BackgroundDiabetes is the leading cause of end-stage kidney disease (ESKD). Because of conflicting results in observational studies, it is still subject to debate whether in diabetic patients the dialysis modality selected as first treatment (haemodialysis or peritoneal dialysis) may have a major impact on outcomes. We therefore aimed at performing a systematic review of the available evidence. MethodsMEDLINE, EMBASE and CENTRAL databases were searched until February 2014 for English-language articles without time or methodology restrictions by highly sensitive search strategies focused on diabetes, end-stage kidney disease and dialysis modality. Selection of relevant studies, data extraction and analysis were performed by two independent reviewers. ResultsTwenty-five observational studies (23 on incident and 2 on prevalent cohorts) were included in this review. Mortality was the only main outcome addressed in large cohorts. When considering patient survival, results were inconsistent and varied across study designs, follow-up period and subgroups. We therefore found no evidence-based arguments in favour or against a particular dialysis modality as first choice treatment in patients with diabetes and ESKD. However, peritoneal dialysis (PD) as first choice seems to convey a higher risk of death in elderly and frail patients. ConclusionsThe available evidence derived from observational studies is inconsistent. Therefore evidence-based arguments indicating that HD or PD as first treatment may improve patient-centred outcomes in diabetics with ESKD are lacking. In the absence of such evidence, modality selection should be governed by patient preference, after unbiased patient information.