Causality assessment of adverse events following immunization
Appears in WER 23 March 2001:
Since the inception of vaccination, it has been recognized that adverse events following immunization (AEFIs) will occur. The frequency of AEFIs is directly related to the number of vaccine doses administered. AEFIs can be causally related to the inherent properties of the vaccine, linked to errors in the administration, quality, storage and transport of the vaccine (programmatic errors), but it must be recognized that when large populations are vaccinated, some serious events that occur rarely with or without vaccination will be observed coincidentally following vaccination. Thus, investigating causality of AEFIs, particularly those that are most serious, is challenging.
The clearest and most reliable way to determine whether an adverse event is causally related to vaccination is by comparing rates of the event in a vaccinated and non-vaccinated group in a randomized clinical trial. Such trials, however, can never be large enough to assess very rare events, and postmarketing surveillance systems are required to identify events potentially related to vaccination. Postmarketing surveillance capability is improving; more countries now have AEFI monitoring systems, and more importance is attached to the reporting of suspected links between vaccination and adverse events. These systems have been successful in bringing to light serious AEFIs after vaccines have been marketed. A recent example is intussusception after administration of reassortant rhesus rotavirus vaccine.
Assessments of whether a given vaccine causes a particular adverse reaction vary from the casual observation to the carefully controlled study. The majority of individuals are not trained in interpreting such studies and are unlikely to understand the enormous difference in significance between these two extremes. Nonetheless, the public frequently forms a decision about a vaccine safety based on the information available to them, often a report based on unscientific observations or analyses that fail to stand the scrutiny of rigorous scientific investigation. Certain reports of AEFIs published in the medical literature over the past few years have resulted in controversy. The studies on which these reports are based, while generating provocative hypotheses, have generally not fulfilled the criteria that would be needed to be able to draw conclusions about vaccine safety with any degree of certainty. Yet these reports have had a major influence on public debate and opinion-making. When this debate spills over to the political arena, to policy-making and to determining the public acceptance of a vaccine by balancing the known benefits against possible but unverified risks, it is clear that a correct assessment of causality is vital.
Submitting a study to a scientific process rather than to partially informed opinion is crucial in determining whether a vaccine actually causes a given reaction. If undertaken carelessly or without scientific rigour, the study results will be inconclusive at best, may result in the inappropriate withdrawal of a valuable vaccine from use, or at worst may result in the exposure of a population to a dangerous vaccine. In 1999, WHO launched the Immunization Safety Priority Project to establish a comprehensive system to ensure the safety of all immunizations given in national immunization programmes. The development of mechanisms to respond promptly and effectively to vaccine safety concerns is a major area of focus of this project. As part of this effort, the Global Advisory Committee on Vaccine Safety (GACVS) was constituted by WHO in September 1999. The Committee's mandate is to enable WHO to respond promptly, efficiently and with scientific rigour to vaccine safety issues of potential global importance.
Building on the seminal work on determining causality of the Surgeon Generalís Advisory Committee on Smoking and Health (1964), the generally established criteria underpinning vaccine adverse event causality assessment that the GACVS uses may be summarized as follows:
- Consistency. The association of a purported adverse event with the administration of a vaccine should be consistent, i.e. the findings should be replicable in different localities, by different investigators not unduly influencing one another, and by different methods of investigation, all leading to the same conclusion(s).
- Strength of the association. The association should be strong in the magnitude of the association (in an epidemiological sense), and in the dose-response relationship of the vaccine with the adverse effect.
- Specificity. The association should be distinctive, the adverse event should be linked uniquely or specifically with the vaccine concerned, rather than its occurring frequently, spontaneously or commonly in association with other external stimuli or conditions.
- Temporal relation. There should be a clear temporal relationship between the vaccine and the adverse event, in that receipt of the vaccine should precede the earliest manifestation of the event or a clear exacerbation of an ongoing condition. For example, an anaphylactic reaction seconds or minutes following immunization would be strongly suggestive of causality; such a reaction several weeks after vaccination would be less plausible evidence of a causal relation.
