The aim of this systematic review was to investigate the influence of the presence and position of mandibular third molars on angle fractures. An electronic search was conducted in the PubMed, Scopus, Web of Science, Cochrane Library, and VHL databases, through January 2016. The eligibility criteria included observational studies. The search strategy resulted in 704 articles. Following the selection process, 35 studies were included in the systematic review and 28 in the meta-analysis. Twenty studies presented a score of ≤6 stars in the Newcastle–Ottawa scale assessment, indicating a risk of bias in the analysis. The presence of a mandibular third molar increases the chance of an angle fracture (case–control and cross-sectional studies: odds ratio (OR) 3.83, 95% confidence interval (CI) 3.02–4.85, I 2 = 83.1%; case–control studies: OR 3.27, 95% CI 2.57–4.16, I 2 = 81.3%). The third molar positions most favourable to angle fracture according to the Pell and Gregory classification are class B (OR 1.44, 95% CI 1.06–1.96, I 2 = 87.2%) and class II (OR 1.67, 95% CI 1.36–2.04, I 2 = 72.4%). Class A (OR 0.60, 95% CI 0.45–0.81, I 2 = 87.1%) and class I (OR 0.51, 95% CI 0.37–0.71, I 2 = 89.4%) act as protective factors for angle fracture. The results suggest that the presence of the third molar increases the chance of angle fracture by 3.27 times and that the most favourable positions of the third molar for angle fracture are classes B and II, whilst classes A and I act as protective factors.
The prevalence of mandibular fractures among facial fractures is high, reaching 76% of all facial fractures. The mandibular regions that fracture most frequently are the mandibular condyles (56.5%), mandibular symphysis (45.0%), body (25.5%), and angle (16.5%) . The locations of mandibular fractures are related directly to the fragility of these bone areas: the condyle is the mandibular area with the lowest bone thickness and is more frequently fractured. Furthermore, the mechanisms of incident forces on the mandible capable of causing fractures are also determining factors in the location of mandibular fractures .
The prevalence of angle fractures varies in different studies, with reported values of 16.5% , 24.5% , 27% , 30% , and 37% . This discrepancy can be attributed to the presence of the mandibular third molar . The third molar, when present, may generate a weak area in the mandibular angle and predispose this region to fracture . On the other hand, it may decrease the occurrence of fractures in the condylar region, which is usually the area of greatest weakness of the mandible .
In 2004, a meta-analysis was performed to verify the association between the presence of mandibular third molars and mandibular angle fractures , and the results revealed a 2.4-times higher chance of a mandibular angle fracture occurring when the third molar is present. Thus there has been evidence of the increased chance of mandibular angle fracture when the third molar is present since 2004. However, only six studies were included in that meta-analysis . Moreover, no forest plot was presented, nor was there an assessment of the chance of mandibular angle fracture occurring taking into account the different positions of the third molar. Twelve years have passed since the meta-analysis in 2004 and new articles on this subject have been published. It would thus be timely to review again the scientific evidence on this issue.
The aims of this systematic review were to evaluate the current scientific evidence on the relationship between the presence of mandibular third molars and fractures of the mandibular angle and to assess the relationships of the third molar with these fractures.
Materials and methods
This systematic review and meta-analysis was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. The clinical question was formulated using the PECO process (i.e., P = patient, problem, or population, E = exposure, C = comparison, and O = outcomes), as follows: the population comprised patients with mandibular fractures; the exposure was the presence or absence of a third molar and the different positions of the third molar (A, B, C, and I, II, III of the Pell and Gregory classification, and vertical (V), horizontal (H), mesioangular (MA), and distoangular (DA) of the Winter classification); the comparison was with other mandibular fractures; the outcome was an angle fracture.
The following inclusion criteria were applied in this systematic review: observational studies (cross-sectional, case–control, and prospective and retrospective cohort studies) that evaluated the influence of the presence and position of a mandibular third molar in mandibular angle fracture.
Exclusion criteria were as follows: (1) case reports, case series, opinion articles, articles that did not present an abstract in the databases, abstracts, and review articles; (2) studies that reported mandibular fracture during the extraction of third molars; (3) studies that evaluated the bad split fractures in orthognathic surgery due to the presence of the third molar; (4) studies including pathological mandibular fractures due to the presence of lesions associated with the presence of the third molar; (5) studies that included patients with diseases of bone metabolism (osteopenia and osteoporosis); (6) studies without a comparison group or without a statistical analysis comparing two or more groups.
