Caroliene Meijndert

121 The effect of implant-abutment connections on peri-implant bone level change Data extraction Following a pre-specified form, the following data were extracted cooperatively by C.M.M. and H.J.A.M.: authors, year of publication, study design, follow-up time, type of implant, type of implant-abutment connection, number of implants, number of implant failures, bone level changes and, if available, additional data on the mid-buccal mucosa level changes. It was decided to group the studies according to the properties of the internal configuration. When an implant connection was fully or partially conical/ tapered along the inner wall of the implant body and the corresponding portion of the abutment, the implant was classified as conical and placed in the ‘platform switched conical’ group (PS-conical). Any connection where the inner portion of the implant and abutment was parallel-walled were classified based on the presence or absence of a platform switch and were placed accordingly into either the ‘platform switch parallel’ (PS-parallel) or ‘platform matched parallel’ group (PM- parallel). Statistical analysis Inter-observer agreement was calculated with IBM SPSS Statistics 20 (SPSS, Chicago, Illinois, USA). Publication bias was assessed by plotting the log odds ratio against its standard error. Pooling of bone loss, implant survival and mid buccal mucosa changes was performed by using the software Comprehensive Meta-Analysis, Version 3.3.070 (CMA, Biostat, Englewood, NJ 07631, USA). A random effects model was used on the pooled outcomes. To analyse sources of heterogeneity between studies, a meta- regression analysis (random effects model) was performed on the connection types, i.e., PS-conical, PS-parallel and PM-parallel. Results Study identification and selection The study selection procedure is summarised in figure 1 . A total of 5,513 publications was identified after the electronic and hand search, up to the 1 st of July 2020. After discarding the duplicates from the output, a total of 2,395 publications underwent title abstract selection whereupon 2,071 did not meet the inclusion criteria. A total of 324 articles remained for full-text analysis and, of these, 281 did not meet the in- and exclusion criteria, or they were a follow-up of an earlier study. A last update was done on the 1 st of December 2020 and, of the 111 new results, 1 was suitable for analysis. A total of 44 publications was finally included for data extraction. There was substantial 7