Twenty-eight studies (2,825 patients) were included: 12 RCTs and 16 non-RCTs. Four studies reported adequate methods for allocation concealment. Six studies used a power calculation. Eighteen studies achieved baseline comparability. Eleven studies blinded outcome assessors and three studies blinded patients. Ten studies reported results for a minimum of 85% of the recruited patients. Where reported, follow-up ranged from five days to five years.
There was no statistically significant difference between groups for Harris hip score (MD 1.49, 95% CI -0.08 to 3.06; 17 studies) or for surgical duration (MD -4.65, 95% CI -9.45 to 0.15; 14 studies). There was a statistically significant difference in incision length (MD -7.56, 95% CI -8.17 to -6.95; 17 studies, units not reported) and less perioperative blood loss (MD -42.4, 95% CI -60.1 to -24.7; 12 studies, units not reported) in favour of minimally invasive surgery when compared to conventional surgery.
There were no significant differences for surgical duration (MD -4.65 minutes, 95% CI -9.45 to 0.15; Ι²=91%; 14 studies), drained postoperative blood loss (six studies), total blood loss (15 studies), requirement for blood transfusion (seven studies) and all six radiological outcomes analysed.
Minimally invasive surgery was associated with greater incidence of iatrogenic nerve palsy (OR 5.3, 95% CI 2.6 to 10.9; 13 studies). The result remained significant for transient lateral femoral cutaneous nerve injury (three studies) but not for sciatic nerve palsy (three studies). There were no significant differences for any of the other complication outcomes.
The authors reported that there was limited evidence for publication bias (the funnel plot was slightly asymmetrical).
The forest plot for the primary outcome (Harris Hip score) appeared to present results for a different outcome (surgical duration).