In the November 2013 issue of the BJGP, Huang et al reported their findings of a systematic review on the use of point-of-care C-reactive protein tests to reduce antibiotic use in respiratory infections in primary care.1 This topic is highly relevant to GPs, as our specialty faces a major challenge in reducing excessive antibiotic prescribing.
To this end, it is important that scientific stringency is exercised when analysing the current evidence on an intervention that may be widely used if found to be effective. We are therefore more than concerned to discover that the present review has significant methodological flaws that impede the interpretation of the results. In their review, the authors state that all included papers were assessed by two independent researchers. Nevertheless, several errors were made.
First, the same data (regarding the same patients) from the Cals 2009 trial have been included three times. Next to the original trial comprising 431 patients, the authors have also included the antibiotic prescribing rates from two additional publications describing the cost-effectiveness and the long-term follow-up of that same trial cohort.2,3 Of note, these two papers clearly describe that they present additional outcomes of that one original trial in all sections of the papers (including the title). They all include a reference to the same published protocol paper and the BMJ trial publication. This error has a significant impact on all tables and figures, as the erroneous triple inclusion resulted in an additional 609 (24.6%) participants.
Secondly, one may question the inclusion of the Gonzales study as this was conducted at an accident and emergency setting. This is not primary care, as the authors claim in title, abstract, and text.
Thirdly, when assessing effects on antibiotic prescribing during 28-day follow-up in Figure 2a (including the index consultation as assessed in the Cals trials), Huang et al have also included data from the Diederichsen and Kavanagh studies regarding antibiotic use after the index consultation only. These different effect measures should not be included in the same meta-analysis.
Finally, we were surprised to find that the main results are presented as a single meta-analysis aggregating data from all included studies regardless of study design. This is not consistent with current methodological practice.4
We think that this review needs to be fully redone using the correct studies and corresponding interpretation of analyses. A systematic review and meta-analysis should be 100% correct, as clinicians and researchers will rely on it, cite it, and use it for implementation in clinical practice.
- © British Journal of General Practice 2014