The electronic survey was conducted in the PubMed, Scopus, Web of Science, Cochrane Library, and VHL (Virtual Health Library; BIREME (PAHO/WHO)) databases, and included publications through January 2016, without language restriction. A combination of medical subject heading (MeSH) terms was used for the search. The terms used in the databases were: (mandible fracture OR mandibular fracture) AND (third molar OR wisdom tooth OR wisdom teeth).
After searching the databases, the titles and abstracts of articles were read by two authors independently (ACVA and SGMF). Studies that could potentially meet the inclusion criteria for the review were identified at this stage. After independent reading, the authors compared results for the studies that would pass to the next stage (article read in full), reaching an agreement on their inclusion or not. In the case of disagreement, a third author (ELG) was consulted to obtain a consensus. The studies selected after reading the title and summary were read in full. At this point, it was determined whether or not the study should be included in the systematic review. The inclusion or not of the study was determined according to the predetermined criteria.
The assessment of the quality of the studies included was performed using the Newcastle–Ottawa scale (NOS) for case–control studies and a modified NOS for cross-sectional studies . The case–control studies were assessed for the following three components of the study: selection, comparability, and exposure. The cross-sectional studies were assessed for the following three components of the study: selection, comparability, and outcome. The NOS for both study types had a maximum possible score of 9 stars/points for each study.
The data extraction was performed in two stages. In the first stage, the following data were extracted: author, year of publication, study design, country where the research was performed, sample (male, female, and total), type of analysis (patients and fractures), mean age, and main cause of mandibular fracture (see Table 1 ). In the second stage, the following data were extracted: mandibular angle fracture (present or absent), third molar (present or absent), Winter classification when a third molar was present (V, MA, H, DA), and Pell and Gregory classification when a third molar was present, for the occlusal plane (A, B, C) and for the mandibular ramus (I, II, III) (see Table 2 ).
|Author||Year||Study design||Country||Sample||Analysis type||Age, years||Main cause of fracture|
|Abbasi et al.||2012||CS||Pakistan||87||15||102||Patients||29.85 ± 11.21||NR|
|Choi et al.||2011||CS||South Korea||333||52||385||Patients||NR||Violence|
|Mah et al.||2015||CS||South Korea||348||92||440||Patients||NR||Violence|
|Duan and Zhang||2008||CS||China||563||137||700||Patients||NR||Traffic accident|
|Fuselier et al.||2002||CC||USA||981 a||228 a||1209||Patients||30.8 ± 10.4||NR|
|Gaddipati et al.||2014||CS||India||95||15||110||Fractures||18–55||Traffic accident|
|Halmos et al.||2004||CC||USA||1449 a||1 a||1450||Fractures||30.6 ± 10.4||NR|
|Kim||2004||CC||South Korea||78||29||107||Patients||29.99 ± 11.46||Fall|
|Iida et al.||2004||CC||Japan||250||96||346||Patients and fractures||NR||Violence|
|Iida et al.||2005||CC||Germany||169||49||218||Patients||NR||Violence|
|Inaoka et al.||2009||CS||Brazil||38||5||43||Patients||26.35||NR|
|Kasamatsu et al.||2003||CC||Japan||121||30||151||Patients||29.9 ± 14.9||Traffic accident|
|Kumar et al.||2015||CS||India||123||19||142||Patients and fractures||NR||NR|
|Revanth Kumar et al.||2015||CS||India||56||8||64||Patients and fractures||16–69||Traffic accident|
|Lee and Dodson||2000||CC||USA||290||77||367||Patients||31.7 ± 10||Violence|
|Ma’aita and Alwrikat||2000||CC||Jordan||488||127||615||Patients||33.2 ± 11.4||Traffic accident|
|Meisami et al.||2002||CS||Canada||NR||NR||214||Patients||Violence|
|Naghipur et al.||2014||CC||Canada||377||69||446||Patients||29.3 ± 11.3||Violence|
|Patil||2012||CS||India||164||26||190||Patients||21–30 (46%)||Traffic accident|
|Rahimi-Nedjat et al.||2016||CC||Germany||915||304||1219||Fractures||32.5 ± 19.3||Fall|
|Rajandram et al.||2013||CC||Malaysia||186||28||214||Patients||27.36 ± 7.97||Traffic accident|
