Diagnostic tests for bacterial infections
- Division of Neonatology and Pediatric Critical Care
- University Children’s Hospital
- D-89070 Ulm, Germany
Editor—Fowlie and Schmidt reviewed numerous publications on haematological parameters and C reactive protein for the diagnosis of bacterial infections and tried to select these publications according to well chosen criteria.1 However, they omitted to examine the publications for two criteria that substantially influence the study results.
Eight studies on band counts or immature to total neutrophil ratios were reviewed. But only the authors of two studies2 3actually defined how an immature neutrophil was differentiated from a segmented neutrophil—that is, morphological criteria, including the width of the connection between the nuclear segments. Segmented neutrophils are defined variably in published studies: most authors require an indentation of the nucleus to less than a third of the maximal nuclear diameter,3 but others require an indentation to 50% or that the connections between nuclear segments are filiform. Discrepancies in the definition of bands and segmented neutrophils may be one of the reasons that the results for sensitivity and specificity vary largely between studies.
Diagnostic parameters and especially C reactive protein have characteristic kinetics in the course of a bacterial infection: C reactive protein has a low sensitivity at the onset of clinical signs of infection but the sensitivity improves with the course of infection.4 Unfortunately, Fowlie and Schmidt included two studies in which the timing of blood sampling was not precisely defined.5 6
References
Drs Fowlie and Schmidt respond:
Editor—Franz and Pohlandt raise interesting points. We agree that differences in how the various tests were carried out may explain some of the heterogeneity in the results. In addition, only 51% of the studies included in the review described the test in sufficient detail such that it could be repeated1 so the extent of this problem was unknown. This was one of the reasons why we decided not to perform a meta-analysis of the results for any given test.
We did not set out to examine the usefulness of serial testing, but agree it is an area that may merit further investigation. However, it is important to bear in mind why diagnostic tests are performed: although the accuracy of tests may improve as the disease progresses, serial results or “late” results cannot help in deciding whether or not to start antibiotic treatment when the infant first presents. Equally, as time goes by it becomes more likely that the definitive results of the “gold standard” will become available, thereby giving the clinician the best evidence on which to base a decision whether or not to stop treatment.1
