Comparison of CBCT and panoramic radiography for the assessment of bone loss and root resorption on the second molar associated with third molar impaction: a systematic review

Abstract

Objective:

To investigate whether the use of cone beam CT (CBCT) changes the diagnosis of external root resorption (ERR) or marginal bone loss (MBL) involving a second molar adjacent to an impacted third molar.

Methods:

A systematic search was applied in PubMed, EMBASE, Scopus, Web of Science, LILACS, Google Scholar, OpenGrey, and ProQuest. Studies assessing the detection of ERR or MBL in a second molar adjacent to an impacted third molar through CBCT and panoramic radiography (PAN) were included. Prevalence and agreement between PAN and CBCT on the detection of ERR and MBL were collected. The risk of bias was assessed using the MAStARI.

Results:

A total of 593 papers were identified, and after a 2-phase selection, 5 studies were included in the narrative synthesis. Regarding ERR, its prevalence in PAN was reported from 5.31 to 19.5% and from 22.8 to 62.0% in CBCT. The percentage of agreement varied from 28.5 to 74.0%. The prevalence of MBL varied from 21.9 to 62.9% in PAN, while those values varied from 21.6 to 80% in CBCT images. The percentage of agreement between PAN and CBCT for the detection of MBL ranged from 66.0 to 85.0%. Four studies presented low risk of bias and one had moderate risk.

Conclusions:

More ERR and MBL are assessed in CBCT compared to PAN. There is a considerable agreement between PAN and CBCT assessment of ERR and MBL, however, mostly related to the absence of the pathology rather its presence.

Introduction

A tooth is considered impacted when it remains inside the jaw’s bone beyond the normal time for eruption¹ due to malposition, lack of space, or a physical obstacle in its eruption course.² As third molars are the most common impacted teeth,³ their removal is very frequent surgical procedure in dentistry.⁴

The impaction of a third molar can be associated with some pathological conditions,² such as external root resorption (ERR) in the second molar, marginal bone loss (MBL) of the distal surface of the second molar, and cystic lesion related to the third molar.⁵ The sensible point is that the pathological conditions associated to impacted third molars might evolve without clinical signs and symptoms.^4,6

Despite the cause–effect relationship is not completely clear, it seems that the mechanical force of an impacted tooth triggers ERR in an adjacent one in addition to the inflammation of the periodontium around this tooth.⁷ Moreover, it seems logical that a misplaced and non-erupted tooth may predispose the MBL, that is the destruction of bone.⁴ Depending on the severity of these conditions, however, the patient may be asymptomatic⁸; thus ERR and MBL of the second molars are frequently and solely detected on radiographic examinations that precedes removal of third molars.⁵

Radiographic images provide information about the third molar itself, the surrounding bone, the adjacent tooth and anatomical structures.⁹ Panoramic radiography (PAN) is the first-choice method for assessment of third molars and for treatment planning.^2,4,9 This imaging modality, however, offers a two-dimensional image, which involves several potential sources of misinterpretation, such as image magnification, distortion by projection errors, blurred images, and complex maxillofacial structures projected onto a two-dimensional plane.¹⁰ On the other hand, cone beam CT (CBCT) offers the possibility of sectioning the volume in all three planes without distortion to detect small details.⁴ Although the possibility of adjusting exposure parameters and field of view to optimize the radiation dose according to the diagnostic requirements,² CBCT units still provide higher radiation doses to the patient in comparison to PAN.⁹ For this reason, the discussion on when performing CBCT prior to surgical removal of third molars remains a hot-topic. The European Academy of DentoMaxilloFacial Radiology (EADMFR) has published a position paper which stated that CBCT-imaging of the mandibular third molar should only be requested when the surgeon has a specific clinical question that cannot be answered by PAN and/or intraoral imaging.¹¹ However, second molar pathologies - ERR and MBL - were not assessed or mentioned as one of these questions.

When ERR and MBL in the second molars are early and correctly diagnosed, the consequences of these pathological conditions are reduced or ceased. In some cases, such as the presence of a severe ERR in the second molar, removal of the second molar should be considered instead of the third molar.⁵ There is no clinical protocol, however, guiding to remove the second instead of the third molar.⁵ It is worth mentioning that the removal a third molar itself as a prophylactic procedure is still a very debatable topic.³

Despite the importance of correct diagnosis of ERR and MBL in order to avoid clinical complications, the scope of these conditions is not established yet. This systematic review (SR) aimed to assess whether the use of CBCT changes the diagnosis of ERR and MBL involving a second molar adjacent to an impacted third molar.

