Bifid Mandibular Canal: A Proportional Meta-Analysis of Computed Tomography Studies

Introduction Growing body of evidences showed different grades in prevalence of bifid mandibular canals. Because the previous reviews focused solely on patient-level occurrence, hemi-mandible-level prevalence, bilateral symmetry, length, and diameter of bifid mandibular canals were required to be estimated collectively. The research question of this meta-analysis was “What is the prevalence of bifid mandibular canal among patients seeking computed tomography examinations”? Materials and Methods In vivo, computed tomography, and cross-sectional studies were eligible. Studies, with less than 100 subjects or anatomic site restriction or controlled class of bifid mandibular canal, were excluded. Joanna Briggs Institute (JBI) critical appraisal tool for prevalence studies was used to assess methodological quality of all included studies. Random effect meta-analyses for proportion of bifid mandibular canal were done. Results 40 studies met the inclusion criteria. All studies were selected for both systematic review and meta-analyses. Totally, 17714 patients and 31973 hemi-mandibles were included. All eligible studies showed moderate risk of bias on average. Resulting from the random effect model, more than 20% of patients seeking computed tomographic examinations had bifid mandibular canals (BMCs) which penetrated into slightly more than 14% of hemi-mandibles. Of the patients having bifid mandibular canals (BMCs), nearly 23% exhibited such anatomy on both sides of their mandibles. Estimated mean length and diameter of the accessory canals of bifid mandibular canals were 12.17 mm and 1.54 mm, respectively. Conclusion The geographical locations, classifications, reliability test, and voxel size of computed tomography were all implicated in the prevalence of bifid mandibular canals along with gender and laterality, although considerable heterogeneity and bias were detected.


