Evaluating a surrogate endpoint at three levels, with application to vaccine development

Abstract

Identification of an immune response to vaccination that reliably predicts protection from clinically significant infection, i.e. an immunological surrogate endpoint, is a primary goal of vaccine research. Using this problem of evaluating an immunological surrogate as an illustration, we describe a hierarchy of three criteria for a valid surrogate endpoint and statistical analysis frameworks for evaluating them. Based on a placebo-controlled vaccine efficacy trial, the first level entails assessing the correlation of an immune response with a study endpoint in the study groups, and the second level entails evaluating an immune response as a surrogate for the study endpoint that can be used for predicting vaccine efficacy for a setting similar to that of the vaccine trial. We show that baseline covariates, innovative study design, and a potential outcomes formulation can be helpful for this assessment. The third level entails validation of a surrogate endpoint via meta-analysis, where the goal is to evaluate how well the immune response can be used to predict vaccine efficacy for new settings (building bridges). A simulated vaccine trial and two example vaccine trials are presented, one supporting that certain anti-influenza antibody levels are an excellent surrogate for influenza illness and another supporting that certain anti-HIV antibody levels are not useful as a surrogate for HIV infection.

Figure 1

Frequency distributions of anti-Weiss Strain A ((a) placebo and (b) vaccine) and anti-PR8 Strain A ((c) placebo and (d) vaccine) antibody levels for the placebo and vaccine study groups of the influenza vaccine field trial.

Figure 2

Nonparametric and parametric (logistic regression model based) estimates of the incidence of hospitalization with Weiss Strain A (a) or PR8 Strain A (b) influenza infection for the placebo and vaccine study groups.

Figure 3

Nonparametric and parametric (logistic regression model based) estimates of mVE(s₁) for (a) Weiss Strain A and (b) PR8 Strain A under the anti-equipercentile assumption. As a sensitivity analysis, panels (c) and (d) show these estimates under the equipercentile assumption.

Figure 4

For data simulated to reflect the design of the PAVE-100 HIV vaccine efficacy trial, estimates of VE(s₁, 0) = mVE(s₁) under (a, b) the baseline irrelevant predictor design; (c) the close-out placebo vaccination design; and (d, e) the combined design. The solid line is the true mVE(s₁) curve, the dashed line is the empirical average estimate of mVE(s₁) over the 500 simulated data sets, and the dotted lines are 50 estimates from 50 simulated data sets. ρ= 0.5 for (a), (d) and ρ= 0.9 for (b), (e).