Methods

The protocol of this SR was developed following recommendations of PRISMA and the study protocol was registered at the Prospective Register of Systematic Reviews (PROSPERO), available under registration number CRD42020136588.¹²

This SR was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Checklist (PRISMA)¹³ and Synthesis Without Meta-analysis (SWiM) reporting items.¹⁴

Information sources

Initially, the following review question was used to establish a search strategy: Is the use of CBCT determinant for the evaluation of ERR or MBL involving a second molar adjacent to an impacted third molar?

An appropriate search strategy was developed for each of the following electronic database with the support of a heath science librarian. Strings of search terms were constructed, consisting of relevant text words and Boolean links: PubMed: (“Root Resorption”[MeSH] OR “Resorption” OR “Resorptions” OR “alveolar bone loss”[MeSH Terms] OR “alveolar bone loss” OR “Alveolar Bone Losses” OR “Alveolar Process Atrophy” OR “Alveolar Process Atrophies” OR “Periodontal Bone Losses” OR “Periodontal Bone Loss” OR “Alveolar Bone Atrophy”) AND (“Radiography, Panoramic”[MeSH] OR “Panoramic Radiography” OR “Panoramic Radiographies” OR “panoramic image” OR “panoramic images” OR “panoramic imaging” OR “panoramic radiograph” OR “panoramic radiographs” OR “panoramic X-ray” OR “panoramic” OR “Pantomography” OR “Pantomographies” OR “Orthopantomography” OR “Orthopantomographies”) AND (“cone-beam computed tomography”[MeSH Terms] OR “CBCT” OR “Cone-Beam” OR “conebeam”). The full search strategy developed for each of the electronic database is detailed in Supplementary Figure 1.

The literature search was done within the PubMed, EMBASE, Scopus, Web of Science and Latin American and Caribbean Health Sciences (LILACS), until January 24, 2021. Grey literature was searched on Google Scholar, OpenGrey and ProQuest Dissertations & Theses Global.

All references were managed by Mendeley software (Mendeley Ltd., London, UK) and duplicate hits removed. There were no restrictions regarding language or publication period.

Eligibility criteria

Studies assessing the detection and prevalence of ERR or MBL of second molar adjacent to an impacted third molar through PAN and CBCT were included. Reviews, case reports, letters and personal opinions were excluded.

Study selection and data collection

A two-phase process was adopted to study selection. First, two reviewers (LMS and MAM) read titles and abstracts and after the full texts of the selected ones. The first and second authors (LMS and MAM) performed the study selection and data collection independently. A third reviewer (ACO) and a fourth reviewer (LBO) were consulted in the event of a disagreement between the first and second authors.

For data extraction, at first two reviewers independently collected the data and compared it afterwards. A third reviewer, an experienced radiologist, confirmed the extracted data (ACO).

The following information was collected: author, year, country, sample size, outcomes, associated parameters, imaging acquisition parameters, statistical analysis, prevalence of ERR and MBL, and agreement between PAN and CBCT, other findings and authors’ main conclusions.

Risk of bias in individual studies

The methodological quality of the included studies was assessed using the risk-of-bias tool Meta-Analysis of Statistical Assessment and Review Instrument (MAStARI) from the Joanna Briggs Institute (Institute Joanna Briggs Institute Reviewers’ Manual).¹⁵ The evaluation was estimated by three authors (LMS, HGA and ACO) regarding the proportion of “yes” items as follows: high (up to 49% score “yes”), moderate (50–69% score “yes”), and low (>70% score “yes”). In addition, the online tool robvis (Risk-of-bias VISualization) was used to generate figures.¹⁶

Synthesis of the results

Heterogeneity within studies was assessed based on study characteristics, methodological and outcome characteristics. All the included studies presented the data collected; however, they were not grouped together in meta-analysis due to the methodological disparities.

Results

Study selection

The search identified 593 records and after removing duplicates, 486 remained, being 12 selected to full-text analysis. In total, five studies were included in the narrative synthesis.^2,8,17–19 Further information about the seven excluded studies is available in Supplementary Figure 2. The studies’ selection steps are presented in a detailed flowchart diagram in Figure 1.

Figure 1.

Flow diagram of literature search and selection criteria.

