Fifteen studies (n=1,890, of which 884 received total hip arthroplasty and 1,006 received hemiarthroplasty) were included in the review. Four were RCTs, three were quasi-randomised trials and eight were retrospective cohort studies. One RCT received a quality score of 9, one a score of 10 and two a score of 11. The three quasi-randomised trials scored 8, 6 and 5. Cohort studies scored between 1 and 6. Mean follow-up duration ranged between one and 156 months.
Primary total hip arthroplasty showed a significantly lower risk of reoperation rates compared with hemiarthroplasty (RR 0.57, 95% CI 0.34 to 0.96; 14 studies). Stratification by study design showed that findings were not statistically significantly different when the retrospective cohort studies (seven studies) or the RCTs (four studies) were combined separately.
Total hip replacement showed statistically significantly better function at 12 and 48 months compared with hemiarthroplasty using the Harris hip score (WMD -5.4, 95% CI -2.7 to -8.2; three studies) and medium functional advantage using different measurement scores (SMD 0.42, 95% CI 0.24 to 0.61; five studies).
There were no statistically significant differences in risk of dislocation, deep infections, general complications and one-year mortality between the two treatment groups, either overall or when different study designs were combined separately.
There was no evidence of significant statistical heterogeneity. Publication bias was evident for deep infections (p=0.027) and general complications (p=0.033).
Two studies that assessed health-related quality of life showed conflicting findings.