A total of 149 articles were included in the retrieval process, and the records after duplicate removal were 87. Furthermore, after excluding the case reports, case series, review articles, conference abstracts, expert opinions and incomplete data sets, a total of 23 full-text articles were ultimately included in the final review.
Both Garden type I and II fractures are NDFNFs. Garden type I fractures can be treated conservatively, but misdiagnosis and osteoporosis may lead to secondary displacement . The complication rate of type II fractures is 1.5 times higher than that of type I fractures. In a recent retrospective study, Rzesacz et al  revealed that conservative treatments were associated with a higher complication rate as compared to surgical treatments. Therein, the incidence of complications in the conventional treatment group was as high as 30-40%. Moreover, Cserhati et al  also revealed that the complication rate, especially the osteonecrosis of femoral head (ONFH), after conservative treatment was 18.9%, which was much higher than that of surgical treatment (3.2%). Since NDFNFs also involve fracture lines, the CT scan shows the pattern of the fracture more clearly for treatment plan. In addition, compared with the use of Garden classification alone, it is more effective to combine the three-dimensional (3D) reconstruction of hip joint with Garden classification to fully evaluate the NDFNFs. With the advantages of intuitive and multiple-angle vision, 3D reconstruction can display the anatomical shape of the femoral neck and is superior to X-ray in diagnosis and displacement judgment of fracture. The 3D reconstruction of hip joint can be used to reconstruct the exact status of fracture, so as to assist the surgeons to judge the fracture type according to the Garden classification and provide scientific basis for the development of treatment plan and assessment of prognosis [10,11,12].
Compared to the surgical treatments, conservative treatments are associated with much more mobility restrictions, bed-related complications, family care, etc, which increase the total cost, morbidity, and mortality [8, 9, 13].
Surgical treatments mainly include internal fixation and joint replacement (especially hemiarthroplasty or total arthroplasty), which are systematically reviewed and summarized in the following sections.
Timing of internal fixation
Kim et al  studied the timing at which patients underwent the internal fixation. Among 58 patients with a mean age of 74 years, the 1-year mortality rate in the patients who had undergone the internal fixation within 24 h after admission was only 4.7%, which was lower than in patients undergoing internal fixation 24 h after admission (13.3%). The 1-year mortality rate was 7% in patients who underwent the internal fixation within 48 h after the admission, and 72% of the patients recovered their walking ability to their pre-fracture level. Thus, an internal fixation performed shortly after admission (within 48 h after admission) may achieve better clinical outcomes and have fewer complications, such as nonunion and late segmental collapse of the femur.
Many implants are used for treating NDFNFs, including cannulated screws, cancellous bone screws, dynamic hip screws, targon system, emerging dynamic locking plates, and full-thread headless compression screws. The posterior retroversion angle is closely related to the prognosis of the patients with NDFNFs. The retroversion of femoral head and frequently associated comminution of the posterolateral wall compromise blood supply of the femoral head [15, 16]. Palm et al  reviewed 113 patients with NDFNFs. One year after surgery, 25 patients (22%) had a posterior retroversion angle of femoral head more than 20°, and the failure rate of internal fixation was 56%; 78% of patients had a posterior retroversion angle less than 20°, and the failure rate of internal fixation was 14%. The increased posterior retroversion angle is a risk factor for the re-operation of femoral neck fractures. Besides, Yamamoto et al  also suggested that some NDFNFs on X-ray may combine with the posterior retroversion angle of more than 20° on CT images. The posterior retroversion angle of more than 20° is intimately related to the poor prognosis of femoral neck fractures.
Cannulated screw is one of the ideal choices for the fixation of NDFNFs in elderly. Chen et al  treated 37 patients (over 80 years old) with NDFNFs using cannulated screws. Two years after surgery, the total healing rate was 95%, and the total success rate was 84%. The main reasons for revision were the failure of internal fixation, nonunion of fracture, and ONFH [20, 21]. Moreover, the high frequency of revision was also related to the factors such as poor bone quality in the elderly, severe osteoporosis, and surgical techniques .
Fixation with the cancellous bone screw is also one of safe and effective surgical approaches. Lee et al  conducted a retrospective study on 116 elderly patients who were treated with cancellous bone screws. Two years after surgery, 85% of the patients recovered to pre-injury activity level, and 90% of patients had no pain. In a retrospective study, Manohara et al  found that 90% of patients could independently walk postoperatively, 20% of patients felt pain during exercise, and the nonunion rate was 3%. Chiu et al  also showed that the non union rate of NDFNFs in patients treated with cancellous bone screws was only 2%.
Dynamic hip screws effectively treat NDFNFs with few complications and a low re-operation rate. Compared to the cannulated screws, patients treated with dynamic hip screws had a higher Harris score one year after surgery and a lower re-operation rate [25, 26]. Watson et al  conducted a comparative study on 62 patients treated with dynamic hip screws and cancellous bone screws, and there were six deaths (19.3%) in both groups. The re-operation rate (3.2%) was lower in the dynamic hip screw group than in the cancellous bone screw group (10.3%). In addition, Makki et al  studied elderly patients aged 70 on average and found that compared to the dynamic hip screws alone, dynamic hip screws combined with anti-rotation screws may not reduce the incidence of ONFH and revision rate.