Introduction
Te formation of bifd mandibular canal (BMC) originates from two fundamental theories. More recent hypothesis was inspired from the investigation of Chaves Lomeli and colleagues [1]. Tey speculated that bifurcation of mandibular canal may be molded by partial fusion of three accessory mandibular canals (Figures 1(a)-1(c)). Tey confrmed that these three canals were supplying tooth germs of mandibular deciduous incisors, deciduous ,molars and permanent frst molars in the mandible of human fetus. However, they did not mention how these canals fused with each other.
Another one, a historical fnding, was explored by observation of Serres. Tis French embryologist examined the vein, in one additional mandibular canal, draining below main canal [2]. Tis canal was confrmed by one recent cadaver report [3], but the authors did not report it as Serres's canal. Also, one human embryonic study [4] corroborated the existence of Serres's vein draining both pterygoid venous plexus posteriorly and venous plexus at mandibular symphysis anteriorly along with Meckel's cartilage (Figure 1(d)) of human embryo. Unfortunately, all these studies never stated where the mechanism of forming such anomaly came from.
Prior to forming BMCs, there are numerous amounts of anatomical variations along the extra-osseous course of inferior alveolar nerve (IAN) branching pattern. Proximally, IAN was found entrapped in the muscle fbers of inferior head of lateral pterygoid muscle [5]. Occasionally, it attained neural anastomosis with lingual nerve, long buccal nerve, auriculotemporal nerve, retromolar nerve, and mylohyoid nerve [6] before entering into the mandibular foramen. Also, at this entrance, it may seldom be penetrated by the maxillary artery [7]. In addition, progressive bone resorption in the edentulous mandible may expose main mandibular canal [8] and accessory lingual canal in the anterior mandible [9].
Subsequently, the mental nerve, an intra-osseous branch of IAN, could encroach anteriorly from its branching point. Tereby, it turns backwards and exits through the mental foramen, forming a loop called anterior loop [10]. In coronal section of CBCT, it can be seen as a numerical "8" character. Around a mental foramen, the existing mental nerves may be accustomed to multiple openings called accessory mental foramina [11]. Te most common position of these foramina was documented at the location posterior and inferior to the main mental foramen [11]. So, along the route of IAN, several anatomical variations can take part as a series of events. Tese may become associated with unavoidable clinical complications during oral surgical procedure.
Inadequate local anesthesia of IAN can be associated with presence of BMC [12]. Tis aberrant anatomy was also found in ffteen percent of the patients with postoperative neurological disturbance after mandibular third molar extraction [13]. Interestingly, one systematic review fgured out that BMC was seldom entrapped between the roots of mandibular third molar [14]. In this case, a more complex treatment option was needed to be planned to undergo third molar extraction. Surprisingly, inferior alveolar nerve was thought to have neurological anastomosis with long buccal nerve through retromolar canal [6]. So, infrequently, aberrant long buccal nerve was investigated during last molar surgery [15]. Hypoesthesia, partial or total loss of sensation, of buccal gingiva was found to be an unpleasant consequence of such procedure. Presence of neurovascular bundles in retromolar type of BMC was confrmed by one cadaver study [16]. Occasionally, inferior alveolar neurovascular bundle was recorded to become injured by endodontic procedure, implant installation, and osteotomy procedure [12]. As a result, we should keep in mind that anatomical variations of inferior alveolar nerve should be identifed preoperatively with a proper investigation method.
In comparison with and stating cone beam computed tomography (CBCT) as a reference standard, sensitivity of orthopantomogram (OPG), in detecting BMCs, ranged from 11% to 76% [13,17,18]. By defning magnetic resonance imaging (MRI) as reference gold standard, only 16.67% of BMCs were found in CBCT [19]. Sometimes, bone marrow [20], instead of vein, artery, and nerve (VAN) assembly, was also observed in the accessory canal of BMC in histological section.
Several numbers of systematic reviews and meta-analysis showed diferent grading in prevalence of BMCs. At patientlevel prevalence of BMCs, Shan and coworkers recently investigated 38% [21]. Valenzuela and associates analyzed 57% of this anomaly at individual level [22]. Also, Hass and colleagues identifed 16.25% of this aberrant anatomy at patient level [23]. One literature review stated that BMC ranged between 0.05% and 69% of the population [12]. In the review of Shah and Mehta [24], retromolar canal, one type of BMCs, revealed 3.2% to 93.5% of dry mandibles. Castro and fellows [25] contributed classifcations of BMCs. Tey focused mainly on the radiographic methods of current classifcations. However, units of analysis, such as patientlevel or hemi-mandible-level prevalence of BMCs, and similarity of diferent classifcations were not considered in such review. Nearly all the reviews included had a wide variety of research methodology, not objectively on computed tomographic examination.
Mostly, these reviews comprised especially patient-level incidence of BMCs. Hemi-mandible-level occurrence, bilateral symmetry, length, and diameter of this variation were required to be noted and pooled proportionally. Also, laterality, sexual dimorphism, prevalence across diferent continents, and diferent classifcations of such variation were still questioned to be found as combined efect size (pooled proportion). Additionally, we also needed to know how reliability test before computed tomography examination and voxel size of CBCT were infuencing the prevalence.
Finally, the question of this meta-analysis was "What is the prevalence of bifd mandibular canal (BMC) among patients seeking computed tomography examinations"?
Te objectives were (i) To observe patient-level prevalence of BMCs (ii) To fnd hemi-mandible-level prevalence of BMCs (iii) To identify bilateral symmetrical distribution of BMCs (iv) To estimate mean length and diameter of BMCs

