Risk factors for implant failure of proximal femoral nail anti-rotation (PFNA –II) in the treatment of intertrochanteric fractures (AO/OTA 31 A1 and A2)

Intertrochanteric fracture is associated with severe morbidity and mortality. The results of postoperative implant failure are catastrophic. The aim of this study was to determine risk factors for implant failure in intertrochanteric fractures treated with proximal femoral nail anti-rotation (PFNA–II) through the assessment of early therapeutic effects. A single-center retrospective study was conducted on a continuous series of 123 intertrochanteric fracture patients treated with PFNA-II between Dec 2018 and Oct 2019. Perioperative medical and imaging data were collected. The patients were divided into two groups according to whether implant failure of not. The differences in reduction quality, nail length and tip apex distance (TAD) were analyzed to determine the risk factors of implant failure.


Introduction
Intertrochanteric fractures are common in the elderly population with an incidence of 0.1% 1,2 . They are characterized by the high disability and high mortality, with a one-year mortality rate as high as 36% 3 .
Surgical treatment especially intramedullary nailing has been the common treatment method due to the central xation, less postoperative pain, and earlier recovery of mobilization 4,5 . Page et al. 6 reported that the rate of intramedullary nailing in intertrochanteric fracture has increased from 3.8% in 2011 to 57.6% in 2015.
At present, proximal femoral nail anti-rotation (PFNA-II) has become the main implant in the treatment of intertrochanteric fractures 7 . However, with the widespread use of intramedullary nails, the failure rate increases to as high as 13.3%~20.5% 8,9 . Salvage procedures are invariably challenging and expose a population to further complex surgery 10 . Given the gravity of the event, it is important to determine risk factors which contribute to and are associated with implant failure. Kraus et al. 11 reported that tip-apexdistance (TAD) greater than 30 mm was the main risk factor. Turgut et al. 12 believed that if the coxa vara cannot be corrected, the lag screw will be in a unfavorable position. Even if the TAD is small, implant failure may be inevitable. Imeric et al. 13 found that PFNA-II showed signi cantly higher failure rate in reverse intertrochanteric fractures. Hao et al. 8 reported that poor reduction quality and loss of posteromedial support are predictors of implant failure in reverse oblique and transverse intertrochanteric fractures treated with PFNA. Whereas, the risk factors for PFNA-II failure are still controversial. Moreover, the length less than 240 mm of PFNA-II are commonly used in Asia, but there was rarely reported the risk factors of short PFNA-II failures. Hence, in this study, a group of intertrochanteric fracture stabilized with short PFNA-II has been chosen for explored the risk factor of the implant failure. It may predict which patients might suffer a complication and guide surgeons in preventing the implant failure in the patients with intertrochanteric fracture treated with PFNA-II.

Operation procedure
The patients were lies on a uoroscopic surgical traction table in supine position after anesthesia (nerve block anesthesia is mostly selected to reduce the impact on cardiopulmonary function). First, the closed reduction was carried out by traction. The fracture site was relieved by abduction and external rotation.
Secondary, under traction, the affected limb was adduction and intorsion at the same time until the foot was slightly adduction and intorsion. The reduction quality was checked by uoroscopy. The disinfection and draping were made after reduction quality was accepted. PFNA-II was implanted into the femur according to the operation procedure provided by the manufacturer. If the reduction quality is poor, additional small incision assisted reduction by instrument. Suture the deep fascia and skin without drainage after intramedullary nailing.

Data Collection
General demographic data of patients were collected from medical record, including gender, age and so on. Image data were extracted from the imaging browsing system of the trauma center. The image data were analyzed by two experienced attending surgeon and the data was adopted when they agree with each other. Discrepancies were resolved by the chief surgeon.
The preoperative evaluation was evaluated by the American Society of Anesthesiologists (ASA score, and bone quality was evaluated by the Singh index of the contralateral hip 14 . The fracture type was described by AO classi cation 15 . The reduction quality adopts the standard proposed by Baumgaertner et al. 16 The caput-collum-diaphysis angle was normal or slightly varus in the anterior and posterior position of X-ray or the angle of fracture site on lateral radiography is not more than 20°. The distance between fracture fragment is less than 4 mm. It is de ned as good when the both conditions can be reached, accept when one condition can be reached and poor with none condition can be reached. The imaging standard of fracture union 17 was de ned as the blur of fracture line on the X-ray plain lm, and the continuous callus passing through at least three cortical bone of the fracture site. The implant failure 18 de ned as that: lag screw cut out, coxa vara (The caput-collum-diaphysis angle less than 5°above the contralateral hip) or nonunion, implant broken, periprosthetic fracture without high energy damage.