Study characteristics

Studies were conducted from 2014² to 2020^18,19 (Table 1). All studies retrospectively retrieved patients presenting both imaging modalities from imaging data banks. Two studies evaluated the prevalence and agreement of ERR through PAN and CBCT images,^2,8 one study evaluated the agreement of ERR through PAN and CBCT images,¹⁹ one study evaluated the prevalence and agreement of MBL through PAN and CBCT images,¹⁸ one study evaluated the prevalence and agreement of both ERR and MBL through the same imaging exams.¹⁷

Table 1.

Summary of descriptive characteristics of included articles that compared PAN and CBCT for detection of pathology on second molar associated with third molar impaction: ERR and MBL (n = 5)

Author, year, country	Sample size	Outcomes	Associated parameters	Imaging acquisition parameters	Statistical Analysis	Main findings (Prevalence and agreement between PAN and CBCT)	Other Findings	Main conclusions
Oenning et al, 2014 Brazil2	188 third molars (91 maxillary and 97 mandibular)	ERR in second molars	1.Inclination of third molar (Winter classification) vs ERR	PAN: OP100 D (Instrumentarium) operating at 66kVp, 2.5mA, exposure time of 17.6 s CBCT: Classic i-CAT (Imaging Sciences International) operating at 120kVp, 6mA, 0.25 mm voxel size, 13 cm FOV	X2 test Fisher exact test 2-proportion Z test	Prevalence of ERR: PAN 5.31%, CBCT 22.88% Total agreement: 151 cases (80.3%) ERR presence: 8 (4.3%) ERR absence: 143 (76%)	1.Mandibular third molars in mesioangular and horizontal inclinations were more likely to cause ERR on second molars.	A significantly greater number of cases of ERR was detected using CBCT. When direct contact between the second molar and an unerupted mandibular third molar is observed on PAN, especially in mesioangular or horizontal third molars, CBCT should be used to provide better evaluation.
D’Costa et al, 2017 India8	120 third molars (39 maxillary and 81 mandibular)	ERR in second molars	Location and severity of ERR Grading of ERR (Nemcovsky criteria) Position of third molar (Pell & Gregory ABC classification)	Not mentioned	Chi-squared test McNemar test	Prevalence of ERR: PAN 13.3%, CBCT 52.5% Total agreement: 71 cases (59%) – ERR presence: 15 (12.5%); ERR absence: 56 (46.5%)	1.Location of ERR: Cervical >Middle>Apical Severity: Slight >Moderate 2.Grade: gradeA >gradeB>gradeC 3.Position B is more prevalent. Sensitivity 23.8% and Specificity 98.2% (CBCT as a gold standard and PAN as a parameter).	ERR was “better” detected with CBCT. If on PAN a close contact is detected between the impacted third molar and adjacent second molar, CBCT can be advised considering the “risk vs reward ratio.”
Matzen et al, 2017 Denmark17	379 mandibular third molars	[td]	Frequency of removal based on pathological findings Interobserver reproducibility Factors in PAN that could predict ERR and MBL in CBCT (e.g. location of overprojection, third molar inclination)	PAN: Cranex Tome (Soredex) or ProMax (Planmeca) CBCT: NewTom 3G (QR SRL) operating at 0.3 mm voxel size, 6 × 8 cm FOV or Scanora 3D (Soredex) operating at 0.13 mm voxel size, 6 × 6 cm FOV	Descriptive statistics: Percentage of agreement, frequency of removal (PAN, CBCT vs pathology) Interobserver reproducibility: percentage of accordance and κ Logistic regression	Prevalence of ERR: PAN 16–19.5%, CBCT 22.4–62% Prevalence of MBL: PAN 21.9–55.6%, CBCT 21.6–74% Agreement PAN/CBCT: ERR 54–74%, MBL 66–85% (varied among four observers)	1.Frequency of removal decision based on pathologies: PAN 47%, CBCT 65% percentage of accordance: PAN 28.1–99.7%, CBCT 61–97.4%. κ: poor to excellent Mesioangulated / horizontally positioned third molars overprojecting the cervical part of the second molar in PAN are strongly associated with pathology observed in CBCT.	ERR and MBL in the second molars are more often observed in CBCT than in PAN. More third molars would be removed if pathological findings are based on CBCT.
Dias et al., 2020 Uruguay18	124 mandibular third molars	MBL at the second molars	Third molar angulation Severity of the MBL /age	PAN: OP200 (Instrumentarium) operating at 66kVp, 8mA, exposure time of 14 s CBCT: i-CAT (Imaging Sciences International) operating at 120kVp, 36mA, 0.25 mm voxel size, 16 × 13 cm FOV	McNemar test, κ statistics, Bowker test, Fisher’s exact test	Prevalence of MBL: PAN 62.9%, CBCT 80% Agreement PAN – CBCT: MBL 74.2%	Mesioangular and horizontal third molars had a significant association with MBL of the adjacent second molars PAN underestimated the severity of MBL compared to CBCT / the mean age of patients with severe MBL was higher	Diagnosing second molar MBL associated to impacted third molar in PAN may be challenging because of false negatives. Impacted third molars justify preoperative CBCT scans if second molar MBL needs to be assessed.
Mendonça et al., 2020 Brazil19	221 mandibular third molars	ERR in second molars and contact area between second and third molars	Comparison between PAN and CBCT evaluation Changes in diagnosis after CBCT evaluation and the impact in the surgical planning and post-operative complication expectation	PAN: PaxX-400C (Vatech) CBCT: i-CAT FLX (Imaging Sciences International) operating at 120kVp, 36.12mAs, 0.25 mm voxel size	Percentage agreement, McNemar test, logistic regression analysis	Prevalence of ERR: PAN 10.00%, CBCT 22.40% Agreement PAN – CBCT: ERR 28.5%	The diagnostic change regarding second molar ERR after CBCT evaluation was not capable of altering the surgical planning EER in the second molar and contact area with the third molar assessed by CBCT seems to decrease expectation of post-operative complications	Diagnostic changes of the third molar relationship with adjacent structures do not change the decision to remove the tooth, but may alter the intraoperative procedures

