We considered Aabenhus and colleagues’ comments about the shortcomings of our article. However we disagree with the main points in their letter. First, initially, we noticed that Cals et al may use the same dataset to publish duplicate studies. However, their results are different. In order not to miss any studies, we conducted the analyses both including them and excluding them. We found whether to include or exclude them would not affect the overall results. In this systematic review, we did not want to miss any studies that met the inclusion criteria.
Secondly, C-reactive protein (CRP) is a kind of biomarker for improving the assessment of infection. CRP is an inherent and natural inflammatory protein in patients irrespective of where a patient is, such as in primary care or emergency. The relationship of CRP to infection does not change. With the same condition, first contact may be flexible in primary care or accident and emergency; thus, we also included Gonzales’ study. In addition, the sample size in Gonzales’ study is very small and the exclusion of this study did not affect the results.
Thirdly, Aabenhus and colleagues are wrong. There is no Figure 2a in this study. We assume they meant Figure 3a, which studied antibiotic prescribing at any time during the 28-day follow-up. All the studies included in Figure 3a have same effect measure.
Finally, there were many publications of meta-analysis which combined the results from both randomised controlled trials and observational studies, such as cohort studies. In our study, we had already conducted subgroup analyses according to the study design.
Editor’s note
Readers are invited to comment on this correspondence, in which some methodological questions about the selection of studies and patients for the meta-analysis are discussed.
- © British Journal of General Practice 2014
REFERENCE
- 1.