Assessment of Methodological Quality.
We investigated the research methods of the included studies using the Joanna Briggs Institute (JBI) critical appraisal tool for systematic reviews of prevalence studies. 9 questions were included in the appraisal. 9 th question of these was excluded. As a consequence, eight questions were retained.
Ten, we categorized the identifed articles into three subgroups: JBI score (8, 7, and 6), (5 and 4), and (≤3). JBI 8, 7, and 6 were consistent with low risk of bias, 5 and 6 were consistent with moderate risk of bias, and less than or equal to 3 was consistent with high risk of bias [26]. Te percentage of JBI score gained by each category was calculated by the following formula: (the summation of JBI scores obtained from each study/total JBI scores) × 100. Finally, average JBI score of all included studies was estimated.
Subsequently, the research methodology of all included studies contributed to chart about the domains of frame of sampling, calculation of sample size, methods of sampling (convenience, randomization, and consecutive sampling methods), reporting of setting detail, and reliability measure before computed tomography examination, validity of measurement instrument, coverage of sample, and completeness of outcome reporting.

Statistical
Analysis. Tree formulas [26] for the corresponding objectives (Ι, ΙΙ, and ΙΙΙ) were (number of patients with BMCs/total number of patients) × 100 for objective Ι, (number of hemi − mandibles with BMCs)/(total number of hemi − mandibles) × 100 for objective ΙΙ, and (number of patients with bilateral presence of BMCs/total number of patients with BMCs) × 100 for objective ΙΙΙ.
Te abovementioned numerator and denominator variables were put into Excel spreadsheet of MetaXL software to undergo meta-analysis by the random efect model. Standard errors (SEs) for mean lengths and diameters (objective ΙV) of BMCs were calculated by the following formula [27]: SE � (SD/ � n √ ), where SD � standard deviation and n � sample size. Ten, mean lengths and diameters together with their corresponding SEs and number of observations were all put into an Excel spreadsheet of Meta-Essentials software to meta-analyze. Te generic inverse variance method and random efect model were used for such analysis [27].
To explore heterogeneity, subgroup analyses by the random efect model were conducted through male versus female, right versus left, patient-level prevalence of BMCs across diferent continents, hemi-mandible-level prevalence of BMCs across diferent continents, patient-level and hemi-mandible-level prevalence of BMCs among the different classifcations, Naitoh's classifcation, Norje's Classifcation, and Langlais's classifcation, and BMC with two mandibular foramina (Langlais ΙV or Norje ΙΙΙ). Heterogeneity was measured with I 2 statistic for proportion of BMC [27].
Sensitivity analysis by random efect model was done by excluding the studies which did not undertake reliability test before CBCT examinations. Te rest of the studies with calibration test were subjected to meta-analysis again.
Moderator analysis was carried out by correlating the voxel of CBCT in millimeters and prevalence of BMCs. For this analysis, standard errors (SEs) for proportions of BMCs were calculated by the following equation [27]: SE � ����������� (P(1 − P)/n), where p � proportion of BMCs and n � sample size. Ten, proportions of BMC prevalence, their corresponding SEs, number of observations, and values of voxel size (moderator) were put together into Excel spreadsheet of Meta-Essentials software to be metaanalyzed. Regression lines were drawn, and the random efect model was used for this analysis.

Publication Bias Test.
Publication bias tests were performed through objectives Ι, ΙΙ, ΙΙΙ, and ΙV by inspecting funnel plots. X-axis of the plot was arcsine prevalence and Yaxis was standard error for objectives Ι, ΙΙ, and ΙΙΙ. For objective ΙV, X-and Y-axes were efect sizes (mean length or diameter) and standard error.
For objectives Ι, ΙΙ, and ΙΙΙ, visual inspection of funnel plot asymmetry was justifed by Doi plot and LFK index [28]. Doi illustrates "no asymmetry" concerning with lack of publication bias, "minor asymmetry" indicating small amount of bias, and "major asymmetry" confrming presence of publication bias. ±1 LFK index reveals the certainty of publication bias [28].
For objective ΙV, funnel plot was intended to be repaired by the trim-and-fll method [29]. Tis reveals how many studies are needed to be flled to neutralize pooled efect size when asymmetry (publication bias) is present.
All the analyses were accomplished in MetaXL and Meta-Essentials software. Te extensions of Naitoh's classifcation, which were not included in Table 1, were as follows. Inferior bifd canal (bicanal): an accessory canal branching inferior from main mandibular canal and then running forward.
TMC (trifd mandibular canal) was not counted for the meta-analysis when primary studies had reported both BMC and TMC.

