The effect of root orientation on inferior alveolar nerve injury after extraction of impacted mandibular third molars based on propensity score-matched analysis: a retrospective cohort study

The sample size is set according to the previous literature [15‐17]. In this study, the sample size is originally set to 300 people. After the data analysis, we used the software PASS v15 (Power Analysis and Sample Size) to verify the sample size of this study. Using the “Tests for Two Proportions” in the software, we set α to 0.05 and β to 0.2, then input the incidence of IAN injury of the buccal group (P1), the incidence of IAN injury of the non-buccal group (P2) and the ratio of the case number of the two groups (N2/N1) into the software. The result shows that the total sample size needed for the study is 449.

This study is a retrospective cohort study with a follow-up period of 6 months. Patients with IMTMs extracted in the department of oral and maxillofacial surgery of the Hospital of Stomatology, Sun Yat-sen University, from December 2021 to November 2022, were screened in this study. The IMTMs with root-IAC contact were preliminarily screened by panoramic radiography and were further diagnosed by cone beam computed tomography (CBCT). A total of 539 IMTMs, including 268 IMTMs from the left side and 271 IMTMs from the right side, were chosen from 379 patients that included 124 males and 255 females and featured an average age of 29.1 years old (18 ~ 59 years old). The complete table of all IMTMs is shown in Supplementary Table S1. Considering that IAN dominates unilateral sensory and IAN injury is dependent on unilateral tooth extraction, each tooth was independently numbered for statistical analysis. Epidata v3.1 was used for sample data entry, and Microsoft Excel 2021 was used for data integration.

Inclusion criteria: (1) the patient having complete data of panoramic radiography and CBCT, (2) the imaging of CBCT showing root-IAC contact, (3) The patient met the indications for impacted mandibular third molar extraction, having no contraindication for tooth extraction and successful tooth extraction. (Refer to Supplementary Table S2 for indications and contraindications for impacted third molar extraction).

Exclusion criteria: (1) the patient having incomplete data of panoramic radiography or CBCT, (2) the imaging showing root-IAC separation, (3) the patient having his/her tooth partially extracted.

The collection of all the data of panoramic radiography and CBCT and the diagnosis of the impaction features of all IMTMs was conducted separately by two radiologists on the same computer in consensus. When there was disagreement, the final decision was made by consultation with the third radiologist. The three radiologists all had more than 5 years of expertise. The repeatability of both the radiologists who performed the diagnosis was evaluated by Cohen's kappa in SPSS v22. The results of Cohen's kappa for both radiologists were over 0.8.

When using CBCT for image analysis, the process of section selection is as follows:

The classification of IMTM is as follows:

All patients were informed before tooth extraction that IAN injury might occur after tooth extraction and signed a surgical consent form. All IMTMs were extracted by the same senior oral surgeon using the same surgical instruments under local anesthesia. The process is as follows: First, the doctor used articaine plus epinephrine for the inferior alveolar nerve, buccal nerve, and glossal nerve block anesthesia plus local infiltration anesthesia. Second, the gingiva was incised using a triangular flap design and the alveolar bone was removed using a dental handpiece. Then, the third molar was extracted using an elevator and dental forceps. Finally, the tooth extraction wound was irrigated with normal saline and sutured. and the bleeding was stopped by compression with cotton balls. Antibiotics were used after tooth extraction when necessary. When the stitches were removed a week post-operation, IAN-relevant maxillofacial areas were examined by 2-point discrimination, pin-prick, and light touch, to determine whether IAN injury occurred. Patients with numbness of the lower lip were followed up by telephone concerning IAN recovery at 2 weeks, 1 month, 3 months, and 6 months post-operation. The patient with disappeared numbness was asked for a subsequent visit for verification of IAN recovery.

The chi-square / Fisher’s exact test was used to compare the incidence of IAN injury among IMTMs with different impaction depths, root orientation, or contact degree of root-IAC. The correlation between root orientation and impaction depth/contact degree of root-IAC was analyzed based on the Lambda coefficient. The datasets were divided into the buccal group and the non-buccal group in terms of root orientation. Age, sex, impaction depth, and contact degree of root-IAC, were taken as confounding variables. Binary logistic regression was used for modeling. The propensity score of individual cases was calculated by the nearest neighbor matching method with no replacement, and the caliper was set to 0.2. The two groups (buccal and non-buccal) were matched at a ratio of 1:1. The Fisher’s exact test was used to analyze the difference in IAN injury between the two groups before and after PSM, and p < 0.05 was considered statistically significant. All data were analyzed in IBM SPSS Statistics (v22) and R (v4.1.2). To be precise, we used package "CreateTableOne" for baseline data statistics, package "tableone" to calculate the standard mean error (SMD) of each variable, and functions "glm" and "predict" to calculate each case's propensity score (PS). Then package "Matching" was used to match the cases with similar propensity scores at the ratio of 1:1 (caliper set to 0.2 with no replacement). And finally, packages "ggplot2" and "tidyverse" were used for plotting.