|Rajkumar et al.||2009||CC||India||136||18||154||Patients||33.2||Traffic accident|
|Zhu et al.||2005||CC||South Korea||359||80||439||Fractures||28.5||Violence|
|Safdar and Meechan||1995||CC||England||167||33||200||Patients||37.95||Violence|
|Yu et al.||2011||CC||South Korea||152||36||188||Fractures||20–29 (24%)||Traffic accident|
|Subhashraj||2009||CC||India||1617||416||2033||Patients||29.8 ± 9.6||Traffic accident|
|Cho et al.||2006||CC||South Korea||432||114||546||Patients||20–29 (44%)||Fall|
|Park et al.||2009||CC||South Korea||165||40||205||Fractures||20–29 (31%)||Fall|
|Tevepaugh and Dodson||1995||CC||USA||80||21||101||Patients||31.7 ± 9.2||Violence|
|Thangavelu et al.||2010||CC||India||345||115||460||Patients||31.2||Traffic accident|
|Ugboko et al.||2000||CC||Nigeria||369||121||490||Patients||30 ± 9||Traffic accident|
|Yadav et al.||2013||CS||India||NR||NR||289||Patients||18–45||Traffic accident|
|Yaltirik et al.||2002||CS||Turkey||84||16||100||Patients||26||Traffic accident|
|Yamada et al.||1998||CC||Japan||27||3||30||Patients||24.5 ± 9.9||Sport activities|
|Lee et al.||2012||CS||South Korea||NR||NR||86||Patients||NR||Violence|
|Author||With angle fracture||Without angle fracture|
|Third molar a||Winter classification||Pell and Gregory classification||Third molar a||Winter classification||Pell and Gregory classification|
|Abbasi et al.||116||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR|
|Choi et al.||43||18||NR||NR||NR||NR||87||56||7||54||81||15||149||251||NR||NR||NR||NR||212||107||35||162||136||56|
|Mah et al.||130||14||36||59||21||7||52||57||21||19||87||24||103||73||39||33||9||1||48||33||22||29||54||20|
|Duan and Zhang||152||45||NR||NR||NR||NR||87||90||9||47||99||40||218||285||NR||NR||NR||NR||569||258||85||478||223||211|
|Fuselier et al.||269||57||NR||NR||NR||NR||153||62||48||156||88||24||568||316||NR||NR||NR||NR||371||85||104||376||149||42|
|Gaddipati et al.||60||11||6||34||6||12||25||22||11||10||30||18||6||41||2||5||1||0||0||0||8||0||5||3|
|Halmos et al.||605||128||NR||NR||NR||NR||329||142||118||331||207||51||1364||803||NR||NR||NR||NR||900||199||262||900||361||100|
|Iida et al.||102||23||13||57||24||NR||28||56||10||10||56||28||89||134||33||117||62||NR||56||124||32||22||102||88|
|Iida et al.||67||11||8||34||2||3||13||31||5||12||25||12||94||46||24||62||12||5||27||67||16||11||71||28|
|Inaoka et al.||38||5||24||10||4||NR||24||9||5||20||17||1||NA||NA||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR|
|Kasamatsu et al.||48||5||1||17||3||3||NR||NR||NR||NR||NR||NR||63||35||0||9||4||0||NR||NR||NR||NR||NR||NR|
|Kumar et al.||21||3||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR|
|Revanth Kumar et al.||142||NR||2||10||1||7||13||6||1||1||19||0||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR|
|Lee and Dodson||79||20||NR||NR||NR||NR||53||18||8||46||21||12||170||98||NR||NR||NR||NR||120||19||26||107||29||29|
|Ma’aita and Alwrikat||127||25||19||16||11||17||29||58||40||43||51||33||299||164||1||14||2||18||181||110||8||202||89||8|
|Meisami et al.||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR|
|Naghipur et al.||163||84||41||140||13||3||93||75||29||75||90||32||70||129||98||213||27||4||179||112||51||145||143||54|
|Rahimi-Nedjat et al.||159||46||NR||NR||NR||NR||217||160||58||200||208||27||343||374||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR|
|Rajandram et al.||55||36||11||27||16||NR||NR||NR||NR||NR||NR||NR||31||152||3||11||17||NR||NR||NR||NR||NR||NR||NR|
|Rajkumar et al.||49||3||9||27||2||1||27||9||3||25||11||3||87||15||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR|
|Zhu et al.||136||34||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||38||108||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR|
|Safdar and Meechan||64||29||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||34||73||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR|
|Yu et al.||32||16||17||13||0||2||10||15||7||14||12||6||160||168||99||60||0||1||83||60||17||66||75||19|
|Cho et al.||228||40||80||90||40||18||68||118||42||112||78||38||176||102||84||44||34||14||98||64||14||106||44||26|
|Park et al.||67||17||19||39||7||2||27||33||7||20||41||6||188||138||73||94||11||10||110||57||21||99||78||11|
|Tevepaugh and Dodson||30||3||NR||NR||NR||NR||24||5||1||17||13||0||43||25||NR||NR||NR||NR||34||8||1||33||9||1|