Results
All stages of identifying and selecting the records were illustrated in the fow diagram ( Figure 2).
Forty studies [13,18,, which met the eligibility criteria, were chosen for both methodological quality assessment and quantitative meta-analyses.
Te detected research methodology of all included studies is summarized in Figure 3.
Population, country, number of patients with genders, number of hemi-mandibles, age of the patients, geographic location, various defnitions of BMC, settings, and study design of the eligible studies are described in Table 2.
Totally, 17714 patients were identifed from the included studies of the review. 6475 males and 7947 females were reported. 31973 hemi-mandibles were found for this review. Age of the patients ranged from 6 to 103 years.

Patient-Level Prevalence of BMCs.
Tirty six studies [33-62, 64-68, 70] revealed patient-level prevalence of BMC. Te total number of dental patients in the meta-analysis was 17239 of which 2985 had BMCs.
As a result, BMC was more investigated on the right side of mandible than in the left predominantly (chi-square statistic: 5.1607, p value � 0.023103).  (2) 117 case reports, (3) 148 pre-surgical assessment for third molar surgery, (4) 61 DTA and agreement studies, (5) 225 BMC restricted to specifc anatomical location, (6) 99 of literature reviews, (7) 12 which were not allowed to be downloaded and interpreted, (8) 12 whose samples were < 100, (9) 26 whose method was OPG and (10) 13 in-vitro, (11) 8 book chapters, (12) 2 BMC associated with infammation, (13) 9 presurgical assessment for implant prosthesis, and (14) 3 letter to editor, opinion and comparative dental anatomy 60 of full-text articles screened for eligibility 20 of full-text excluded due to (1) 3 which set the separate defnitions for BMC & RMC, (2) 3 which used unclear working defnitions of BMC, (3) 3 which do not allow for translation service, (4) 5 in which data calculations were incorrect, (5) 2 duplications which were published in national journals by mother languages, (6) 1 conference paper, (7) 1 that solely investigated coronoid canal, (8) 1 not peer reviewed previously and (9) 1 which restrictedly found RMC 40 studies selected for both systematic review and metaanalyses  Tere were no data to pool the estimates for both African and Australian populations.

Publication Bias.
Test for publication bias indicated that funnel plot asymmetry was found in objective Ι (patient-level prevalence of BMC) and objective ΙΙ (hemi-mandible-level prevalence of BMC) (Figures 10(a) and 10(b)). Tese fgures illustrated the absence of studies at right-hand top of the plots. "Major asymmetry" for both investigations was also detected in Doi plots. LFK indexes were 2.66 for Ι and 2.78 for ΙΙ.
"No asymmetry" of the plot was seen in objective ΙΙΙ bilateral symmetrical distribution of BMC. Tis fgure showed symmetrical distribution of the included studies. −0.92 was detected as LFK index for such case.
Publication bias was not found in the meta-analysis of the pooled estimated length of BMC (objective ΙV). On the other hand, it was investigated in the analysis of the pooled diameter of BMC (objective ΙV). Te trim-and-fll test confrmed that three studies were needed to be flled on the left side of the funnel plot (Figure 10(c)).

