The use of polysaccharide pneumococcal vaccination in elderly or high risk populations remains controversial. Evidence from less developed countries is the most persuasive in the absence of HIV, but in more developed countries the nine randomised controlled trials to date are inconclusive. They have now been the subject of some five meta-analyses^1–5 and to this can be added a recent meta-analysis of both the randomised controlled trials and observational studies.⁶ There has even been a review of the meta-analyses!⁷

The most recent review is helpful because it compares the trials and the observational studies using similar quality criteria and methods of pooling results. The strongest evidence is based on pneumococcal bacteraemia as the end point. Here studies of all types—case-control, cohort and randomised controlled trials—show consistent evidence of around 50% protection. However, it remains unclear whether this represents true protection or a suppression of bacteraemia without affecting the underlying disease (pneumonia). When it comes to pneumonia, the picture is much more confusing—and here we mean all cause pneumonia. Both observational studies and randomised controlled studies show significant heterogeneity between studies. If an estimate of protection is made despite this heterogeneity, then the trials show a tiny harmful effect compared with a beneficial 35% in observational studies (although the confidence interval included no protection).

In this situation it is very good to see some new primary data—especially from a trial. In this issue of Thorax a study by Alfageme et al⁸ examines whether the vaccine would prevent community acquired pneumonia in individuals with chronic obstructive airways disease. The study was relatively small and so, although the overall efficacy was 24%, this did not reach statistical significance. However, they did find statistically significant evidence of protection in two subgroup analyses—namely, in individuals aged less than 65 years and in those with severe disease. It was not clear if these subgroups were specified a priori. The authors conclude by recommending vaccination in these groups. Given that these are a relatively small population at high risk, this would seem reasonable. Assuming the vaccine does no harm—on which point the evidence is reasonably strong—then the cost will be modest for the potential benefit.

In contrast, the UK has adopted a policy of vaccinating all individuals aged over 65 years—at considerable cost. This programme is being evaluated, although currently only by pneumococcal bacteraemia. As the introduction was phased in—beginning with those aged over 80 years in August 2003, followed by those over 75 years from April 2004, and finally those aged 65 and older from April 2005—the results are currently only robust for those older than 80. These were presented at the HPA conference in Warwick this September. They show a 9% decline in the rate of bacteraemia with no effect on case fatality. A particular problem with using surveillance data for estimating effects in this programme are that the coverage rate was only estimated at 26% in 2003/4. There is therefore a large potential for confounding by selection of those in contact with health services and with differing levels of health from the whole population. It is to be hoped that pneumonia can be included as an end point in this surveillance but, even then, confounding will remain an issue unless risk factors are collected on both those affected and on the vaccinated population.

One concern with all of these studies has been the misclassification inherent in using all cause pneumonia as an outcome—an effect on pneumococcal pneumonia could be diluted. Methods of diagnosing pneumococcal disease have improved markedly in recent years. The time has surely come to apply these to this issue, with adequate numbers and in a carefully designed study that can cope with confounding.

This issue has become more complex with the advent of pneumococcal conjugate vaccines. The USA introduced a 7-valent vaccine into the routine childhood programme in 2000. Although this vaccine was designed for the common pneumococcal serotypes affecting children, an effect has been seen on adult disease.⁹ Rates of invasive pneumococcal disease have declined in all age groups from 50 years upwards with the greatest decline (a remarkable 35%) seen in those aged over 84 years. This is likely to be causal since an effect on adult pneumonia was seen in the original US trial of childhood pneumococcal conjugate vaccine. It is presumed that this represents an effect on transmission from young children since it is known that carriage of the pneumococcus is reduced by the conjugate vaccine. Any future evaluation will therefore need to take account of any concurrent pneumococcal conjugate vaccination in the population.

It has also been proposed that the conjugate vaccines—possibly with more serotypes added—might be used directly in the elderly. The rationale here would be to induce better immune memory with a priming dose of conjugate followed by a booster dose of the 23-valent polysaccharide. Clearly trials of this approach are warranted but, as Conaty and colleagues point out,⁶ we must not allow the same situation to arise in the high risk and elderly as has happened with polysaccharide vaccine. We need carefully designed clinical trials of adequate size using the best modern methods of diagnosing pneumococcal pneumonia to determine the end point.