Background: Infectious complications that belong to the most serious complications of cancer treatment can be prevented using antibiotics or colony stimulating factors (CSFs). Determining the role of the each of these strategies and their combination was the aim of a network analysis. Since most trials aimed to evaluate the efficacy of either CSFs or antibiotics, not all trial protocols specify the prophylaxis to be used in both arms of the trial. Excluding such trials may lead to bias. Methods: A Bayesian network analysis of RCTs was performed. The following four treatment options were compared: antibiotics, CSF, the antibiotics plus CSFs and no prophylaxis. Trials which could not be allocated into these categories were excluded from the network. However they were included in sensitivity analyses: pair-wise meta-analyses of CSF versus no CSF (or antibiotics versus no antibiotics) were performed including subgroup analyses by use of antibiotics (CSFs) in both arms of the trial. Results: The search retrieved 125 trials. The prophylaxis in both trial arms was unclear in 50 trials. For CSFs, odds ratios and 95% credibility intervals for infections in the full network were: CSF versus none 0.63 (0.44 to 0.84) and CSFs versus CSF plus antibiotics 0.61 (0.46 to 0.76). There was no evidence for heterogeneity or inconsistency in the network. In the sensitivity analyses of efficacy for CSFs, no subgroup differences were found between trials mandating antibiotic prophylaxis, without antibiotic prophylaxis and trials with unclear prophylaxis. Results were similar for antibiotics. Conclusions: When treatments can be combined, a systematic review examining different treatment options is hampered by reporting/design of treatments in both trial arms. Excluding trials from network analysis that cannot be allocated to treatment options is problematic. Our solution was to perform both a network analysis of different treatment options and meta-analyses of the efficacy of antibiotics and CSFs separately.