BMJ 2011;343:d7995

Rapid Response:

Re: The idolatry of the surrogate

Yudkin, Lipska, and Montori are right to point to some problems with the use of surrogate end points, which they do largely in relation to the treatment of diabetes mellitus. There are, however, some problems with their overall analysis.

1. They do not clearly distinguish between surrogate end points for benefits and those for harms. It is reasonable to expect, as they suggest, that actual harms should be studied in trials, rather than surrogates. However, although they are correct to stress that when surrogates are used to assess long-term benefits, it is better to study multiple end points, it is often not possible to study actual benefits, particularly in diseases in which the final end points are intermittent, such as epilepsy, in which natural variability in the individual case confounds one's interpretation of the possible effects of interventions, and in population studies of diseases in which long-term treatment is needed to prevent final end points. Hypertension is a good example of the latter. Yudkin et al. cite Law et al.1 as showing that antihypertensive drugs may not reduce the long-term risk of stroke, but that is not what the latter showed at all in their meta-analysis of 147 trials. To quote from their paper: "... all the classes of blood pressure lowering drugs have a similar effect in reducing CHD events and stroke for a given reduction in blood pressure"; beta-blockers gave extra protection shortly after a myocardial infarction. Blood pressure is clearly an excellent surrogate for the long-term benefits of treating hypertension.

2. Their taxonomy of types of surrogate is incomplete. The details are too complex to describe here, but elsewhere I have described five different types of surrogates, and I do not think that I exhausted all the possibilities.2

3. Their use of the term "hard" to refer to final endpoints is misleading. The terms "hard" and "soft" in relation to types of data have not been widely discussed, but a definitive description was given by Feinstein some years ago.3 Briefly, he outlined three characteristics of “hard” data: that they are acquired objectively not subjectively; that they are preservable (an X-ray can be reassessed and a blood measurement can be repeated); and they are measured on a dimensional rather than a ordinal or nominal scale. It is clear that many surrogates are “hard” end points by this definition. It would be better to contrast surrogate end points with "final" end points.


1. Law MR, Morris JK, Wald NJ. Use of blood pressure lowering drugs in the prevention of cardiovascular disease: meta-analysis of 147 randomised trials in the context of expectations from prospective epidemiological studies. BMJ 2009; 338: b1665.

2. Aronson JK. Biomarkers and surrogate endpoints in monitoring therapeutic interventions. In: Glasziou P, Irwig L, Aronson JK, editors. Evidence-based Medical Monitoring: From Principles to Practice. Oxford: Wiley-Blackwell Ltd, 2008: 48-62.

3. Feinstein AR. Clinical biostatistics. XLI. Hard science, soft data, and the challenges of choosing clinical variables in research. Clin Pharmacol Ther 1977; 22(4): 485-98.

Competing interests: No competing interests

17 January 2012
Jeffrey K Aronson
Clinical pharmacologist
University of Oxford
Department of Primary Health Care