Literature search
A total of 491 records was preliminarily identified after systematically searching the mentioned databases based on our search strategy using different search term combinations. We screened the titles and abstracts and excluded 456 irrelevant records. We scrutinized full texts of the remaining 35 papers for eligibility, and 14 papers, on a total of 1588 elderly patients with ITFs, remained for meta-analysis finally. The detailed selection process is shown in Figure 1.
Characteristics of included studies
Table 1 summarizes the characteristics of included studies. A total of 774 IF and 814 AR cases with ITFs in the elderly from 14 studies were included. Seven studies were from the Asia (ChinaChang,Shen,Tang,Yang, South KoreaKim,Kim,Park),25-31 6 studies were from the Europe (TurkeyKayali, Görmeli, Desteli, Özkayın,Güven, FranceBonnevialle),32-37 and 1 study was from AfricaGashi.38 Among these eligible trials, the AR included hemiarthroplasty and total hip arthroplasty, the IF included intramedullary system (GN, PFN, PFNA, IN) and extramedullary system (CHS, DHS), which represented the commonly used implants in clinical practice. Table 2 presents the quality of the included studies. They included three RCTs (Kim 2005; Desteli 2015, Özkayın 2015); two nonrandomized prospective studies (Bonnevialle 2011; Gashi 2018), and nine retrospective analyses (Kayali 2006,Chang 2011,Shen 2012,Kim 2014, Tang 2012, Görmeli 2015, Park 2015, Güven 2016, Yang 2017).
Outcome measures
Operative time
Ten studies covered the operative time,25-27,29,30-35,38 but only the data from 8 studies,25,26, 29,30,32-35 including 753 patients were eligible in the form of mean and standard deviation. High heterogeneity was present in these studies (Chi2=129.74; P<0.00001; I2=95%). The sensitivity analysis indicated that no significant change was found in the results after the deletion of any literature, and suggested the result is reliable. Therefore, a random effect model was conducted, and the result showed that there was a significant difference in the operative time between two groups (SMD= -0.80; 95% CI= -1.47, -0.12; P =0.02) (Fig.2). Compared with the IF group, AR group needed more operative time.
Blood loss and Transfusion
Blood loss was documented in 5 studies,25,27,29,31,33 and 2 of them including 201 patients were eligible in the form of mean and standard deviation. 29,33 The result of fixed-effect model showed that there was a significantly different in the blood loss between the two groups (SMD=-4.08; 95% CI=-4.58, -3.59; P<0.00001) with no heterogeneity (Chi2=0.20; P=0.65; I² = 0%) (Fig.3). IF group was superior to AR group in the blood loss.
Eight studies covered the blood transfusion,25-27,29,30,32,33,38 but only 4 studies of them provided the media volume of transfusion.26,29,30,32 358 patients were included in the remaining 4 literatures which were eligible in the form of mean and standard deviation. The heterogeneity test indicated there was a moderate heterogeneity (Chi2=10.40, P=0.02, I2=71%). However, sensitivity test was unnecessary because significant difference was observed in each trial, indicating the result reliable. The result of random effect model showed that the IF group has less transfusion than AR group (SMD= -0.67; 95% CI= -1.08, -0.26; P=0.001) (Fig.4).
Complications-related general condition
The general complications after surgery were reported in 8 studies (1221 patients, 189 events),26,27,29,30,32,33,37,38 which mainly included pneumonia, cardiovascular complication, urinary tract infection, neurologic complication, thromboembolic complication, pressure sores and wound infection. The heterogeneity test showed a low level among these studies (Chi2=8.75, P=0.27, I2=20%) and fixed-effect model was adopted. The incidence of postoperative complications-related general condition in AR group (91/634, 14.4%) was slightly less than that of the IF group (98/587, 16.7%), but there was no significant difference between the two groups (OR= 1.24; 95% CI= 0.90, 1.70; P=0.19) (Fig.5).
Complications-related operation
Ten studies,25-27,29,30-32,35,37,38 on a total of 1212 patients, covered major method-related complications, which mainly included the dislocation, loosening, cut out, breakage, intra-operative fracture of femur, postoperative fracture of femur, nonunion, delayed union and malunion. The heterogeneity test showed a low level among these studies (Chi2=14.48, P=0.11, I2=38%) and fixed-effect model was adopted. The result indicated that there is a significant difference between AR and IF groups in the complications-related operation (OR= 2.21; 95% CI= 1.41, 3.45; P=0.0005) (Fig.6), and it was clear that AR is superior to IF for the complications-related operation.
Reoperation
The rate of reoperation was reported in 7 studies.27,29, 31,32, 33,37, 38 The total reoperation rates in the AR and IF groups were 3.83% (19/496) and 9.83% (45/458), respectively. A fixed effect model was followed to conduct due to a low heterogeneity (Chi2=8.07; P=0.15; I² = 38%), and the result showed that the AR group has a significant lower rate of reoperation than IF group (OR= 2.74; 95% CI= 1.57, 4.76; P=0.0004) (Fig. 7).
Hospital stay
Nine studies covered the hospital stay,25,27,29-31,32-34,38 but just 6 studies including 575 patients were eligible in the form of mean and standard deviation.25,29,30,32-34 The heterogeneity test showed a high level in these studies (Chi2=73.14, P<0.0001, I2=93%) and random effect model was conducted. The sensitivity analysis indicated that no significant change was found in the results after the deletion of any literature, suggesting the result reliable. A random effect model was followed to conduct, and the result showed that there is no significantly different between two groups in the hospital stay (SMD= 0.16; 95% CI= -0.50, 0.82; P =0.64) (Fig. 8).
Harris hip score
Nine studies reported the HHS after surgery,25-27,29,31,33,35,36,38 but just 7 eligible studies including 704 patients in the form of mean and standard deviation were performed in the forest plot diagram.25,27,29,33,35,36,38 The heterogeneity test showed a high level among these studies (Chi2=101.52, P<0.00001, I2=94%) and random effect model was conducted. The sensitivity analysis indicated that no significant change was found in the results after the deletion of any literature, suggesting the result reliable. The result indicated that there is no significant difference between two groups in the HHS (SMD= -0.12; 95% CI= -0.79, 0.54; P=0.71) (Fig. 9).
Mortality within 1-year
A total of 10 studies reported the mortality after surgery,25-27,28,29,31-33,37,38 but just 8 of them, on a total of 1919 patients, provided the valid mortality with 1-year after surgery. 25,27,29,,31-33,37,38 A fixed-effect model was adopted due to no heterogeneity in these studies (Chi2=3.78, P=0.88, I2=0%). The result showed that IF has a statistically less mortality with 1-year than AR (OR= 0.70; 95% CI= 0.51, 0.97; P=0.03) (Fig. 10).