Substitutes/surrogates are proxy or indirect measures for clinical endpoints, which are not always useful. Clinical outcome (i.e., increased survival) is of research and regulatory interest, not changes in laboratory measurements.

In survival analysis, no surrogate will explain all the effects of the treatment being studied. Even if it did, there is always measurement error, and there wouldn’t be a perfect correlation.

In observational studies, it may be difficult to separate surrogate exposures from the risk factor of interest. This would be the case when surrogates are in the causal chain between the risk factor of interest and the disease. If the surrogate is treated as a confounder, and adjusted for, the association between risk factor and disease may disappear. This could be seen as over-adjustment.