Assignment De nitions
For risk factor analysis, age, BMI, Singh index, reduction quality, ASA index and TAD were classi ed and assigned values (Table 1).

Statistical analysis
Statistical analysis was performed with SPSS for windows software (Ver.22.0, SPSS, Chicago, USA). The continuous parameters were expressed by mean with standard deviation (mean ± SD), and it was analysis by t test. The categorical parameters was expressed by percentage and it was analysis by χ 2 test. Analysis of risk factors were evaluated by logistic regression, and odds ratios (ORs) with their 95% con dence intervals (CIs) were also obtained. P<0.05 was de ned as signi cant in all tests.

Results
There were 80 patients met the inclusion criteria with 27 males (33.3%) and 53 females (66.7%). Average age of those patients was 80.2 years (range from 55-93) and the mean follow-up time was 12.3 months (range from 8 to 19). There were 37 patients with the left hip fracture and 43 for right. Implant failure occurred in 6 patients (7.5%). According to implant failure, the patients were divided into implant failure group and non-failure group. The general conditions of the two groups are shown in table2. There was no signi cant difference in demographic data between two groups, so the baseline characteristics were the same.
In implant failure group, there were 2 cases of A1 type intertrochanteric fractures and while 4 of A2 type, 4 case of nail length were 200 mm and 2 of 240 mm, the difference of fracture type and nail length between the two groups was no statistically signi cant (p>0.05). The implant failure rate was 5.7% in patients with good and acceptable reduction quality and 33.3% in patients with poor reduction. The difference between the two groups was statistically signi cant (p<0.05). The TAD was 18.3±6.1 mm in failure group and 22.3±6.3 mm in non-failure group. In term of hospital stay, the failure group was 10.3±3.3 days and the non-failure group was 8.9±3.2 days. The difference of TAD and hospital stay between the two groups was no statistically signi cant. The perioperative data of the two groups are shown in Table 3.
The reduction quality was the independent risk factor of implant failure through logistic regression analysis of multiple factors (p<0.05). It was found in table 3 that the incidence of implant failure of patients with poor reduction quality was 8.75 times higher than that of patients with good and acceptable reduction (table 4).
The failure of internal xation revealed that three patients had cut out, one patient had the periprosthetic fracture, two patients had slight coxa vara after reduction and lead to nonunion during the process of weight bearing. Among the patients with implant failure, we found that four patients have been not reconstructed the medial femoral support, two patients showed good reduction at rst in postoperative Xray but then they were found the reduction loss, hip varus and cut out of helical blade during the followup. In order to clarify the causes of implant failure, we conducted further analysis of postoperative X-ray and classi ed three typical types of xation effect in all patients. There was cortical support type (Fig 2), the helical blade support type (Fig 3) and the nail support type (Fig 4).