|Thangavelu et al.||150||50||17||150||9||23||70||82||28||30||100||50||110||150||134||100||26||52||148||115||77||204||88||48|
|Ugboko et al.||65||11||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||343||71||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR|
|Yadav et al.||262||27||69||119||18||26||87||162||13||97||154||11||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR|
|Yaltirik et al.||51||NR||20||12||18||1||NR||NR||NR||NR||NR||NR||49||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR|
|Yamada et al.||45||0||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR||35||10||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR|
|Lee et al.||51||12||NR||NR||NR||NR||22||24||5||37||6||8||12||11||NR||NR||NR||NR||NR||NR||NR||NR||NR||NR|
The meta-analysis was conducted using R software, with ‘meta’ and ‘metafor’ packages (R Foundation for Statistical Computing). Statistical heterogeneity was tested by I 2 statistic. This statistic expresses the percentage of the variation across studies. Heterogeneity of I 2 < 25% was considered low, of I 2 = 50% was considered moderate, and of I 2 > 75% was considered high. When I 2 was equal to 0, a fixed-effects model was used; when I 2 was >0, a random-effects model was used. The dependent variable was fracture of the mandibular angle. Analyses were done for the following categorical variables: presence of mandibular angle fracture (‘yes’ or ‘no’) and presence of mandibular third molar (‘present’ or ‘absent’). When a third molar was present, the variables extracted were: (1) angulation of the third molar according to the Winter classification (V, MA, H, DA), which was assessed to determine the risk of mandibular angle fracture in each position; (2) position of the third molar according to the Pell and Gregory classification (A, B, C and I, II, III), which was assessed to determine the risk of mandibular angle fracture for each classification. Forest plots were generated for each meta-analysis. The forest plot is a graphical display of the estimated results of the individual studies included in the meta-analysis, along with the overall result. Publication bias was calculated using Egger’s test when at least 10 studies were to be included in the model. Funnel plots to assess the possibility of publication bias were analysed visually. The Funnel plot is a graph designed to evaluate the existence of publication bias. When asymmetry is observed in a funnel plot, publication bias may be present. A cumulative meta-analysis was performed to observe the variation in the odds ratio (OR) and confidence interval (CI) over time for the studies included.
A total of 704 articles were identified (Cochrane seven articles, PubMed 253 articles, Scopus 217 articles, VHL 33 articles, and Web of Science 194 articles). Of this total, 336 were duplicate articles. These were excluded, resulting in 368 articles for analysis. The titles and abstracts of the articles were read by two independent authors to determine whether the studies fulfilled the inclusion criteria for this review. At this stage, 322 articles were excluded for not fulfilling these inclusion criteria. Thus, the full texts of 46 articles were read. Eleven articles were excluded after full-text reading. Finally 35 articles were included in the review : 22 were case–control studies and 13 were cross-sectional studies . Twenty-eight of the 35 articles were included in the meta-analysis ( Fig. 1 ).
Description of the studies
All continents were represented by at least one paper, demonstrating the universality of this research. The articles evaluated the outcome (angle fracture and the presence of third molars) in three different ways. Some studies evaluated patients with mandibular angle fractures, not taking into consideration two fractures when there was a bilateral fracture of the mandibular angle . Some studies evaluated the angle fracture considering the presence of two other fractures when bilateral . Other studies evaluated both patients and fractures as the number of fractures . The highest average age was 37.95 years , while the lowest average age was 24.5 years . Traffic accident, fall, and assault were the most frequent causes ( Table 1 ).