Summary of Main Findings.
Resulting from the calculation, more than 20 out of hundred patients undergoing computed tomography examinations had BMCs. Tose BMCs penetrated into 14% of hemi-mandibles. Of the patients having BMCs, over 23% exhibited bilateral distribution of such anatomy in the mandible.
Patient-level prevalence of BMCs ranged from 1% among people in Lucknow of India [18], 2% in Brazilians [66], and 3% in both Rasht population of Iran [56] and Samsun people of Northern Turkey [58] to 54% of Alexandria Egyptians [36] and 58% in Taiwanese Chinese in New Taipei City [55], and the peak was 67% in North Cyprus of Turkey [44]. At the hemi-mandible level, the prevalence of BMCs ranged from 1% in both Indian [18] and Brazilian populations [66] and 2% in both Turkish [58] and Iranian populations [56] to 42% in both Taiwanese [55] and Eastern Anatolia population of Turkey [34] and 43% in Cairo population of Egypt [63], and the climax was 46% in Northern Cyprus population of Turkey [44].
We emphasize that the extreme variations were seen in the Turkish populations at both patient level and hemimandible level.
Symmetrical occurrence of BMCs ranged from 0% in Brazilian population [38] and 2% in Pathum Tani people of Tailand [35] to 50% in Alexandria Egyptians [36] with the highest occurrence of 69% in Shenzhen population of China [59]. We notice that although Turkish populations were involved in scoring the upper tier of both patient-level and hemi-mandible-level occurrence of BMC, their constitution was almost 40% in the case of bilateral symmetry [44], ranking after China.
Te accessory canals of BMCs lengthened to more than twelve millimeters in the populations of the included studies in our meta-analysis. Te mean lengths of these accessory canals ranged from 7.1 mm in the Spanish patients at University of Santiago de Compostela [70] to 16.9 mm in South Korean population.
Specifcally, forward, retromolar, and buccolingual canals took the longest length in Yemeni [41], South Korean [40], and Spanish [70] populations, out of other classes of Naitoh's classifcation. Some investigators fgured out that accessory canals of BMCs were longer in premolar region than in molar [63].
On average, the canals widened to over 1.5 millimeters in diameter in the populations of the selected primary studies. Te mean diameters of these accessory canals ranged from   0.9 mm in Taipei population [52] to 2.28 mm in Yemeni people of Sanaa City [41]. Comparatively, retromolar and dental canals occupied the widest, among the categories of Naitoh's classifcation, in South Korean [40] and Yemeni [41] populations. Te accessory canal with large diameter was more prone to be detected in OPG [70]. One study mentioned that the caliber of vessel being more than one millimeter allows blood fow to fow at three milliliters per second [71]. Te diameter may broaden to 3.4 millimeters in some cases [47].
BMCs were more commonly found in the right half of the mandible than the left. But, a limited number of studies had right side predominance of BMC specifcally in Taiwanese [55] and Turkish Istanbul populations [39]. Left side dominance of such morphology was detected only in Milan people of Italy [67]. Tis subgroup efect has never been identifed signifcantly in the rest of the studies [36, 37, 42-45, 49-52, 56-58, 60, 61, 64-66, 69, 70]. We suggest the biased distribution may result from the increase in sample size of the meta-analysis, previously mentioned in sex diference.
A quarter of Europeans, nearly one-ffth of Asians, and one in ten of American population had BMCs. Subsequently, this anatomy tunneled in approximately 33% (Africa), 17% (Europe), more than 13% (Asia), and over 7% (America) of hemi-mandibles across the world. However, there were only two studies [36,63] conducted in Egypt. Tis should not be considered as a true representative of all African nations.
Also, patient-level data from Africa and Oceania can never be accessed in this review.
At the patient level, the prevalence of BMCs by Naitoh's classifcation was 6.2% greater than Norje's classifcation and 11% more common than Langlais's categories. At the hemimandible level, the prevalence of such bifd anatomy classifed by Naitoh's classifcation was 5.6% more frequent than Norje's classifcation and 15.3% more abundant than Langlais's classifcation.
As a result, BMC was enumerated as more abundant proportion in categories of Naitoh's classifcation than the other two classifcations. We think that CBCT technology was progressively advanced in recent decade immediately after Naitoh and colleagues had used CBCT and invented their classifcation system. Collectively, this classifcation counted an additional coronal section of computed tomography image in contrast to other classifcations. Buccolingual type of Naitoh's classifcation can be detected in this section.
Additionally, inferior bifd canals also known as bicanals accounted for 2.2% in our meta-analysis. Both Elnadoury [36] and Shen [48] reported this anatomy up to more than 4%. Although Saket [63] had not reported complete picture of BMCs, he presented the picture of inferior bifd with confuence type in his article. Of the accessory mandibular canals classifed by Naitoh, 8.5% rejoined into main mandibular canal (confuent type) in our meta-analysis.
Less than one percent of hemi-mandibles orchestrated two mandibular canals originating from two separate mandibular foramina (Langlais ΙV or Norje ΙΙΙ). Although most of the investigators [33][34][35][36][37] had not presented such type of anatomy, some [48] reported that 6% of accessory canal of BMCs drained outside the mandible by separate foramina openings.
We hypothesize that inter or intra-rater reliability test before CBCT examinations could have a slight positive infuence on prevalence of BMC. Because of half of the selected studies not having undergone the test, the pooled estimate of this meta-analysis may be underestimated.