Recent studies have revealed that those systems were associated with a low incidence of postoperative complications in NDFNFs . Compared to traditional screws, the targon system is associated with a lower revision rate and prevents femoral neck-head fragment settlement. Moreover, compared to the cannulated screws (48.8%), the revision rate of targon system within the first postoperative year is 4.7% . Compared to the dynamic hip screws, the targon system is associated with low rates of femoral neck-head fragment settlement and fewer hemiarthroplasty due to fixation failure . In addition to the targon system, the emerging dynamic locking plate is also widely used. Van Walsum et al  used emerging plates for NDFNFs and found that only 6 cases failed among 149 cases.
Most patients experienced the postoperative femoral neck shortening, resulting in the varus hip and poor gait function. Chiang et al  treated 50 NDFNFs using partially-threaded cannulated screws and full-threaded headless compression screws. The two implants provide stable fixation but cannot prevent femoral neck shortening and varus displacement.
Dynamic magnetic resonance imaging (MRI) and surgical selection
In NDFNFs, nonunion and late collapse of the femoral head are the two major complications. Patients aged 69 or younger have a high risk of ONFH after percutaneous screw fixation . Moreover, preoperative traction also increases the risk of postoperative ONFH . The blood supply to the femoral head is a significant factor affecting fracture healing. Morimoto et al  examined blood perfusion of the femoral head using dynamic MRI enhanced integral color imaging and classified the perfusion states into four types. With type A, the color of fracture area was consistent with that of the healthy side, indicating that its blood perfusion was the same; with type B, the color of the fracture area was darker than the femoral head area of the healthy side, indicating that the blood perfusion was reduced; with type C, the fracture area was black, indicating that there was no perfusion in this area. When they treated their patients with three cannulated screws, the nonunion rates of type A, B, and C were 0, 6.7%, and 50.0%, respectively, and the collapse rates of the femoral head were 0, 4.4%, and 0, respectively. For type C blood perfusion, joint replacement should be the first option, and a revision operation should be avoided if all possible.
In recent years, hemiarthroplasty is increasingly performed in elderly NDFNFs. According to the statistics of the Norwegian fractures data center, NDFNF patients treated with hemiarthroplasty rose from 2.1% in 2005 to 9.7% in 2014 .
Although different types of internal fixation have been widely acknowledged in the treatment of NDFNFs, Lin et al  found that the revision, nonunion and delayed union rates were higher in patients over 75 years old with femoral neck fracture receiving internal fixation than in those with hemiarthroplasty, and the prognosis was also worse. Lu et al  conducted a randomized controlled trial comparing hemiarthroplasty and internal fixation for NDFNFs in 78 patients (aged 85 to 100 years). The re-operation rate of the hemiarthroplasty was significantly lower than that of internal fixation (5.41% vs. 21.4%). Only the surgical methods had a significant impact on the occurrence of re-operation as shown by a Cox proportional hazard regression model. There were no significant differences between the two treatments in survival time and mean Harris score 5 years after surgery. However, hemiarthroplasty resulted in a significantly higher excellent-to-good rate. In addition, given nearly one-third of the elderly patients have combined dementia, following the postoperative rehabilitation instructions is difficult. Therefore, internal fixation may be a better choice in these patients .
Regarding this, Olofsson et al  conducted a retrospective study including 180 elderly patients (all aged above 70 years) with NDFNFs and dementia. They found no difference in the incidences of complications and mortality at 4 months and 1 year of follow-ups after the internal fixation and hemiarthroplasty. Dementia may not be a contraindication for hemiarthroplasty. Moreover, because the artificial femoral heads gradually degenerate, the postoperative survival of the autologous femoral head may be longer [42,43,44]. Currently, however, no study has compared the long-term outcomes between artificial and autologous femoral heads.
Controversies on the choice of surgical methods
Currently, the optimal treatment for NDFNFs is still controversial because few RCTs with a high level of evidence were conducted. The major considerations include postoperative function, complications, reoperation rate, and total cost [26, 34, 45,46,47]. You et al  indicated that internal fixation was associated with mild pain and better patient satisfaction but was associated with a high re-operation rate (20%). Moreover, hemiarthroplasty was a cost-effective option compared to internal fixation ($23,467 vs. $25,356) . Hemiarthroplasty was also associated with less complications, lower re-operation rate, and better early functional recovery . However, the downsides include longer operative time, more intraoperative blood loss, and higher intraoperative risk [48,49,50]. Dolatowski et al  conducted an RCT and revealed that hemiarthroplasty didn’t outperform internal fixation in the re-establishment of the postoperative hip function. However, we believe that, with rapid development of IT technologies, 3D printing, computer-assisted navigation and other new technologies will find their application in orthopedic practice [51,52,53,54,55,56]. 3D printing-related personalized technology and the computerized navigation might offer effective solutions to the selection of surgical methods for the NDFNFs in the elderly.