BackgroundMany aspects of pharmacological treatment of Lyme neuroborreliosis in children, such as choice of drug, dosage, and duration are subject to intense debates, leading to uncertainties in patients’ parents and healthcare providers alike. To assess the available evidence for pharmacological treatment for children with Lyme neuroborreliosis we conducted a systematic review.MethodsThe comprehensive systematic literature search included randomized-controlled trials (RCTs) and non-randomized studies (NRS) on treatment of Lyme neuroborreliosis in children (age <18 years). Our primary outcome was neurological symptoms after treatment. Risk of bias was assessed with the Cochrane risk of bias tools for RCTs and NRS. Quality of evidence was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach.ResultsTwo RCTs and four NRS were eligible for inclusion. Risk of bias in RCTs and NRS was generally high. Reporting of studies was generally poor. Regarding the primary outcome neurological symptoms at 1–3 months, no statistically significant difference could be found in cohort studies between doxycycline and beta-lactam antibiotics. In two RCTs comparing penicillin G and ceftriaxone, no patient experienced residual neurological symptoms at the last reported time points. Quality of evidence according to GRADE was judged very low.ConclusionsData is scarce and with limited quality. Several issues could not be addressed due to scarcity of information. No eligible study compared different treatment durations. According to the available evidence, there seems to be no difference between different antibiotic agents for the treatment of Lyme neuroborreliosis in children regarding neurological symptoms. We found no evidence that supports extended antibiotic regimes.Review registrationSystematic review registration: CRD42014008839.Electronic supplementary materialThe online version of this article (doi:10.1186/s12883-016-0708-y) contains supplementary material, which is available to authorized users.