Moderator Analysis.
By undergoing moderator analysis between voxel size of CBCT and hemi-mandible-level prevalence of BMCs, Z value exceeded 1.96 and 0 was not included in the confdence limits of slope (Table 4). Tis indicates the signifcant positive association between hemi-mandible-level prevalence of BMC and voxel size of CBCT. Te voxel sizes were reported ranging from 100 to 400 µm [13, 18, 33-35, 37-44, 46, 47, 49, 50, 52, 53, 56, 58, 63, 65, 66, 70]. Although smaller voxel size potentiates the resolution of CBCT, the reverse can be proved by the analysis. However, the reviewers could not explain why the studies having reported high degree of prevalence of BMCs used larger voxel sizes.

Agreement and Disagreement with the Previous Reviews.
Haas and colleagues [23] previously investigated that patient-level prevalence of BMC was 16.25% by CTor CBCT. Tis is obviously lower than the fndings of our analysis. Te pooled prevalence of BMCs in both in vitro studies and the studies, which used OPG as investigation method, comprised 6.5% and 4.2% in Haas's meta-analysis [23]. Also, these disagreements may be due to dissimilarity between research methods used in selected studies.
Valenzuela-Fuenzalida and coworkers [22] found 57% of BMCs in more than 4000 mandibles. Tis is superior to the fnding of our analysis displaying more than 20 percent of more than 17000 mandibles. Te previous analysis [22] manipulated not only CBCT studies but also cadaver studies and the studies that used dry human skulls. So, we make the assumption that the proportions of BMCs may be infated due to smaller sample size of primary studies, distortion of specimen, and fragility of dry mandibular bone.
Ngeow and Chai [12] showed patient-level prevalence of BMCs from 0.05% to 69%. Te range was complementary with our fnding which ranged between 1% and 67%. Tey also observed mean length of accessory canal of BMCs ranging from 7.1 mm to 16.9 mm. Tis is in agreement with our analysis again. Te previous review also pointed out that the mean diameters of the accessory canal ranged from 0.9 mm to 2.2 mm. We identifed that our range (0.9 mm-2.28 mm) was in turn in agreement with the previous fndings.
In the review of Shah and Mehta [24], retromolar canal comprised 3.2% to 93.5% of dry mandibles. Tis range was fve times wider than ours. As a result, we think that restriction to this subtype of BMC, dry bone sample, and methodological diversity may greatly overwhelm the prevalence of retromolar canal.
By comparing other aberrant anatomy of the mandible, Muinelo-Lorenzo et al. [11] summarized that accessory mental foramina were detected nearly 8% at the patient level. Additionally, Mishra and associates [11] showed that anterior loops of mental nerve were seen up to 41% at the subject level. So, we recognize that BMCs were found more than twice of the accessory mental foramina and less than half of the anterior loops at the individual level. However, we did not attempt to analyze the associations between these aberrant anomalies and BMCs in this review.
Additionally, Mishra's group [11] also stated that the average length of anterior loops ranged from 1.1 mm to more than 8 mm. Te upper limit of anterior loop's length could be overlapped with the lower bound of mean length of accessory canal of BMCs.
Castro and fellows [25] reviewed the classifcations of BMCs. Tey especially concentrated on the radiographic methods used in classifcation systems, two or threedimensionally. Tey also outlined locations of BMCs at mandibular ramus or at body of the mandible. Contrastively, from our standpoint, we emphasize on units of analysis at patient level or hemi-mandible level and similarity among diferent classifcations (Table 1).

