Database confidentiality, image acquisition, and retrieving
In this retrospective observational study, consecutive CBCT images were retrieved and investigated from a CBCT database possessed by the Department of Dentistry, Tri-Service General Hospital, National Defense Medical Center, Taipei, Taiwan (from November 2013 to December 2016). All images were not taken specifically for this project. The qualified images (subjects and teeth) that met the inclusion and exclusion criteria were selected for analysis (Supplemental Figure 1 and Supplemental Table 1). The protocol used in this study was reviewed and approved by the Ethics Committee and Institutional Review Board of Tri-Service General Hospital, National Defense Medical Center (TSGHIRB No. 2-102-05-064).
The CBCT machine (NewTom 5G; QR, Verona, Italy) was operated by board-certified radiologist following the standard manufacturer’s settings as previously described [10, 12]. The skull orientation and region of interest were taken according to previous studies [10, 15, 22]. The maxilla was bilaterally symmetric and the occlusal plane, either in the frontal or sagittal view, was parallel to the ground (Supplemental Figure 2A). The acquired CBCT images were saved in a Digital Imaging and Communications in Medicine (DICOM) format, and these data were confidentially protected.
Inclusion and exclusion criteria of selected CBCT images
The CBCT images had to fulfill the following inclusion and exclusion criteria as previously described [10, 23]. The inclusion criteria had to be as follows [10, 23]:
• Permanent mandibular central incisor, lateral incisor, or canine had to be fully erupted with fully formed apexes;
• Each examined tooth had to be normally positioned with normal alignment, with a harmonious incisal line across the mandibular anterior teeth;
• Opposing maxillary teeth had to be present to provide information for optimal implant angulation and inclination;
Subjects were excluded if images showed [10, 23]:
• Bone screws and plates for surgical treatments, or any grafted materials;
• Preexisting alveolar bone destruction, perforation, dehiscence, or a combination of these caused by periodontal disease or traumatic injury around the investigated region;
• Supernumerary or impacted tooth;
• A pathological lesion, or evident root resorption;
• Incompletely formed apex;
• Dental misalignment, or preexisting dental implant;
• Signs of prosthodontic treatment, root canal treatments, and/or apical surgery;
• Obscurity or distortion due to scattering, or beam-hardening artifact reasons.
Assessment and classification of the crestal and radicular dentoalveolar phenotype (CRDAP) of anterior mandibular teeth
The following reference points and landmarks were defined on reconstructed images by viewing a sagittal-sectioned image of the region of interest and the center section of each investigated tooth (Fig. 1a) [24].
• Point A and Point B: the mid-facial and mid-lingual CEJ of the tooth, respectively;
• Point C: the intersection point of line A-B and line connecting the incisal edge and root apex;
• Length of the root (L): the distance between root apex and the Point C;
• Measuring point 1 (MP1): the point located 4 mm apical to the CEJ on facial root surface;
• Measuring point 2 (MP2): the point located middle of the root on buccal root surface.
Two dentoalveolar zones, “crestal zone” and “radicular zone”, were defined with minor modification to fit the purpose of this study accordingly (Fig. 1b) [25]. The crestal zone of dentoalveolar bone was defined as the region from the facial CEJ extending to a point 4-mm apical (MP1). The radicular zone was dependent upon individual root length and was defined as the region from MP1 to the MP2 (Fig. 1). The thickness of facial dentoalveolar bone on both crestal and radicular zones could be determined as either thick or thin phenotype [25]. For thick phenotypes, the facial bone thickness was defined as ≥ 1 mm, whereas, for thin phenotypes, the thickness was < 1 mm.
Four classes of crestal and radicular dentoalveolar phenotype (CRDAP) of mandibular anterior teeth were categorized according to the thickness of dentoalveolar bone at both crestal and radicular zones at the tooth level (Fig. 1b) [25]:
• Class I: both the crestal and radicular dentoalveolar zones were thick phenotype.
• Class II: the crestal zone was thick, but the radicular zone was thin phenotype
• Class III: the radicular zone was thick, but the crestal zone was thin phenotype
• Class IV: both the crestal and radicular dentoalveolar zones were thin
phenotype.
Measurements of morphological features of mandibular anterior region
The following morphologic and dimensional parameters of the mandibular teeth and alveolar ridge were measured accordingly (Fig. 2) [10, 17, 26].
• Concavity depth (CD): the distance between the deepest point of the facial bone plate (point Q) and a vertical reference line perpendicular to the mandibular plane, passing through the most external point of the labial plate (point P) (Fig. 2a).
• Concavity angle (CA): the angulation between line Q-P (that is, the line connecting points Q and P, with point P defined as the most external point of the labial plate) and line Q-R (that is, the line connecting points Q and R, with point R defined as the most external point of the labial plate inferior to point Q, and relatively lower than the apex of the tooth along the apico-coronal continuum) (Fig. 2a).
• Torque (T): the angle formed between the long axis of a tooth (that is, the line connecting incisal edge and root apex of the tooth) and long axis of the mandible, connecting from root apex to point D (that is, the lowest point of the mandible) (Fig. 2b) [26].
