Objective: A fundamental information needed to conduct economic evaluations of vaccines is effectiveness against disease. However, effectiveness is not always observed directly and relies on an immunological response that predicts protection. Typical immune responses which are predictive of protection are neutralizing antibodies, called surrogates or correlates of protection (COP). Often the COP is reduced to a threshold value that differentiates between protected and susceptible. COPs are relied on in place of estimates of effectiveness and for immunization policy, however there are no consistent criteria or statistical methods for establishing candidate immune response as predictive COP. Our aims were to review proposed hierarchies of evidence necessary to establish a COP and statistical methods used to relate immune responses to protection. Methods: The strength of evidence for demonstrating a COP based on different frameworks and early and modern statistical methods approaches to establish a COP were reviewed. Findings and Recommendations: Different frameworks define different levels of confidence in COPs. The Prentice framework is significance testing-driven and requires protection to be related to vaccination, the correlate related to the vaccine and correlate related to clinical endpoint. Moreover vaccination should not add additional information on protection over that explained by the correlate. A framework by Qin proposes levels of evidence based on single or multiple randomized trials. To estimate thresholds, early vaccine studies relied on inspection of disease rates observed in discrete intervals of assay values. Modern examples employed Chang-Kohberger method, but this requires an estimate of vaccine efficacy based on occurrence of disease before it can be used. The scaled-logit model permits estimation of continuous protection curves by antibody titer. In addition to statistical criteria, other considerations include clear endpoint definition, laboratory assays, host and population factors. New statistical methods should be developed and tested within evidence frameworks to better obtain estimates of vaccine effectiveness.