Immunological Correlates of Vaccine-Mediated Protection

2.1

Correlates of Protection

A variety of different terms, definitions, and interpretations of them exist in the literature surrounding correlates of protection. The lack of a consistent approach to defining correlates leads to confusion but also reflects the increasing complexity of knowledge of the host response that determines protection. Here we use a simple practical definition: a correlate of protection is a biomarker that is statistically associated with protection. Thus, in theory, any measurement that can be made of the host response pathways from the first interaction of a pathogen with the host (perhaps binding to a pattern recognition receptor on dendritic cells) through to the effector of the human immune system (eg, a bactericidal antibody response), could be related statistically to protection and be a correlate ( Fig. 7.1 ). It is also important to consider that there may also be bystander effects of the activation of the immune response, which are not responsible for protection either directly or indirectly, but could be statistically related to protection and thus a correlate. While this latter category has not been much considered in the past, the current use of transcriptomic and proteomic approaches to analyze immune responses is likely to result in the identification of such activated pathways that are not driving the protective responses but are statistically related to them. While it may appear that any of the previously mentioned measures that correlate consistently with protection would be a useful marker, there are advantages in identifying the effector(s) of protection, as these are by definition the most robust measurements to provide assurance of effectiveness ( Fig. 7.1 and Table 7.1 ).

2.2

Relative and Absolute Correlates

An additional important consideration is that correlates may be relative or absolute. An absolute correlate is one in which there is a defined and accepted threshold above which there is protection and below which there is not. Unfortunately, some “absolute” correlates of protection are not absolute, as will be discussed later. A relative correlate is where the protective biomarker has been defined but there is no threshold that relates to absolute protection, although generally higher numbers (eg, higher level of RSV antibody) are related to more protection. For relative correlates, there may be some individuals who are protected even with rather low levels of antibody, resulting in a frequency distribution of protection.

2.3

Co-correlates

In some circumstances there is more than one mechanism of protection that can be measured. Indeed, it seems likely, especially with live vaccines, that there are multiple pathways to protection that could be induced. For example, live viral vaccines may induce both neutralizing antibodies and cytotoxic T cells, which could both be important in early defense against infection. It is plausible that high avidity antibodies produced in response to meningococcal conjugate vaccines could provide some protection through bactericidal activity, opsonophagocytosis, and antibody-dependent cellular cytotoxicity, yet only bactericidal activity is firmly accepted as the major contributor to protection.

2.4

Surrogates of Protection

Surrogates of protection, from the dictionary definition of the word “surrogate,” are measurements that substitute or are a proxy for protection, which would make this essentially the same as the definition of a correlate of protection described earlier. However, others have considered that the word “surrogate” relates to the “correlate” (rather than the protection) and therefore a surrogate may be defined as a marker that substitutes for the correlate of protection, but does not itself confer protection. The terms correlate and surrogate have often been used interchangeably in the primary literature leading to much confusion. Qin et al. defined correlates of protection as either correlates of risk (for correlates not in the mechanistic pathway to protection), or as different levels of surrogates of protection that relate laboratory measurements to vaccine efficacy ( Table 7.2 ). For Qin et al. a surrogate of protection is necessarily on the mechanistic pathway to protection, with a Level 1 surrogate of protection representing a correlate on the mechanistic pathway to protection (our terminology) from a single setting (eg, a single vaccine trial), and a Level 2 surrogate of protection representing a correlate on the mechanistic pathway to protection that is predictive of vaccine efficacy across a number of settings. Although useful in highlighting the importance of validating correlates across different settings, their terminology have been recently simplified, as discussed later.

The major regulatory authorities have included the term “surrogate” in their own definitions, but are not consistent in their usage ( Table 7.2 ). The World Health Organization defines a surrogate as “a marker that is statistically associated with clinical protection and that lies on the causal pathway leading to protection” whereas a correlate is “a marker that is statistically associated with clinical protection, but not necessarily on the causal pathway leading to protection,” which is almost exactly the opposite of the definition given earlier and of the US Food and Drug Administration (FDA) definition ( Table 7.2 ). While some have argued that the term surrogate should be abandoned as a result of this discrepancy in definition, it is established in official definitions and it is therefore important to recognize that there are substantial differences in the definitions used both in the primary literature, and reviews and official documents.