Risks of Bias.
Although overall average JBI score of all included studies in this meta-analysis had been 54.69% (moderate risk of bias), some limitations were seen at the domains of sampling characteristics, sample size calculation, population coverage, reliability test, and outcome reporting.
Te worst domain that seems to be sufering from risk of bias was sample size calculation. Te prior estimation of sample was never attempted in 90% of the included studies ( Figure 3). Also, 65.85% of the studies used the records within inadequate time frame and took the sample from single centers or university, not from several centers. Tis may lead to under-coverage of target population and could not be true representative of such population.
Consequently, 60% of the eligible studies neglected population characteristics (age, gender, and ethnicity) to record thoroughly. 50% of the studies did not obtain inter or intra-examiner agreement test before CBCT examinations. Additionally, 45% of the studies did not report patient-level and hemi-mandible-level prevalence of BMCs, their bilateral distribution, and other suitable outcomes sufciently. Tese factors could be prone to misclassifcation of BMCs and incomplete outcome reporting.

4.7.
Heterogeneity. Heterogeneity grew considerably in case of the meta-analyses at patient-level prevalence, hemimandible-level prevalence, and bilateral symmetrical distribution of BMCs.
To explore the source of heterogeneity, we conducted multiple subgroup analyses. By doing so, we identifed some heterogeneities that originated from the classifcations used in primary studies. When we had categorized the studies into their corresponding classifcations, the meta-analyses demonstrated some relief from heterogeneity.
For Norje's classifcation, I 2 statistic decreased from 98% to 51% at the patient-level prevalence of BMCs and from 99% to 83% at hemi-mandible-level prevalence.
For Langlais's classifcation, I 2 dropped from 98% to 73% at patient level and from 99% to 87% at the hemi-mandible level of BMCs, respectively. Across the diferent classes of Langlais classifcation, the parameter decreased from 98% to 9% in Langlais Ι and to 0% in Langlais ΙΙΙ at the patient-level prevalence of BMCs.
At the hemi-mandible-level prevalence of BMCs, I 2 statistic of heterogeneity fell from 99% to 0% in case of inferior bifd type of BMCs and to 87% for the confuent type BMCs.
No substantial loss of heterogeneity was found in Naitoh's classifcation.
For African continent, I 2 reduced from 99% to 8% at hemi-mandible-level prevalence of BMCs. In such case, we speculate that it may be due to the scarcity of evidences in the African nations.
Finally, we conclude that the heterogeneity in prevalence of BMCs can be partly explained by the diferent classifcation systems used in selected primary studies.