CBCT, cone beam CT; ERR, external root resorption; MBL, marginal bone loss; PAN, panoramic radiography.

Risk of bias in the studies

Four studies were considered to have a low risk of bias and one had a moderate risk of bias (Figure 2A and B). In three studies, the risk of bias was judged low for all domains. It is important to emphasize that only one study¹⁹ obtained score “not applicable”, which occurred in two domains (D5 and D6).

Figure 2.

RoB summary of author’s judgments for each included study, evaluated by the MAStARI tool and graphically represented by “Traffic-light” plot (A) and Weighted bar plot (B) (generated using the online tool robvis (Risk-Of-Bias VISualization).

Results of individual studies

The included studies evaluated from 120⁸ to 379¹⁷ teeth. Two studies evaluated maxillary and mandibular third molars^2,8 and three studies evaluated mandibular third molars only.^17–19 From the five included studies, two reported prevalence and agreement in the detection of ERR between PAN and CBCT images^2,8 and one reported only the agreement in the detection of ERR between these imaging exams.¹⁹ From these five studies, one reported prevalence and agreement for the detection of MBL in PAN and CBCT images¹⁸ and one reported prevalence and agreement for the detection of both ERR and MBL through these PAN and CBCT images.¹⁷ One study also reported sensitivity and specificity of PAN for ERR detection considering CBCT as reference standard.⁸

The prevalence of ERR for maxillary and mandibular second molars varied from 5.3%² to 13.3%⁸ in PAN, while those values varied from 22.9%² to 52.5%¹⁷ in CBCT images. Only one study² disclosed the prevalence of ERR in maxillary second molars, which was 1% in PAN and 14.3% in CBCT. Three studies assessed the mandibular second molars separately, and the reported prevalence varied from 9.3%² to 19.5%¹⁷ in PAN, and from 22.4%^17,19 to 62%¹⁷ in CBCT (Table 2). Data on prevalence of one study¹⁹ was not available and was retrieved directly with the authors. The percentage of agreement among observers between PAN and CBCT for the detection of ERR was between 28.5%¹⁹ and 74.0%¹⁷ (Table 1).

Table 2.

Comparison between PAN and CBCT images in the detection of external root resorption in second molars (2M)

	Maxillary and mandibular 2M (%)		Maxillary 2M (%)		Mandibular 2M (%)
	PAN	CBCT	PAN	CBCT	PAN	CBCT
Oenning et al., 20142	5.3	22.9	1	14.3	9.3	31
Matzen et al., 201717	-	-	-	-	16–19.5	22.4–62
D’Costa et al., 20178	13.3	52.5	-	-	-	-
Mendonça et al., 202019	-	-	-	-	10	22.4

CBCT, cone beam CT; PAN, panoramic radiography.

Missing values marked with “-“ mean that authors did not assess this variable.

