The principal finding of this pooled analysis was that there were no apparent differences in the incidence of postoperative complications after pediatric FNF between the ORIF and CRIF groups, especially for ONFH, nonunion, coxa vara deformity, LLD, and PPC. Given the consensus that the quality of reduction is the most consistent predictor of successful treatment for displaced FNF [4, 5], we believe that ORIF should be performed in FNF when it is required for anatomical reduction, especially on the fracture irreducible by closed manner.
A lack of randomization exists for the use of open reduction, raising concerns about the effects of bias in previous studies. In addition, in the present synthetic analysis, 8 of 15 studies presented that the decision to use ORIF was determined by whether FNF was reducible using the closed method. As ORIF is performed for fractures that fail CRIF, these fractures may have been more displaced, or more difficult fractures to begin with. Therefore, we cannot interpret the finding that no significant difference exists between ORIF and CRIF.
ONFH in children is the most common complication following FNF, with a reported incidence of 0–92% in the literature . Wang et al.  reported that age and initial displacement were independent predictors of ONFH in pediatric FNF and insisted that, to avoid ONFH, adequate reduction is more important than the type of reduction. Upadhyay et al.  demonstrated that no significant differences existed in the quality of reduction and ONFH between the ORIF and CRIF groups in their prospective, randomized study. However, they excluded four patients in whom an acceptable closed reduction could not be obtained, and these patients were managed using open reduction. Song studied ORIF not by the fracture pattern but by the study period, where CRIF was performed regardless of the fracture pattern during the “CRIF period” . Under these conditions, Song reported that CRIF without considering the fracture pattern demonstrated poor reduction and more complications, including ONFH.
Nonunion in pediatric FNF is much less common than in adult FNF because of the presence of a thick functional periosteum, but once nonunion develops, it is associated with poor patient outcomes . Including very recent study from Wang et al., initial displacement, delayed surgery, inadequate reduction status, and poor fixation are the major causes of nonunion [34,35,36]. Ju et al.  concluded anatomical reduction must be achieved after FNF to reduce the incidence of nonunion, regardless of whether CRIF or ORIF is used. In their meta-analysis, Yeranosian et al.  suggested that the approach to the treatment of pediatric FNF should become more interventional because of the high complication rate with unsatisfactory reduction over the years.
Consistent emphasis has been placed on anatomical reduction in both ONFH and nonunion. Although no significant differences were found between ORIF and CRIF for both ONFH and nonunion in the present study, the outcomes of the ORIF group were confounded by the large number of more severe fractures. We believe that it is important to achieve anatomical reduction regardless of the reduction type.
In the current study, the incidence of coxa vara deformity, LLD, and PPC also showed no differences between the ORIF and CRIF groups. Although these conditions may occur alone or in combination [21, 28, 32], the consensus is that the incidence of coxa vara deformity is related to reduction status, which means that its likelihood could be diminished with anatomical reduction and internal fixation [34, 38]. Although our current synthetic results did not reveal the superiority of ORIF over CRIF with respect to complications, theoretically, ORIF is more beneficial than not in treating FNF because it is required to avoid an unacceptable reduction status, which can lead to serious complications.
Specific fracture patterns of irreducible FNF have been reported, and a more invasive approach has been recommended for these types of FNF . However, all included studies classified fractures according to the Delbet classification system, which divides fractures according to their location . Using this classification, it is difficult to determine whether a satisfactory reduction is possible using only CRIF. Before multiple attempts at closed reduction, it is better to identify which fracture patterns are impossible to repair with CRIF and make an appropriate decision. Therefore, further studies are required to confirm this hypothesis.
The current meta-analysis has several limitations. First, although a satisfactory number of studies were included, all were retrospective in nature. Pooling the results of predominantly retrospective studies may overestimate the outcomes. Nevertheless, considering that our study is the first meta-analysis to provide a general overview of this topic and no publication bias was observed, our synthetic results are meaningful. Second, as limited data were available, we could only conduct the meta-analysis for all variables as dichotomous data and not as continuous data, especially for the degree of coxa vara deformity or LLD. Although the current study showed no significant differences in these variables between the types of reduction, converting these variables to continuous data may make a statistically significant difference. Third, owing to the characteristics of the meta-analysis, we could not control for fracture severity as a confounding factor, even with our best efforts to reduce the bias from each included study. Therefore, we need further high quality studies to analyze these issues more clearly.