• Deep bone thickness (dBT): mandibular deep bone thickness was measured at 0, 5, and 10 mm from tooth root apex and along the long axis of the mandible (that is, the line connecting the tooth apex and point D), presented as dBT (0), dBT (5), and dBT (10), respectively (Fig. 2b). Each measurement line was also perpendicular to the mandibular long axis, then the distance between the facial and lingual outline of the bone was measured [26].
Virtual implant selection, placement, and definition of labial bone perforation (LBP)
The root form taper-designed dental implants were selected from an implant database available in the CBCT software as previously described [10, 12, 16]. The diameter of dental implants was determined by mimicking the corresponding size of each investigated tooth root, 3.0 mm for central and lateral incisors and 4.3 mm for canines (NobelActiveTM, Nobel Biocare, Gothenburg, Sweden) respectively, which were commonly used values for anterior mandibular teeth [27].
A selected dental implant was virtually placed along the long axis of the investigated tooth root with 4 mm of implant anchorage into native bone that was considered the minimum necessary to achieve primary stability [28, 29]. After placing the suitable implant virtually, labial bone perforation (LBP) or non-perforation were determined. The LBP was defined when the virtual implant extruded out of the apical outline of the labial cortical bone in the cross-sectioned and axial-viewed images (Supplemental Figure 3A) [14]. On the contrary, non-perforation was defined as the virtual implant within the apical outline of the labial cortical bone (Supplemental Figure 3B).
Qualification and examination of CBCT images
All selected CBCT images of 1920 × 1080 pixel resolution, displayed on a 19-inch liquid-crystal display monitor (ChiMei Innolux Corporation, Taiwan), were examined by a commercially available three-dimensional (3D) navigation software (ImplantMax® 4.0; Saturn Image, Taipei, Taiwan) in a dimly lit environment. To ensure data reliability and reproducibility, all images were re-oriented so that the morphology of the crown and root, from cementoenamel junction (CEJ) to the apex, in the sagittal planes could be investigated undoubtedly as described previously (Supplemental Figure 2B) [22].
Calibration and reliability between intra- and inter-examiners
The CBCT images were carefully inspected to follow the eligibility criteria by two independent investigators (first and corresponding author) twice, 1 week apart. Prior to the study, intra- and inter-examiner calibrations were performed on 50 randomly selective images based on the diagnosis of anatomic landmarks and anatomical measurements from CBCT images to assess data reliability. The nominal variables (e.g., perforation vs. non-perforation, CRDAP classification, Supplemental Table 2) and continuous variables (e.g., concavity depth, torque, dBT (0, mm), dBT (5, mm), dBT (10, mm)) and calibrated measurement errors, Cronbach’s alpha, and ICC (intraclass correlation coefficient) between intra- and inter-observations were also summarized (Supplemental Table 3). The Kappa statistic values for CRDAP and labial bone perforation were 0.940 and 0.929 for intra-observer agreement, and 0.929 and 0.932 for inter-observer agreement, respectively (Supplemental Table 2). Furthermore, the measurement errors, intraclass correlation coefficient, and Cronbach α values were also performed to confirm the reliability of intra- and inter-observations for continuous variable measurement (Supplemental Table 3). After calibration, the two independent investigators (first and corresponding author) evaluated the images separately, and any disagreement in image interpretation was discussed until a consensus was reached.
Statistical analysis
The occurrences of LBP were expressed as the percentage of the number of sites with perforation divided by the total number of corresponding investigated sites. Pearson’s chi-square tests were used to examine differences with categorical variables, such as the frequency distribution of four types of CRDAP classification and the LBP of the investigated tooth in the anterior mandibular zone. Shapiro-Wilk test was used for testing the normality of data (data not shown). To compare the values of CD, CA, T, and deep bone thickness at perforation and non-perforation sites, Mann-Whitney U tests were performed for the non-normality. After multi-collinearity tests were performed to examine the linearity between each two variables, dBT (5) was removed from final regression analysis (data not shown). For analyzing the risk of LBP, multivariable logistic regression analysis with generalized estimating equation (GEE) method was used to handle repeated measurements of tooth sites in each subject which simultaneously adjusted for within factors of person and tooth sites, categorical (i.e., gender, tooth type, and crestal and radicular dentoalveolar phenotype (CRDAP)) and continuous (i.e., age, cavity depth (CD), cavity angle (CA), torque (T), and deep bone thickness (dBT)) variables. Four models were applied where Model 1 was adjusted for within factors (as our univariate model), Model 2 was adjusted for within factors, gender, and age, Model 3 was adjusted for within factors, gender, age, CD, CA, T, dBT (0), and dBT (10), and Model 4 was adjusted for within factors, gender, age, CA, CD, T, dBT (0), dBT (10), and CRDAP (CRDAP class III teeth were excluded due to no “perforation” teeth in this group). These adjusted variables in sequential models were determined to follow a theme from within factors, subject’s factors, and then anatomic factors influencing labial bone perforation with evidences proven by previous studies. All statistical analyses were performed by SPSS for Windows (PASW Statistics, version 18.0, SPSS, Inc., Chicago, IL, USA) and the level of statistical significance was set at p < 0.05.