Background: For the empirical treatment of communityacquired pneumonia requiring admission to a non-ICU ward, the Dutch guidelines recommend either beta-lactam monotherapy, beta-lactam and macrolide combination therapy, or fluoroquinolone monotherapy. The lack of convincing evidence to preferentially recommend any of the three empiric regimens results from intrinsic limitations of current studies, such as bias by indication and residual confounding in observational studies, and the unknown effects of pre-randomisation antibiotic use in randomised controlled trials. In this paper we discuss the methodological drawbacks of observational cohorts and randomised controlled
trials in antibiotic therapy. Next, we explain why we designed a multicentre cluster-randomised cross-over study to evaluate the effectiveness of three antibiotic treatment strategies, consisting of a preferred treatment regimen of beta-lactam monotherapy, beta-lactam and macrolide combination therapy or fluoroquinolone monotherapy, in adult patients admitted to a non-ICU ward with a clinical diagnosis of communityacquired
pneumonia. Furthermore we outline different
aspects of this design that deserve thorough consideration. Conclusion: We discuss different aspects of a clusterrandomised cross-over trial that is designed to determine the effects of three recommended regimens of antibiotic treatment of CAP.