In this study, 21 cases of IAN injury with numbness of the lower lip occurred after IMTM extraction. The incidence of IAN injury was 3.90% (21/539). The IMTMs information was shown in Supplementary Table S1.

In terms of impaction depth, except that no IAN injury was found for IMTMs with complete eruption (0/25), the incidence of IAN injury for partial eruption, mucosal impaction, and osseous impaction, was 0.79% (1/127), 2.81% (5/178), and 7.18% (3/209), respectively.

In terms of root orientation, except that no IAN injury was found for the encircling group (0/15), the incidence of IAN injury for the buccal group, upper group, and lingual group was 7.62% (16/210), 1.92% (2/104), and 1.43% (3/210), respectively.

In terms of contact degree of root-IAC, the incidence of IAN injury for level I, II, III, and IV, was 1.18% (2/169), 3.08% (5/130), 4.11% (6/146), and 8.51% (8/94), respectively.

Taken together, there were significant differences in the incidence of IAN injury after IMTM extraction among different types of impaction depth (p = 0.006), root orientation(p = 0.008), and contact degree of root-IAC (p = 0.02). The results are shown in Table 1.

As demonstrated above, IMTMs with the root located on the buccal side of IAC were more prone to IAN injury. However, it was unclear whether the difference was attributed to the distribution unevenness of impaction depth and contact degree of root-IAC among different groups of root orientation. Therefore, the chi-square test of the contingency table was used to analyze the correlation between root orientation and impaction depth/contact degree of root-IAC, and the lambda coefficient was 0.276(p < 0.001) and 0.102(p = 0.003), respectively (the results are shown in Table 2). The proportion of level III and IV of contact degree in the buccal group was 70.1% (149/210), which was higher than that of 28.6% in the non-buccal group (94/329). Furthermore, the proportion of mucosal and osseous impaction in the buccal group was 80.5% (169/210), which was higher than that of 66.3% in the non-buccal group (218/329).

The correlation analysis revealed the necessity of reasonable statistical methodologies to balance the distribution of impaction depth, contact degree of root-IAC, and other confounding factors, among different groups of root orientation, for a better insight into the independent effect of root orientation on IAN injury.

Age, sex, impaction depth, and contact degree of root-IAC were considered confounding variables. Binary logistic regression was used to model and calculate the propensity score of each case. Using nearest neighbor matching with no replacement and setting the caliper to 0.2, the buccal group and non-buccal group were matched at the ratio of 1:1. There were 21 cases with IAN injury and 518 cases without IAN injury before PSM and 9 cases with IAN injury and 257 cases without IAN injury after PSM. The case number of buccal/non-buccal was 210/329 before matching and 133/133 after matching. The standard mean difference (SMD) of age, sex, impaction depth, and contact degree of root-IAC, was respectively 0.066, 0.457, 0.357, and 1.245 before PSM, and was respectively 0.05, 0.051, 0.122, and 0.101 after PSM, as shown in Table 3. It was demonstrated that PSM significantly reduced the interference of confounding factors. In addition, the plot of the Kernel density of the PS score before and after PSM was shown in Fig. 3, indicating that group-to-group comparability was significantly improved after PSM.

Before PSM, the incidence of IAN injury was 7.62% in the buccal group and 1.52% in the non-buccal group, and the difference was statistically significant (p <.001). The odds ratio (OR) was 5.344 and the risk ratio (RR) was 5.013. After PSM, the incidence was 6.02% in the buccal group and 0.75% in the non-buccal group, and the difference was statistically significant (p=.04), as shown in Table 4. It is evident that after balancing confounding factors, the risk of IAN injury of IMTMs whose root was located on the buccal side of IAN was significantly higher than that of non-buccal counterparts. The odds ratio (OR) was 8.448 and the risk ratio (RR) was 8.