Neonatal meningitis with human parvovirus B19 infection.

EDITOR,-The article by Lacy and Ohlsson clearly shows how statistical manipulation of figures can produce differing results from the same basic data.' The authors with their 'cautious use of meta-analysis' find insubstantial evidence of the benefit of IVIG in either prevention or treatment ofneonatal sepsis. Using similar data, Weisman et al 2 found the relative risk of infection if IVIG prophylaxis was not given to neonates to be 2-6 (3 2) (mean (SD)), and a relative risk of death in infected neonates not treated with IVIG to be 3 0 (0-7). The authors explain this difference by suggesting use of 'inappropriate statistical methods' by Weisman et al. Lacy and Ohlsson have heavily pruned published data in search of 'good quality' and 'homogeneity'. In the field of IVIG nothing thus far has been homogenous. All the published data good or poor quality have not only differed in entry and outcome criteria but also in basic definitions of variables such as the definition of sepsis and mortality from sepsis. Nor have the authors differentiated between mortality from sepsis and that from unrelated causes or weight groups. Babies that weigh 800 g have a higher mortality from causes other than sepsis than those weighing 2500 g. The authors have also failed to discriminate between studies in which a placebo was used for the control group and studies in which there was no intervention in the control group. Another bias in this analysis was the uncritical use of the large study by Fanaroff3 which accounts for 46% of the total sample. This study was blinded only in phase I and open in phase II. As the authors themselves pointed out, the studies differed in dose regimen, duration of treatment, and the IVIG preparation used. However, they fail to point out two crucially important differences between preparations: bioavailability which depends on the method of preparation,4 and variability in IgG subclass distribution. IgG subclass distribution in the preparations is of greatest relevance, taking into account the organisms which cause infection in the neonatal period. Second, Clapp et al 5 and ourselves have clearly shown how important it is to attain and maintain serum IgG above at least 400 mg/dl to be protective very few studies report or measure serum IgG, thus making any comparison extremely difficult, if not impossible. It may make statistical sense to reject studies which are not prospective, blinded, and controlled, but it makes a nonsense to compare studies without taking into consideration the very principles on which the whole concept ofIVIG is based. Information such as bioavailability of the product, serum concentrations obtained, and well defined outcome measure are crucial principles missing from this meta-analysis. Any conclusions drawn on this basis are therefore questionable to say the least. No clinician would use a subtherapeutic dose of antibiotic, for example, and expect it to be effective. It is clear, however, that well designed, large studies with appropriate WVIG are required. K N HAQUE St Helier Hospital, Wrythe Lane, Carshalton, Surrey SM5 IAA