Fourteen studies were identified for inclusion in the review (2,219 patients, range 14 to 732). The six surgical studies were all case series (324 patients, range 14 to 126). The loss to follow-up for the eight non-surgical observational studies (1,895 patients, range 45 to 732) was 365 (19%, range zero to 32%). Follow-up ranged from 0.5 to 10 years for non-surgical patients and one to 4 years for surgical patients.
For the non-surgical group (eight studies), one study was excluded from the meta-analysis since all the patients with fracture non-union were found to have had inadequate (late) initial treatment. For the remaining seven studies, there was a significantly higher risk of fracture non-union for displaced fractures (RR 4.40, 95% CI 2.23 to 8.67; Ι²=54%) versus undisplaced fractures (OR 5.52, 95% CI 2.48 to 12.3; Ι²=55%).
Four of the six studies in the surgical group did not have a comparative group and had a 100% union rate after surgery, so a meta-analysis was not possible. Only two of the 157 displaced fracture cases in the six studies failed to heal after surgery.
The two-way contingency table analysis gave a significantly higher risk of non-union for treatment with a plaster cast versus surgery (OR 16.8, 95% CI 3.8 to 75.2).
There was no evidence of publication bias for the main analysis and findings were not significantly altered in subgroup analysis.