Regarding the MBL, the included studies assessed its prevalence only in mandibular second molars which in PAN was reported between 21.9%¹⁷ and 62.9%.¹⁸ In CBCT images, the prevalence of MBL of mandibular second molars varied from 21.6%¹⁷ to 80%¹⁸ (Table 3). The percentage agreement between these imaging modalities varied from 66.0 to 85.0%¹⁷ (Table 1).

Table 3.

Comparison between PAN and CBCT images in the detection of marginal bone loss of second molars (2M)

	Maxillary and mandibular 2M (%)		Maxillary 2M (%)		Mandibular 2M (%)
	PAN	CBCT	PAN	CBCT	PAN	CBCT
Matzen et al., 201717	-	-	-	-	21.9–55.6	21.6–74
Dias et al., 202018	-	-	-	-	62.9	80

CBCT, cone beam CT; PAN, panoramic radiography.

Missing values marked with “-“ mean that authors did not assess this variable.

Table 4 shows that three studies found a relation between ERR and MBL of mandibular second molars and the mesioangular and horizontal angulation of mandibular third molars.^2,17,18

Table 4.

Relation of ERR and MBL of second molars and inclination of third molars (3M) in CBCT images

	ERR		MBL
	Angulation of 3M		Angulation of 3M
	Maxilla	Mandible	Maxilla	Mandible
Oenning et al., 20142	-	Mesioangular / horizontal	-	-
Matzen et al., 201717	-	Mesioangular / horizontal	-	Mesioangular / horizontal
D’Costa et al., 20178	-	-	-	-
Dias et al., 202018			-	Mesioangular / horizontal
Mendonça et al., 202019	-	-	-	-

ERR, external root resorption; MBL, marginal bone loss.

Missing values marked with “-“ mean that authors did not assess this variable.

Discussion

Analyzing potential pathological consequences on the second molar is an important step of the third molar imaging assessment. The detection of ERR and MBL involving the second molar might have a significant impact in the treatment decision.¹⁹ In this context, the results of this SR have shown that ERR in second molars adjacent to an impacted third molar was reported to be from 5.3 to 19.5% in PAN while on CBCT images the prevalence of ERR was reported between 22.8 and 62%.^2,8,17 A clear increase in ERR prevalence (i.e. detection) is observed in CBCT-based evaluation of the relationship of second and third molars. This tendency of more ERR detection on CBCT images have been previously discussed,¹⁷ indicating that the three-dimensional analysis, without magnification or distortion, allow this better detectability. In addition to that, the possibility to verify the suspicion of root resorption in more than two planes, without superimposition of adjacent structures, plays a significant role in greater detection for CBCT.^2,17 On the contrary, PAN assessment is restricted to a two-dimensional image prone to magnification, distortion, and third molar image projection over the second molar, precluding the detection of root surface contour alterations related to the ERR.²

Regarding MBL prevalence, two included studies showed these data. Matzen et al¹⁷ reported a prevalence of MBL in PAN images from 21.9 to 55.6%, while on CBCT this prevalence was from 21.6 to 74%. Dias et al¹⁸ reported a 62.9% MBL prevalence in PAN and 80% in CBCT images. In contrast to ERR, for MBL the difference in prevalence between PAN and CBCT may be considered just moderately higher for the last. The advantages of three-dimensional modality over two-dimensional imaging, however, also applies here for the higher MBL detection in CBCT images.^17,18 For example, in PAN image the superimposition of the buccal and lingual cortical plates may smooth or cover the MBL which is clearly observed in the sectional CBCT assessment.

The sample of all included studies was randomly selected, which is important when considering the prevalence of those conditions. Only one study has reported the prevalence of ERR in maxillary second molars, that was markedly lower in PAN (1%) compared to CBCT (14.3%).² Two studies found an association between mandibular third molar angulation (i.e. mesioangular or horizontal) and the detection of ERR, but none for maxillary molars.^2,17 That same association was reported for MBL in mandibular molars^17,18; none of the studies reported the prevalence of MBL in maxillary second molars. A previous study²⁰ assessed MBL and ERR in maxillary second molars comparing PAN and CBCT images. In this study, only the patients who presented a suspicion of pathologies in PAN (MBL, ERR, or increased follicular space) were scanned with CBCT. As we understood that it could change the agreement scenario and the prevalence data could not be extracted, this study was not included in the SR. Nevertheless, the authors report similar detection rates of second molar pathology in both imaging modalities, although PAN may underestimate ERR.