Discussion
PFNA-II has been the mainly method for the treatment of intertrochanteric fractures due to its advantages of short operative time and less bleeding 19 . AO/OTA 31-A1 and A2 accounted for 80%-94.7% 20 of all intertrochanteric fractures. The fracture line distribution of A1 and A2 is different from that of the reverse intertrochanteric fractures (A3), and the implant failure mechanism is also different with each other 21 .
Many studies have not analyzed these two types of intertrochanteric fractures separately 22 , which may lead to different results. For patients with A1 and A2, PFNA-II with nail length less than 240 mm is generally recommended 23 . Therefore, A1 and A2 intertrochanteric fractures stabilized with short PFNA-II (nail length less than 240 mm) is the most common clinical case. It is important to determine the risk factors of implant failure for guiding intertrochanteric fractures treatment.
In this study, the difference of reduction quality between the failure group and the non-failure group was statistically signi cant and the incidence of implant failure in failure group was higher than that in nonfailure group. This may be suggested that reduction quality and implant failure were directly related.
Morvan A et al. 24 retrospective analyzed 228 patients aged over 75 years operated by Gamma 3 and Dynamic hip screw (DHS). Cut-out rate was 0.89% in good reduction and 12.12% in poor reduction. To minimize mechanical complications, great attention should be paid to fracture reduction and lag-screw position. The same results found in De Brujin' s study 25 . In our study, reduction quality is risk factor of implant failure in intertrochanteric fracture treated with short PFNA-II. Implant failure rate was 33.3% in poor reduction and 4.5% in good and acceptable reduction. In multivariate analysis, the risk of implant failure in poor reduction was 8.75 times higher than that in good and acceptable reduction (OR = 8.75, 95%CI 1.215-62.99, p = 0.0313).
According to the postoperative X-rays, the medial femoral cortex was discontinuous in 66.7% of the patients with implant failure in this study. Posterior medial support is considered to be an important factor affecting the stability of intertrochanteric fractures, which has been con rmed in both basic and clinical studies. Do et al. 26 reported that xation strength was approximately 5 times greater in small defect of femoral medial cortex than in large defect of femoral medial cortex. Nie et al. 27 con rmed by biomechanics that medial reconstruction is more important than lateral wall reconstruction. Similar ndings have been found in clinical research. Mariani et al. 28 analyzed 20 patients with nonunion of intertrochanteric fractures and found that all of them were unstable fractures and the reason of implant failure was related to posterior medial comminuted. In this study, 66.7% of the patients with implant failure had poor medial reconstruction suggested that medial reconstruction is a keystone of intertrochanteric fractures (A1 and A2) treated with short PFNA-II.
Even if the continuity of the femoral medial cortex is restored, the implant failure such as cut out may still occur during the weight bearing. Our previous study proposed the concepts of secondary stability and cortical or implant support 29 . In this study, we con rmed this phenomenon ( Fig. 1 to 3). If the medial femoral shaft cortex or helical blade forms a support to the medial cortex of the head-neck fragment, the fracture is healed (Figs. 1 and 2). Otherwise, the fracture will result in loss of reduction, secondary stabilization may result in implant support (Fig. 3). According to our team's early research in recent years 30 , most of the anterior cortex of intertrochanteric fracture is simple fracture, so the anterior medial cortex needs to provide stable support. In patients with intertrochanteric fractures, osteoporosis is severe, and the strength of the anterior medial cortex is limited. In Ender's classi cation 31 , this fracture of varus deformity and retrocurvature proximal fragment impacted into trochanteric spongiosa, leaving a cavity after reduction danger of secondary varus was de ned as impacted fractures. In 2013, Gotfried et al. 32 also reported a similar phenomenon and de ned it as negative and neutral support. Since then, there have been a lot of relevant reports, but the reasons for the failure of neutral support have not been explained 33,34 . We found that the compressed area of cancellous bone in the medullary cavity forms a triangular void structure after reduction of head-neck fragment. If the medial femoral cortex fails to provide stable mechanical buttress during the weight bearing, the head-neck fragment is prone to varus again. Hence, we considered this triangular void structure may have a strong association with postoperative implant failure.
There are some limitations in this study. First of all, it was a retrospective study that the data which were not collected initially could not be evaluated. We only included a few common indicators for evaluation. Nonetheless, we have speci ed strict inclusion and exclusion criteria to reduce confounders and minimize interference with the results. Besides, the sample size is relatively small which may cause statistical bias.

Conclusion
Reduction quality is a risk factor for implant failure of short PFNA-II in the treatment of Intertrochanteric fractures (OA/OTA 31-A1 and A2). Anatomic reduction and reconstruction of medial femoral support are important means to reduce the implant failure. Even if the reduction is good, there might be still a triangle void area between PFNA-II and medial femoral cortex, which is prone to cause displacement of medial cortex, coxa vara or implant failure.     Cortical support is shown in a typical case. a: perioperative radiographs of intertrochanteric fracture. b postoperative radiographs showed that the medical cortex of the head-neck fragment was supported by the medical cortex of the femur. c: radiographs of six months after surgery showed fracture union without reduction loss. Figure 3 helical blade support shown in radiographs. a: intertrochanteric fracture with lesser trochanter free. b: The radiographs on the second day postoperative showed that the inferior cortex of head-neck fragment was blocked to varus by the helical blade (long red arrow). A gap was formed in fracture site (short red arrow). c: The femoral cervicodiaphyseal angle in three months was consistent with that the second day postoperative due to the helical blade support (long red arrow), but the gap in the fracture site was blurred (short red arrow).