To systematically assess studies analyzing peri-implant bone loss in implants placed in crestal and subcrestal position. Following the recommended methods for systematic reviews and meta-analyses (PRISMA), an electronic search was conducted in the PubMed (MEDLINE), EMBASE and LILACS databases to identify all relevant articles published up until April 2017. The search included human studies comparing marginal bone loss (MBL) between a control group and a study group with a minimum of 10 patients and a minimum follow-up of 6 months after prosthetic loading with rough neck implants. Two independent reviewers assessed the risk of bias in the selected studies based on the Newcastle-Ottawa scale for observational studies and the Cochrane Collaboration for clinical trials. Of 342 potentially eligible items, 7 complied with the inclusion criteria. One article was retrieved through the manual search. Eight articles were finally included: five experimental and three observational studies. The risk of bias assessed by the Cochrane Collaboration and Newcastle-Ottawa showed a high risk of bias. The mean follow-up period was 21 months (range 6-36 months). In four studies, implants placed in a crestal position presented higher MBL than subcrestal implants - the differences being significant in one study, while in three studies, implants placed in a subcrestal position presented greater MBL than crestal implants, with significant differences in only one study. Despite its limitations, the present systematic review did not find better outcomes between crestal and subcrestal implant placement, however, new studies will be needed, involving improved designs and the standardization of protocols to allow statistical comparisons and the drawing of firm conclusions.