- Biological plausibility. The association should be coherent; that is, plausible and explicable biologically according to known facts in the natural history and biology of the disease.
Clearly, not all these criteria need to be present, and neither does each carry equal weight for a causal relationship between an adverse event and the vaccine to be determined. In addition to the general principles mentioned above, there are a number of provisos or considerations that need to be applied for determining causality in the special field of vaccine safety. They are:
- The requirement for biological plausibility should not unduly influence negatively a consideration of causality. Biological plausibility is a less robust criterion than the others described. If an adverse event does not fit into known facts and the preconceived understanding of the adverse event or the vaccine under consideration, it clearly does not necessarily follow that new or hitherto unexpected events are improbable. ìBiological plausibilityî is most helpful when it is positive; it is less so when negative.
- Consideration of whether the vaccine is serving as a trigger (trigger in this context is an agent that causes an event to happen which would have happened some time later anyway). When acting as a trigger, the vaccine may expose an underlying or pre-existing condition or illness. An example of the latter would be an auto-immune condition triggered non-specifically by the immune stimulus of the vaccine.
- In the case of live attenuated vaccines, if the adverse event may be attributable to the pathogenicity of the attenuated vaccine microorganism and thus not be distinguishable (except, perhaps, in severity) from the disease against which the vaccine is being administered, a causal connection is more plausible. Identification of the vaccine organism in diseased tissue and/or in the body fluids of the patient in such a situation would add weight to causality. There are exceptions to both these above points.
An association between vaccine administration and an adverse event is most likely to be considered strong when the evidence is based on:
- Well-conducted human studies that demonstrate a clear association in a study design that is determined a priori for testing the hypothesis of such association. Such studies will normally be one of the following, in descending order of probability of achieving the objective of the study: randomized controlled clinical trials, cohort studies, and case-controlled studies and controlled case-series analyses. Case reports, however numerous and complete, do not fulfil the requirements for testing hypotheses, although on occasion such reports can be compelling if there are clear biological markers of the association, as is the case for vaccine-associated paralytic poliomyelitis.
- An association that is demonstrated in more than one human study and consistent among the studies. The studies would need to have been well conducted, by different investigators, in different populations, with results that are consistent, despite different study designs. Demonstrable association in the studies between dose and the purported adverse effect (either the dose or the number of doses administered, or both) will, in many cases, strengthen the causal association between the vaccine and the adverse event. This is not always the case, especially if there is an immunological relationship.
- A strong similarity of the adverse event to the infection the vaccine is intended to prevent, and there is a non-random temporal relationship between administration and the adverse incident.
It is important that there should be a strict definition of the adverse event in clinical, pathological and biochemical terms, as far as that is achievable. The frequency in the nonimmunized population of the adverse event should be substantially different from that in the immunized population in which the event is described, and there would not normally be obvious alternative reasons for its occurrence that are unrelated to immunization.
An adverse event may be caused by a vaccine adjuvant or excipient, rather than by the active component of the vaccine. In this case, it might spuriously influence the specificity of the association between vaccine and adverse event. As far as possible, safety issues should be clarified in premarketing controlled clinical studies, with attention being given in such studies to safety issues and their monitoring, although with extremely rare unexpected events, this may not be achievable because of the need for extremely large sample sizes to detect them.
When adverse events are attributable to a vaccine, it is important to determine whether there is a predisposed set of subjects (by age, population, genetic, immunological, environmental, ethnic, sociological or underlying disease conditions) for any particular reaction. Such predisposition is most likely to be identified in case-controlled studies.
A systematic effort should always be made to exclude confounding programmatic errors and variability and aberrations in vaccine manufacture. The latter quality issues are most likely to be revealed by careful attention to batch and lot testing.
Since observational studies are not randomized and since individuals who are ill are generally less likely to be immunized (but more likely to have an adverse outcome), epidemiological studies on vaccine safety need to pay special attention to contraindications as potentially confounding factors. The consequences of this bias may be false-negative studies.