Abstract
Introduction
Invasive meningococcal disease is a devastating infection and early treatment is considered critical in reducing the risk of death. To this end, in some countries general practitioners (GPs) are advised to give parenteral (i.e. intramuscular or intravenous) antibiotics when they suspect meningococcal infection, without waiting for a confirmed diagnosis. However, a number of studies reporting results from small case series have failed to demonstrate strong evidence in support of this approach, while two larger observational studies, one from the United Kingdom and one from Denmark, reported an increased likelihood of death in cases given pre-hospital antibiotics. For ethical and practical reasons, a randomised controlled trial of pre-hospital antibiotics is not considered feasible.
Aim
The primary aim of the current study was to resolve this controversy by estimating the effect of pre-hospital parenteral antibiotics on case fatality risk (CFR) in meningococcal disease.
Methods
This was an observational study of New Zealand (NZ) meningococcal disease surveillance data for 1995 – 2006 (n=5340). Causal modelling indicated that minimising bias from confounding would be a major challenge of the analysis. Adjusting for multiple confounders in regression analyses resulted in a large number of cases being dropped, mainly because of missing data patterns in the model covariates. This situation led to a decision to impute missing data using chained equations. The validity of the observed estimate of effect was investigated using quantitative bias analysis methods to assess the likely impact of bias from unmeasured confounders and measurement error. These bias analysis techniques were also applied to earlier studies to explore reasons for the contradictory results reported in the literature.
Results
Among the 3796 patients in the dataset who were seen by a GP before admission the CFR was 2.9%, and pre-hospital antibiotic treatment was associated with a reduced CFR of 1.9% (adjusted and imputed RR 0.54; 95%CI 0.33 – 0.90). No likely sources of bias were identified that would alter the conclusion of a treatment benefit.
Conclusions
This is the largest reported investigation of pre-hospital antibiotics in meningococcal disease. The large study size improves precision but does not protect against bias, hence the decision to address and quantify bias in a systematic way – which did not substantively alter findings. This approach strengthens the study conclusion that pre-hospital treatment can reduce the risk of death from meningococcal disease. Following on from this investigation, the next steps will be to consider how to translate these findings to clinical and public health practice.
Investigations using surveillance data are vulnerable to bias from several sources, such as lack of data on likely confounders. These bias analysis methods could usefully be applied to other public health research questions.
A final observation is that this study was made feasible by the fact that in NZ, pre-hospital antibiotic treatment is recorded routinely in surveillance data. There is a need to consider what other information could be collected by the notification system to guide the management of meningococcal disease and related conditions of public health importance.