Case–control studies showed a risk of bias score of 5 to 8 points out of a total of 9 on the NOS. Five studies scored 8 stars in the evaluation of quality , and four studies scored 7 stars . The majority of the studies ( n = 10) scored a total of 6 stars . Only three studies scored a total of 5 stars . In the evaluation of case–control studies, the domain least contemplated was ‘selection’ (definition of controls), followed by ‘comparability’ (see Table 3 ).
|Author||Selection (maximum 4 stars)||Comparability (maximum 2 stars)||Exposure (maximum 3 stars)||Total|
|Case definition||Representativeness of the cases||Selection of controls||Definition of controls||Comparability of cases and controls on the basis of the design or analysis||Ascertainment of exposure||Same method of ascertainment for cases and controls||Non-response rate|
|Fuselier et al.||★||0||★||0||★||★||★||★||6/9|
|Halmos et al.||★||★||★||0||★★||★||★||★||8/9|
|Iida et al.||★||0||★||0||★||★||★||★||6/9|
|Iida et al.||★||★||★||0||★||★||★||★||7/9|
|Kasamatsu et al.||★||0||★||0||★||★||★||★||6/9|
|Lee and Dodson||★||★||★||0||★★||★||★||★||8/9|
|Ma’aita and Alwrikat||★||★||★||★||★||★||★||★||8/9|
|Naghipur et al.||★||★||★||0||★★||★||★||★||8/9|
|Rahimi-Nedjat et al.||★||★||★||0||★||★||★||★||7/9|
|Rajandram et al.||★||★||★||★||0||★||★||★||7/9|
|Rajkumar et al.||★||0||★||0||0||★||★||★||5/9|
|Zhu et al.||★||★||★||0||0||★||★||★||6/9|
|Safdar and Meechan||★||0||★||★||0||★||★||★||6/9|
|Yu et al.||★||0||★||0||★||★||★||★||6/9|
|Cho et al.||★||0||★||0||★||★||★||★||6/9|
|Park et al.||★||0||★||0||★||★||★||★||6/9|
|Tevepaugh and Dodson||★||0||★||0||★★||★||★||★||7/9|
|Thangavelu et al.||★||0||★||0||★||★||★||★||6/9|
|Ugboko et al.||★||0||★||0||0||★||★||★||5/9|
|Yamada et al.||★||0||★||0||0||★||★||★||5/9|
a Selection: (1) (a) yes, with independent validation (1 star), (b) yes, e.g. record linkage or based on self-report, (c) no description; (2) (a) consecutive or obviously representative series of cases (1 star), (b) potential for selection biases or not stated; (3) (a) community controls (1 star), (b) hospital controls, (c) no description; (4) (a) no history of disease (endpoint) (1 star); (b) no description of source. Comparability: (1) (a) study controlled for the most important factor (position of third molar) (1 star), (b) study controlled for any additional factor (1 star). Exposure: (1) (a) secure records (e.g. surgical records) or (b) structured interview where blind to case/control status (1 star), (c) interview not blinded to case/control status, (d) written self-report or medical record only, (e) no description; (2) (a) yes (1 star), (b) no; (3) (a) same rate for both groups (1 star), (b) non-respondents described, (c) rate different and no designation.
Cross-sectional studies showed a risk of bias score ranging between 4 and 8 stars out of a total possible score of 9 stars. Among the 13 studies evaluated, only two obtained a score of 8 stars in the quality assessment ; four had a total of 7 stars , two had 6 stars , four had 5 stars , and one had 4 stars . In assessing the quality of cross-sectional studies, the domain least contemplated was ‘comparability’ (see Table 4 ).
