Being one of the severest complications of middle ear surgery, iatrogenic facial nerve paralysis has always been the accident that every otologist tries his best to avoid. Fortunately, the incidence has declined. In our center, it has dropped from 0.13% (7/5491) to 0.04% (6/14552) over the past 2 decades. Worldwide, it was reported between 0.6% and 3.6% [4, 5] last century, and is assumed to be lower than 0.10% [3] at present. One reason contributing to this is the application of advanced surgical instruments, such as facial nerve monitor and high-definition microscope. However, iatrogenic facial nerve paralysis still happens now and then. When iatrogenic facial nerve paralysis happens anyway, knowing how to properly manage it will be beneficial to the recovery, that is the reason why we carry out this study.

For the injury happened at our center, most happened in the surgery for cholesteatoma of middle ear (4, 50.0%), because surgery for cholesteatoma was often more complicated than that of other middle ear disease [6]. Besides, most surgeries (37, 82.2%) and severe injury (partial or complete transection, 23, 95.8%) were conducted at outside nonacademic centers, lack of clinical experience and adequate temporal bone dissection training may contribute to the facial nerve injury.

Malformation of facial nerve. Dehiscence of fallopian canal is the most common high-risk factor; it may be congenital or eroded by lesion. The incidence was reported as high as 56% in an anatomy of 1000 temporal bones [7] and 30%-33% in cholesteatoma surgeries [8, 9]. In our study, the incidence was 62.5% in the cases injured at our center. The risk of dehiscence comes from exposure of facial nerve directly to drills, sharp surgical instruments, or even some irritating substances [10], and preoperative HRCT could help identify it. In addition, deformed facial nerve would make localization and identification difficult. In our study, anomalous course and bifurcation of facial nerve were also identified. History of congenital facial nerve paralysis is uncommon but may add risk to idiopathic facial nerve paralysis. One patient (No. 44 case in Supplementary Table 1) with congenital H-B grade II paralysis suffered from H-B grade V postoperative paralysis; during exploration, fallopian canal was intact and facial nerve was only mildly edematous, we assumed it may be increased nerve susceptibility.

Revision surgery. During revision surgery, localization of the facial nerve becomes difficult due to removed landmarks and fibrosis. In our study, four (8.9%) cases of injury happened in revision surgery.

Misdiagnosis. Some facial nerve tumors are difficult to distinguish from cholesteatoma or granulation tissues [11] (our cases were excluded by exclusion criteria), once misdiagnosis is done before surgery, facial nerve paralysis is very likely to happen. It is crucial an otologist bore in mind the lesion on the surface of facial nerve might be the tumor originates from the nerve trunk, although preoperative facial nerve function could be normal.

The most venerable site was tympanic segment (32, 71.1%), and many patients (22, 48.9%) suffered from multisegment injury, which were consistent with other report [12]. In cholesteatoma and chronic otitis media surgery, lesions are often near tympanic sinus, so distal tympanic segment, and sometimes the second genu are easy to be injured. The tympanic segment near oval window is prone to be injured when lesions locate near stape, especially when fallopian canal is dehiscent [13]. The geniculate ganglion should be attention to when clearing the lesions near the cochleariform process [14]. The injury of labyrinthine segment is rare; however, when clearing cholesteatoma above tympanic segment in the attic, labyrinthine segment is at high risk because facial nerve turns backward and inward from geniculate ganglion.

