Surveillance data for assessing impact of vaccination are valid

EDITOR—Olowokure et al suggest that routine surveillance of Haemophilus influenzae is incomplete, that completeness declined after the vaccine was introduced, and that the effectiveness of vaccination programmes is overestimated.1 We question this.

The authors suggest that the same weakness may affect the surveillance of the group C meningococcal vaccination programme. We accept that routine reporting is often incomplete, but, because of this, all vaccine preventable infections are under enhanced surveillance. The 15-fold reduction in the incidence of H influenzae type b disease after the introduction of the H influenzae type b vaccine was observed in an enhanced active surveillance scheme operating in five NHS regions,2 not by the use of routine laboratory reports to the Public Health Laboratory Service (PHLS) Communicable Disease Surveillance Centre.

This surveillance scheme was established before the vaccine was introduced and was continued until 1995, when an enhanced national scheme was implemented.3 Completeness of data can be maximised by reconciling reports to the Communicable Disease Surveillance Centre, isolates referred to the PHLS Haemophilus Reference Unit, and notifications of H influenzae type b meningitis, and by active reporting of cases to the British Paediatric Surveillance Unit.

A similar scheme—reconciling reference laboratory reports, notified infections, cases known to consultants in communicable disease control, and laboratory reports to the Communicable Disease Surveillance Centre—was used to determine the burden of infection before the introduction of meningococcal group C vaccine4; this scheme has now been extended nationally.

Olowokure et al do not mention the most important weakness of surveillance systems for vaccine preventable disease. When the incidence of a disease is being ascertained after the introduction of a vaccine, the specificity of clinical case definitions and laboratory tests is critical. If specificity is low, when the true incidence of an infection declines the predictive value of the case definition falls and the proportion of false positive diagnoses increases.

Since 1990 in five regions, and since 1995 in the whole of England, all reports of confirmed invasive haemophilus infections have been followed up by referral of the isolate to the PHLS Haemophilus Reference Unit, where additional confirmation is carried out with molecular typing techniques. Between 1995 and 1999, of 136 putative type b isolates referred, only 108 were confirmed as type b.

The evaluation of surveillance data for H influenzae type b vaccine should be based only on confirmed type b infections. The collaboration between the Communicable Disease Surveillance Centre and the national reference laboratories for meningococcal and haemophilus infections ensures that national surveillance data for England and Wales for assessing the impact of vaccination are valid.

Authors' reply

EDITOR—Ramsay et al misrepresent the conclusions of our study. We suggested that routine surveillance was incomplete, that the underascertainment worsened after the vaccine was introduced, and that if routine surveillance data are used the effectiveness of the vaccine is overestimated. Thus we make the argument for the introduction of enhanced surveillance before the introduction of the intervention (to set baselines) and its continuation, with the same methods, after the introduction (to detect change).

At the time that we performed our analysis the relevance to the introduction of meningococcal type C vaccine was that no funding had been agreed for ongoing enhanced meningococcal surveillance, even though the immunisation programme had started. Enhanced national surveillance has now been set up, although there are methodological differences to the subnational system that was in place before the vaccine was introduced.

Because our paper was published as a short report (maximum 600 words) we did not have space to mention several weaknesses in surveillance systems. The use solely of laboratory confirmed cases obtained by testing routinely generated clinical specimens presupposes that no changes occur in the clinical practice that generates these isolates. Experience with meningococcal infection suggests that preadmission antibiotics and an increased reluctance to perform lumbar puncture in recent years have reduced the likelihood of obtaining an isolate of the infecting organism1 and that other methods of case ascertainment are required.2

Ramsay et al, and others, expend enormous effort in improving surveillance data. They should interpret our report as supportive of the need for their efforts.