|Author||Selection (maximum 4 stars)||Comparability (maximum 2 stars)||Outcome (maximum 3 stars)||Total|
|Representativeness of the sample||Sample size||Ascertainment of the exposure||The subjects in different outcome groups are comparable, based on the study design or analysis; confounding factors are controlled for||Assessment of the outcome||Statistical test|
|Abbasi et al.||★||0||★★||0||★★||0||5/9|
|Choi et al.||★||★||★★||★||★★||★||8/9|
|Mah et al.||★||★||★★||★||★★||★||8/9|
|Duan and Zhang||★||★||★★||0||★★||★||7/9|
|Gaddipati et al.||★||★||★★||0||★★||★||7/9|
|Inaoka et al.||★||★||★★||0||★★||★||7/9|
|Kumar et al.||★||★||★★||0||★★||0||6/9|
|Revanth Kumar et al.||0||0||★★||0||★★||0||4/9|
|Meisami et al.||★||★||★★||0||★★||0||6/9|
|Yadav et al.||0||0||★★||0||★★||★||5/9|
|Yaltirik et al.||0||0||★★||0||★★||★||5/9|
|Lee et al.||0||0||★★||0||★★||★||5/9|
a Selection: (1) (a) truly representative of the average in the target population or (b) somewhat representative of the average in the target population (1 star), (c) selected group of users, (d) no description; (2) (a) justified and satisfactory (1 star), (b) not justified; (3) (a) validated measurement tool (2 stars) or (b) non-validated measurement tool, but the tool is available or described (1 star), (c) no description of the measurement tool. Comparability: (1) (a) the study controlled for the most important factor (position of third molar) (1 star), (b) the study controlled for any additional factor (1 star). Outcome: (1) (a) independent blind assessment or (b) record linkage (2 stars), (c) no description; (2) (a) the statistical test used to analyze the data is clearly described and appropriate, and the measurement of the association is present (1 star), (b) the statistical test is not appropriate, not described, or incomplete.
Figure 2 shows the meta-analysis of the case–control studies that assessed the chance of angle fracture (experimental group) compared to other types of mandibular fracture without angle fracture (control group), in the presence of third molars. When the third molar was present, there was a greater chance of mandibular angle fracture than other mandibular fractures (OR 3.27, 95% CI 2.57–4.16, I 2 = 81.3%). Figure 3 shows the meta-analysis of the case–control and cross-sectional studies that assessed the presence or absence of a third molar in relation to mandibular angle fracture. When the analysis was the presence or absence of mandibular angle fracture in the presence or absence of the third molar, the presence of the third molar increased the chance of mandibular angle fracture (OR 3.83, 95% CI 3.02–4.85, I 2 = 83.1%).
Studies evaluating the characteristics of the third molar associated with a higher chance of mandibular angle fracture, showed that the presence of a third molar of Pell and Gregory class B in relation to the occlusal plane (events of experimental group) had a higher probability of having a mandibular angle fracture (OR 1.44, 95% CI 1.06–1.96, I 2 = 87.2%) compared with the other classes (A + C) ( Fig. 4 ).
A third molar of Pell and Gregory class A (events of experimental group) acted as a protective factor for mandibular angle fracture (OR 0.60, 95% CI 0.45–0.81, I 2 = 87.1%), when compared to the other classes (B + C) ( Fig. 5 ). No significant difference was found for class C third molars when compared to the other classes (A + B) (confidence interval crossing 1) (OR 1.19, 95% CI 0.57–2.46, I 2 = 96.1%).
For the Pell and Gregory classification assessing the position of the third molar in relation to the mandibular ramus, it was found that a class II third molar (events of experimental group) was more likely to be associated with mandibular angle fracture when compared to the other classes (I + III) (OR 1.67, 95% CI 1.36–2.04, I 2 = 72.4%) ( Fig. 6 ). A third molar of class I (events of experimental group) acted as a protective factor for mandibular angle fracture when compared to the other classes (II + III) (OR 0.51, 95% CI 0.37–0.71, I 2 = 89.4%) ( Fig. 7 ). No significant difference was found for class III third molars when compared to the other classes (I + II) (confidence interval crossing 1) (OR 1.51, 95% CI 0.94–2.42, I 2 = 90.3%).
The meta-analysis of third molar angulation according to the Winter classification did not show a significant difference between each position compared to the other positions: vertical (OR 0.69, 95% CI 0.45–1.06, I 2 = 77.7%), mesioangular (OR 1.42, 95% CI 0.96–2.10, I 2 = 81.7%), horizontal (OR 0.88, 95% CI 0.65–1.18, I 2 = 30.1%), and distoangular (OR 0.95, 95% CI 0.63–1.43, I 2 = 30.8%).
The cumulative meta-analysis showed that from 1995 (when the first article was published) to 2004 (when the first meta-analysis on the topic was undertaken), there was a variation in the odds ratio of 2.78 to 5.21 favouring the influence of the presence of the third molar in mandibular angle fracture. In addition, the confidence interval varied widely between those years with the inclusion of new studies. From the year 2005, the odds ratio varied from 3.10 to 3.38. Furthermore, the confidence intervals remained proportional to the results of the present meta-analysis (OR 3.30, 95% CI 2.59–4.20) ( Fig. 8 ).