Intended for healthcare professionals


Vaccination and its adverse effects: real or perceived

BMJ 1998; 317 doi: (Published 18 July 1998) Cite this as: BMJ 1998;317:159

Society should think about means of linking exposure to potential long term effect 

  1. Tom Jefferson (zorria{at}, Coordinator
  1. Cochrane Vaccines Field, Ministry of Defence, Ash Vale, Hants GU12 5RR

    Vaccines have been spuriously linked to sudden infant death syndrome,1 paediatric asthma,2 autism,3 inflammatory bowel disease,4 and permanent brain damage.5 Recently US researchers have suggested that vaccination after 28 days after birth may induce type 1 (autoimmune) diabetes mellitus in susceptible individuals.5 This theory, pounced on earlier this year by the US media, may have led to a lowering of confidence in childhood routine immunisation. In May several institutions (including the National Institute of Allergy and Infectious Diseases, Centres for Disease Control, the World Health Organisation, and the UK's Department of Health) sponsored a workshop at the US National Institutes for Health to assess the evidence of a possible causal link.

    Immunologists, diabetologists, epidemiologists, policymakers, and observers debated the available evidence for two days and concluded that it does not support a causal link between vaccination and the onset of type 1 diabetes. Some short and longer term observational studies to test the hypothesis are currently underway. However, the results of a large randomised controlled trial of vaccine against Haemophilus influenzae type b carried out in Finland in 1985-76 were reanalysed by Tuomilehto et al and showed no association between the incidence of diabetes mellitus and the addition of another antigen to the schedule, irrespective of timing (unpublished data). Data reanalysis was made possible by prospective linking of individual information on exposure (in this case infant vaccination or placebo administration) with the Finnish diabetes register.

    Neil Halsey, head of the Institute for Vaccine Safety at Johns Hopkins University, summed up features common to recent vaccine scares:

    • A casual link is usually claimed with a disease or condition of unknown or unclear aetiology.

    • The association is claimed by one investigator or a group of investigators.

    • The association is not confirmed by peers or by subsequent research.

    • The claims are made with no apparent concern for potential harm from public loss of confidence and refusal to vaccinate children.

    Additionally, findings of subsequent studies that fail to confirm the original claim never get the publicity given to the “original” finding; thus the public never gets a balanced view.

    It is time to think hard about how society can deal with the difficult issue of possible long term and rare adverse effects of vaccination. Attention to the issue is unlikely to fade, as new and better vaccines are produced and as public expectations of effective and safe interventions increase. The first obvious source of data on rare and long term effects is the original clinical trials of the vaccine, with direct observation of the incidence of events in one of the double blind randomised arms. But, early trials, usually conducted for registration, are too small and too short.7 Additionally, assessment of adverse effects is probably best done by comparing events in one or more intervention arms with those in a placebo arm, thus restricting observation to trials of new or partially tested vaccines for which a placebo arm is ethically admissible.

    One possible solution could be to increase the duration and power of trials to detect rare and long term adverse effects. Apart from cost, however, there are major ethical problems in continuing a trial with a non-immunised cohort in an effort to detect possible rare and long term effects once the short term safety and effectiveness of the vaccine have been shown.

    The use of case-control studies and case series is helpful in defining the likelihood of an association but, given the possible presence of multiple unknown biases, such studies do not allow estimation of the attributable risk, essential for safety assessment. An additional problem with any prospective approach is that some adverse effects become known only years after the development, marketing, and registration of the vaccine, making “data dredging” the only way in which they might be observed and later recognised. Data dredging is likely to be inefficient and unable to assess unexpected associations, which are likely to take place periodically.

    One way out of the dilemma could be the linking of individual exposure to vaccination to possible adverse events in later life in a similar fashion to the reanalysis by Tuomilehto et al of the original Haemophilus influenzae type b trial data. This would allow the creation of retrospective exposure cohorts linked to historical controls for testing any of the hypotheses generated. This approach would require access to individual immunisation data and the ability to identify and locate individuals in later life. As most immunisation schedules vary from country to country the international extension of such a scheme would allow comparisons based on exposure to different schedules, thus enhancing the power of the system. An enhancement of this method could be its use within the context of a systematic review of the known effects of vaccines in question made available within the Cochrane Library.

    Whatever methods are used, governments and manufacturers will be increasingly involved in assessing long term adverse effects of vaccines and will need to reassure the public of the overwhelming safety record of vaccines, when the seriousness of the target disease is forgotten. This at present is the only certainty.


    I thank Drs Neal Halsey, David Salisbury, Dirk Teuwen, and Jaakko Tuomilehto for their help.


    View Abstract