2.5

Mechanistic and Nonmechanistic Correlates

Plotkin and Gilbert have recently proposed that correlates are simply divided into mechanistic and nonmechanistic correlates. The former being those assessments that directly measure the effector mechanism of protection and the latter being measurements of other responses that correlate with protection, but are not responsible for it. An example of a mechanistic correlate, according to Plotkin and Gilbert, is meningococcal bactericidal antibody (measured using the serum bactericidal assay), which is believed to be the effector of vaccine-induced protection after immunization with meningococcal vaccines and correlates with protection. Meningococcal antibodies can also be measured by enzyme-linked immunosorbent assay (ELISA) and these in turn also correlate with protection, but since antibodies, measured in this way, raised by vaccination do not necessarily actually confer protection, ELISA antibodies are considered in this framework to be a nonmechanistic correlate of protection. Conversely, in the context of the herpes zoster (shingles) vaccine, these authors have argued that T cells are responsible for protection providing a mechanistic correlate, but the most convenient measurement to make clinically, which also correlates with protection, is the antibody response, which is not thought to be responsible for protection and is therefore a nonmechanistic correlate. Both of these examples highlight difficulties in understanding the correlate. The mechanistic correlate, meningococcal bactericidal antibody, is a subset of the antibodies contained in the total antibody measured by ELISA, leading to the conclusion that both of these measurements contain mechanistic correlates. In the case of the shingles vaccine, it is possible that there is some minor contribution of antibody-mediated viral neutralization (preventing spread between cells) to the containment of the varicella zoster virus (VZV) after immunization and thus the nonmechanistic correlate may also be a partial mechanistic correlate.

2.6

Pathway and Effector Correlates of Protection

In view of the difficulties mentioned earlier in defining correlates and surrogates of protection, and the different understanding about the terms among individuals and official agencies, it is important that authors routinely define the terms when using them in order to maintain clarity. We suggest that a more precise and descriptive definition of correlates, as proposed by Plotkin and Gilbert, might improve understanding of the differences in meaning that have appeared in the literature. However, to be prepared for the rapid expansion in measurements that will be produced following the widespread adoption of the new technologies, we propose the following descriptive definitions ( Fig. 7.1 )

• 1.

  Established mechanistic correlates of protection:

  • a.

    Effector correlate of protection: measurement that is the effector mechanism of protection and does correlate with protection (eg, meningococcal bactericidal antibody)

  • b.

    Pathway correlate of protection: measurement of a biomarker that is on the pathway of responses that leads to the protective response and does correlate with protection (eg, T follicular helper cells)

• 2.

  Undefined correlate of protection: measurement of a biomarker that is not established as directing the protective response but does correlate with protection (eg, a gene expression profile that correlates with protection)

There are many different types of response that may be an effector correlate but an almost unfathomable number of correlates that may be pathway or as yet undefined correlates (examples are show in Table 7.1 ).

2.7

Protection Endpoints

Before discussing correlates of protection further, it is important to consider the endpoint: protection. It is readily assumed that a correlate of protection relates to the defined endpoint of sterilizing immunity at the individual level, that is, if the biomarker is present at a certain level then the individual is fully protected from infection. While this might be an ideal situation, the reality is that not all vaccines deliver sterilizing immunity and, furthermore, obtaining this endpoint might be clinically unimportant or logistically impossible. For example, the trials of rotavirus vaccines (which enrolled more than 60,000 infants) focused on prevention of hospitalization and severe disease as achievable endpoints, and had a correlate been established, this would have been the protective endpoint defined. Mild rotavirus infection is not clinically important and did still occur in the clinical trials, despite high efficacy against severe disease. By contrast, the endpoint for the original trial of the pneumococcal conjugate vaccine (PCV7), involving over 37,000 infants, was invasive pneumococcal disease, measured by blood culture. This study led to the definition of a level of antibody (0.35 μg/mL) as the correlate of protection that has been widely adopted. From a global perspective, invasive pneumococcal disease is not actually the most important clinical endpoint, since most of the hospitalizations and deaths from pneumococcal disease are caused by pneumonia. Pneumococcal pneumonia in children, where approximately only 10% of cases are blood culture positive, is difficult to define with high specificity but the development of a consensus on a WHO radiological definition of endpoint pneumonia has made the use of pneumonia accessible as a useful clinical endpoint.

It is now established that herd immunity (herd protection) is an important component of the protection of populations against almost all communicable diseases that are acquired through contact with other humans. In the case of pneumococcal disease, herd immunity is most readily measured by studying the reduction in nasopharyngeal carriage of vaccine-type pneumococci among vaccinated populations, and is now the basis of effectiveness studies being undertaken in many settings to assess the impact of the roll out of new PCV10/13 vaccine programs. Once colonization of toddlers with vaccine serotypes is blocked by vaccine-induced antibody, disease rates fall among children and adults who are vaccinated and those who are not as transmission is interrupted.

Suitable clinical endpoints, with which a biomarker might be usefully correlated, therefore include: (1) absolute prevention of infection, (2) prevention of death, (3) prevention of severe disease/hospitalization, (4) prevention of sequelae, (5) prevention of certain syndromes associated with the infection (eg, pneumonia), and (6) a clinical measure of herd immunity (eg, blocking of colonization).

While the gold standard for establishing a correlate is to consider individual protection in randomized controlled clinical trials, some biomarkers are best established after licensure, particularly where there has been no prelicensure efficacy trial. In this circumstance, the correlate of protection will relate only to population protection and cannot quantitate the level of the correlate that is required to confer individual protection. A good example of this is the capsular Group C meningococcal vaccine that was evaluated after licensure in the United Kingdom in an effectiveness study. The proportions of the population in different age groups whose serum bactericidal antibody titer was over the putative protective threshold ( ≥ 1:8 with rabbit complement) was related to the population estimate of protection to derive a threshold for protection at the population level.