Mild injury normally only leads to nerve edema. In our study, 28.8% patients suffered from such injury. Edema could result from retraction of nerve, to avoid it, the lesions should be cleared along the nerve trunk. It could also come from compression by bone chips or hematoma, so careful examination and hemostasis before the end of surgery was important. Heat from drills, electrical stimulation from coagulator lead to thermal damage and nerve edema, so timely rinse is important to taking away the heat. Notably, two cases with canal intact were also found facial nerve mildly edematous after decompression. One of them suffered from congenital H-B grade II paralysis, we assumed it may be increased nerve susceptibility; the other one may suffered from herpes simplex virus reactivation, which was consistent with the pathology of Bell palsy [13, 15]; for there was no direct injury to the facial nerve, they might have recovered without exploration. In addition, prognosis of facial nerve with sheath defective is nearly as good as with edema because the axons remain intact. In our study, 13.3% patients suffered from sheath defective and none recovered to worse than H-B II. This kind of injury often comes from accident bruise in an unclear operative field, so try to rinse and identify the facial nerve, nerve is unlikely to be injured when it has been identified [16]. When the nerve cannot be distinguished from lesions, the facial nerve monitor can help, the current should not be excessive but be gradually adjusted to the threshold. What is more, facial nerve monitor cannot take the place of proficient anatomical knowledge and rich clinical experience despite its wide application.

The severe injury is direct transection, which accounts for 57.8% cases in our study. To avoid it, the movement of high-speed drill or micro cutting equipment should be along the course of facial nerve and appropriate sizes should be chosen for different lesions. Besides, half patients with nerve transection developed neuroma, more than 90% happened over 3 months after paralysis. Before nerve repair, traumatic neuroma has to be removed, which sacrifices the length of facial nerve and potentially makes direct anastomosis unavailable, so early exploration could reduce the possibility of autologous nerve graft.

Immediate and severe paralysis (H-B grade V or VI) should receive more attention, because the earlier and severer a paralysis, the more likely the nerve suffers from substantial damage. However, tight dressings and packing should be released and time allowed for any local anesthetic effects to dissipate [14], and some patients can still close eyes upon forceful eyelid closure even a few hours after a complete transection of facial nerve [3]. Once immediate and severe paralysis is diagnosed, whether exploration be conducted should be further weighed. If the otologist is certain of nerve integrity, such as the whole course of the facial nerve was identified in surgery or the fallopian canal remained intact, then a conservative approach can be adopted. When there are some uncertainties, HRCT scan could help identify the integrity of fallopian canal. If fallopian canal is not integral, exploration should be conducted at the earliest time, if integral, conservative treatment could be tried. However, if HRCT indicates no obvious injury but there is no sign of clinical and electrophysiological improvement after conservative treatment, exploration should also be conducted timely. In our study, all the patients suffered from immediate and severe paralysis, with no sign of improvement after conservative treatment, and no otologist was certain of nerve integrity. During exploration, a general rule is decompression for mild injury, re-approximation of axions for superficial transection, anastomosis for partial or complete transection, and graft is adopted when end-to-end tension-free anastomosis (even with rerouting) is not applicable.

The severity of facial nerve injury. The milder the injury, the better the outcome. Most of patients (85.7%) with mild injury recovered to H-B grade I and II, and none recovered to worse than H-B III. On the contrary, for those with facial nerve partially or completely transected, the best recovery was H-B III. The prognosis of anastomosis seemed better than graft because no patient recovered to worse than H-B III, but statistical analysis did not show significance (P = 0.575). Maybe it was due to limited sample size, or there was just no difference between them [17], in the future a larger series is needed to validate it.

Duration before revision surgery. For decompression, those received surgery within 2 months after paralysis had higher possibility of H-B grade I or II recovery than those over 2 months (P = 0.026), and for patients recovered to H-B I, seventy-five percent were also decompressed within 2 months. Prolonged edema or compression was found to reduce the probability of a satisfactory recovery [18, 19]. For graft, those received surgery within 6 months were more likely to get H-B III recovery (P = 0.041). This is supported by the theory that nerves regenerate at about 1 mm/day [20], and muscles undergo irreversible atrophy from 12 months after denervation [21], considering the distance for nerve regeneration from sectioned ending to distal motor endplate, graft had better been conducted within 6 months other than 12 months as previously recommended [22, 23]. As to anastomosis, all patients in our study were conducted within 6 months, and all recovered to H-B III.

Re-revision surgery. Four patients received exploration at nonacademic centers before transferring to our center for re-revision surgery, they all gained worse recovery than those without previous exploration (P = 0.001). The main reason may be additional surgical trauma from nonstandard revision surgical procedure and consequent delayed re-revision surgery.