One of the most important factors in designing studies to evaluate new treatments for a particular disease is the choice of the endpoint used to assess treatment efficacy. Frequently, a true endpoint is costly or difficult to measure, requires a long-term follow-up or a large sample size.

In such cases, the use of the true endpoint increases the complexity and/or duration of the research. To overcome these problems, a potentially attractive solution is to replace the true endpoint by a surrogate endpoint, which is measured earlier, more conveniently, or more frequently. However, to be acceptable, a proposed surrogate needs to be evaluated and validated. The mere existence of association or correlation between a surrogate and the endpoint, although necessary, is not sufficient to validate surrogacy. What is required is that the effect of treatment on the surrogate endpoint reliably predicts the effect on the true endpoint. The issue of validating surrogate endpoints is complex and hundreds of publications have been written on the subject. The recent paper by Li et al. (2068) proposes a Bayesian approach to surrogacy assessment using principal stratification. They applied their proposed methodology to data from the Collaborative Initial Glaucoma Treatment Study (CIGTS). In brief, they were interested in validating the intraocular pressure (IOP) measurement at the twelfth month as a surrogate for the IOP measured at the 96^th month in patients who underwent medical versus surgical treatment. The authors also proposed a new measure to assess surrogacy, the common associative proportion (CAP), which ranges from 0 to 1, with higher values indicating better surrogacy.

Frequently, a true endpoint is costly or difficult to measure, requires a long-term follow-up or a large sample size

Based on CAP values, the IOP measured at the twelfth month was a weak surrogate to the IOP measured at month 96, with CAP of only 0.126. This was expected, as the correlation between these IOPs was only 0.304 and 0.441 in the medicine and surgery groups, respectively. This can probably be explained by IOP variations over time and by the fact that patients in the CIGTS had progressively more aggressive interventions over time depending on their IOP levels.

An interesting question is whether other measurements, such as optic disc or nerve fiber layer assessment by imaging instruments, could be used as surrogate endpoints in these clinical trials.

The statistical methodology for surrogate validation proposed by Li et al. Could find more interesting application in the evaluation of surrogate endpoints for progression in glaucoma clinical trials. The use of visual field endpoints in these trials frequently requires a very large sample size with long-term follow-up. An interesting question is whether other measurements, such as optic disc or nerve fiber layer assessment by imaging instruments, could be used as surrogate endpoints in these clinical trials.

As long as these methods are shown to be reliable surrogates for the clinically relevant endpoints, their use could potentially decrease sample size requirements, duration of follow-up and costs of research on new treatments for glaucoma.