Did the dislocation risk after primary total hip arthroplasty decrease over time? A meta-analysis across six decades.

backgroundWhile continuous optimization is attempted to decrease the incidence of dislocation after total hip arthroplasty (THA), dislocation remains a major complication. This meta-analysis aims to analyze the evolution of the dislocation risk after primary THA over the decades and to evaluate its potential publication bias. PATIENTS AND

methodsA systematic search was performed according to the PRISMA guidelines for this meta-analysis in the literature published between 1962 and 2020. MEDLINE, Cochrane and Embase databases were searched for studies reporting the dislocation risk and length of follow-up. Studies that reported on revision rates only and did not mention separate dislocations were excluded. All study designs were eligible. Study quality was assessed by existing quality assessment tools adjusted for arthroplasty research. Overall risk and yearly dislocation rates were calculated and related to historical time frame, study design, sample size and length of follow-up.

resultsIn total, 174 studies were included with an overall moderate quality. In total there were 85.209 dislocations reported in 5.030.293 THAs, showing an overall dislocation risk of 1.7%, with a median follow-up of 24 months. The overall dislocation risk classified per decade decreased from 3.7% in 1960-1970 to 0.7% in 2010-2020. The yearly dislocation rate decreased from 1.8 to 0.7% within these same decades. There was no significant correlation between the reported dislocation risk and the duration of follow-up (p = 0.903) or sample size (p = 0.755). The reported dislocation risk was higher in articles with registry data compared to other study designs (p = 0.021).

conclusionThe dislocation risk in THA has been decreasing over the past decades to 0.7%. Non-selective registry studies reported a higher dislocation risk compared to studies with selective cohorts and RCTs. This indicates that the actual dislocation risk is higher than often reported and 'real-world data' are reflected better in large-scale cohorts and registries.