Imperfections of BMC's Classifcations.
Among BMC's classifcations, Norje's [36] and Langlais's [37] categorizations were based upon two-dimensional X-ray examinations, while Naitoh and coworkers [35] had investigated by viewing three-dimensional computed tomography. Te most distinguished feature between the latter and former is the inclusion of coronal view in Naitoh's classifcation.
Tis additional view integrates the buccolingual type of BMCs in Naitoh's classes, which is never found in both Norje's and Langlais's contributions (Table 1). On the other hand, two mandibular foramina types were not seen in Naitoh's classifcation, although they had been previously set in Norje's and Langlais's classifcations.
Lateral lingual and median lingual canal stated in other studies [72] could be misinterpreted as buccolingual type of BMC. Dental canal of Naitoh's classifcation can lose its identity after extraction of corresponding tooth. So, it may be misunderstood as Naitoh's retromolar and forward types. We cannot mention precisely how nutrient canals and edentulous condition infuence the classifcation of BMCs in CBCT image.
Te questionable content and construct validity lead to imperfections of the classifcations. 4.9. Content of Accessory Canal of BMC. Vein, artery, nerve, and lymphatic drainage are major constituents of the main mandibular canal. However, in place of the assembly of vein, artery, nerve, and lymphatic drainage, only one large venous vessel supplying base of mandible [3] or nutrient vessels [80] or bone marrow [20] or multiple osteoporosis cavities [81] or proximal branching of mental nerve at the entrance of mandibular foramen [82] or remaining nerve plexus [83] of edentulous mandible can be present in the accessory mandibular canals of BMCs of cadaver sample. Tese structures can correspond to be radiolucency in computed tomography images.
Unfortunately, strictly bony radiographic architecture of the accessory canal can be seen in CBCT bone-contrast image. Soft tissue content of this additional canal cannot be found in the image.
As a result, MRI is the current gold standard method in observing not only BMC but also its contents. Blood vessels and nerve can be well diferentiated by viewing signal intensities of MRI because vein exhibits more intensifed features than nerve in MRI image. So, even VANL assembly can be detected in MRI [19].
Unfortunately, because of MRI being soft-tissue contrast, two mandibular foramina of BMCs may not be seen accurately in this image. So, some investigators [19] advised that they should be confrmed by CBCT, which is hard-tissue contrast, in this case.
Bifd mandibular nerves may not always occupy two mandibular canals. Te mandibular canal wall is mostly formed by facing trabecular bony plates inside while their bony pillar orienting outside [84]. Tis pillar-plate orientation could be destroyed by bone diseases. Furthermore, this proves that the mandibular canal wall does not possess specifc compact or specialized bony structure in nature although, not rarely, radiopaque line is seen along the course of this canal in formal radiographic examinations.
To the best of our knowledge, we conclude that bifd mandibular nerves may be present even in a single hollow of bone cavity. Also, this could be missed during routine radiological examination.

Publication
Bias. Tree studies [54,59,60] from mainland China were translated from Chinese to English, 1 study [42] and 1 thesis [65] from Peru were translated from Spanish to English, and 1 study [69] from Spain was translated from Spanish to English.
Although 5 studies of languages other than English [42,54,59,60,69] and 1 thesis (gray literature) [65] were included in this meta-analyses, major publication bias was subjectively seen in both patient-level and hemi-mandiblelevel prevalence of BMCs and pooled diameter of accessory canals of BMCs.

Future Studies.
In spite of progressive number of evidences investigating BMCs being found, pooled sensitivity and specifcity of CBCT in comparison with gold standard MRI in detecting this anatomy will be needed to be questioned and pooled.

Conclusion
Generally, 20.7% of patients seeking computed tomography examinations and 14.3% of hemi-mandibles displayed BMCs. Nearly 23% of those patients exhibited bilateral distribution of such specifc anatomy. On average, the accessory canal of BMCs lengthens up to 12.14 millimeters and widens to 1.54 millimeters. Sexual dimorphism towards male gender and right-sided predominance of the canal were seen together with high statistical power and sample size of the meta-analyses. Europeans were found to be the population in which BMCs were mostly investigated all over the world. Usage of Naitoh's classifcation and reliability tests may escalate the proportion of BMCs. We uncovered one unexplainable reason in which voxel size of CBCT may have positive correlation with prevalence of BMCs with no regard to considering other resolution parameters.

Data Availability
Te data supporting the fndings of this study are available from the corresponding author upon reasonable request.

Conflicts of Interest
Te authors declare that they have no conficts of interest.