Intended for healthcare professionals

Primary Care

More valid criteria may be needed

BMJ 2003; 327 doi: (Published 04 December 2003) Cite this as: BMJ 2003;327:1327
  1. Paul Little, professor of primary care research (ps13{at}
  1. 1University of Southampton, Aldermoor Health Centre, Southampton SO16 55T

    Zwart et al1 provide a welcome addition to the efficacy literature since a systematic review reports data from only three double blind studies among children and few from typical primary care settings.2 The study shows that—among a selected group of more unwell children (with two of the Centor criteria: absence of cough, history of fever, cervical nodes, and purulent pharynx)—antibiotics do not help symptomatically but may reduce imminent quinsy, impetigo, or scarlet fever.

    There are several reasons to support the authors' reluctance to advocate immediate antibiotics, even in this more unwell subgroup. The study was not powered a priori to detect deterioration of illness, hence secondary findings must be treated with some caution. Furthermore antibiotics did not prevent true major complications (such as rheumatic fever, quinsy) but the worsening illness (mainly “imminent quinsy”), and all cases resolved with further antibiotics and without admission. Finally efficacy trials with close follow up probably improve compliance and effect sizes compared with routine practice, where half of the medication is not taken. If the study findings can be generalised to everyday practice it means that seven children with two out of four of the Centor criteria would have to be treated to prevent one case of worsening of illness, with the associated disadvantages of side effects, reduced local and systemic immunity, and the cycle of recurrence.36

    What evidence do we need to move practice forward? The study highlights two key questions around the targeting of antibiotics.

    Firstly, are the Centor criteria valid? They were developed to predict the presence of streptococci, but with high asymptomatic carriage rates (particularly in children) this study highlights that predicting the presence of bacteria is not enough, particularly for infections of the upper respiratory tract.7 We therefore need to develop criteria based on immunological evidence of infection.7 It is also unclear what cut-off point should be used—in adults three out of the four criteria are normally used, but in this study only two were used.

    Secondly, do the criteria predict benefit from treatment? The next step is—either with the Centor criteria or with more valid criteria—to assess benefit among subgroups in a large, adequately powered trial. We therefore need to show whether antibiotics selectively predict benefit, either symptomatically or for complications, among people who meet the Centor criteria, and what cut-off in the criteria should be used.

    Until such evidence is available, what should clinicians do? It would be reasonable to share the results of this study—that seven children with the Centor criteria need to be treated to prevent worsening illness in one child—with parents. If parents were keen to have antibiotics the case would still be very strong for offering delayed antibiotics for children with two out of the four Centor criteria, since in unwell children a delayed prescribing strategy (waiting for 48 hours, rather than the normal five days) in three placebo controlled trials for patients with streptococcal pharyngitis did not result in complications.46

    Competing interests: PL has received consultancy fees for two half day meetings in the past two years on the role of antibiotics in preventing complications of lower respiratory tract infections that were funded by Abbott Pharmaceuticals.