Overall, the studies included in the present SR presented low risk of bias, with exception of one study. The seventh domain regarding the outcome measure was considered positive for all studies. A previous systematic review on CBCT accuracy for ERR²¹ indicates CBCT as a reliable method for ERR detection, although the included studies were in vitro with simulated ERR lesions. In addition, the indication of further CBCT investigation in cases when ERR is suspected in PAN may change the treatment planning.⁵ On the other hand, Schröder et al²² assessed CBCT accuracy for natural ERR detection from 2.4 to 3.1 mm³ and concluded that CBCT may not be as accurate as previous in vitro studies have claimed. For periodontal bone loss evaluation, CBCT is a well-established method.²³ Nevertheless, data on prevalence of ERR and MBL are reported considering the observers in each study and may diverge if different observers do the same evaluations, mainly for ERR. Thus, it is expected a moderate-to-high observer variation on the imaging-based detection of discrete pathologies, due to the inherent subjectivity of the task, observers’ experience, and observers’ formation background.

In this systematic review, we also aimed at verifying the agreement between PAN and CBCT regarding the detection of ERR and MBL involving second molars adjacent to impacted third molars. The agreement between imaging modalities for ERR was reported between 28.5 and 80.3%,^2,8,17,19 while for MBL it was between 66 and 85%.^17,18 Overall agreement shows slightly higher percentage for MBL. As stated by Dias et al,¹⁸ this fair agreement for MBL could not be the same when dealing with more detailed diagnostic tasks such as ERR.

Although the studies showed a relatively high agreement between PAN and CBCT for ERR detection, this is mostly related to the number of second molars classified with absence of ERR. For example, in the study of Oenning et al,² the agreement for presence of ERR occurred in 4.3%, while in 76% for the absence of ERR. Matzen et al¹⁷ found agreement for MBL presence of 24% while the agreement was 51.6% for MBL absence, considering all four observers together. Conversely, Dias et al¹⁸ found higher agreement for MBL presence (58.9%) compared to MBL absence (15.3%). Such controversial results between those studies may be related to the methodological issues regarding image analysis: while one study invited four oral radiologists to independently assess the images,¹⁷ the latter invited two oral radiologists to assess the images in consensus.¹⁸

Apart from prevalence and agreement between PAN and CBCT for the detection of ERR and MBL, the included studies also revealed additional outcomes. ERR lesions are usually located in the cervical portion of the root of second molars, and, considering CBCT as reference standard, PAN images have low sensitivity (0.238) and high specificity (0.982) for ERR diagnosis.⁸ Oenning et al² and Dias et al¹⁸ concluded that mandibular third molars in mesioangular and horizontal inclinations have a greater potential to induce ERR and MBL on the adjacent second molars, respectively, which corroborated the findings of Matzen et al.¹⁷ In addition, PAN tends to underestimate MBL severity compared to CBCT.¹⁸ The frequency of removal decision of the third molars are increased when assessing CBCT images,¹⁷ although the diagnostic change specifically in ERR presence caused by the CBCT evaluation is not capable of altering the surgical planning.¹⁹

The diagnostic efficacy of imaging exams can be classified according to the hierarchical model proposed by Fryback and Thornbury.²⁴ In this 6-level model, the first level concerns technical aspects, the second level is directed to diagnostic accuracy, sensitivity and specificity of the images, the third level addresses the capability of the images to change one’s diagnostic thinking, the fourth level regards the treatment plan change, the fifth level measures the effect on patient outcome, and the sixth and last level analyzes societal benefit of the imaging exam. In this sense, the studies included in the present systematic review may partially cover levels 1 to 3 of this model.

Limitations

CBCT assessment showed a higher prevalence of both ERR and MBL compared to PAN images. However, due to the retrospective nature of the included studies, none followed-up the patients or used a gold-standard to confirm the presence of such pathologies, and thus, the diagnostic values (accuracy, sensitivity, and specificity) could not be determined for CBCT and PAN in ERR and MBL diagnosis. Although this could be considered a limitation, the applicability of a gold-standard in clinical studies assessing ERR and MBL may be restricted for ethical reasons (e.g. the extraction of the second molar for histology analysis to confirm ERR). In addition, CBCT has proven its value in ERR detection, and this imaging modality has capability to diagnose clinically relevant ERR in the second molars.

Matzen et al¹⁷ discussed such issue, stating that CBCT could not be considered as the gold-standard. This does not invalidate the studies, but when analyzing the results, one must bear in mind that prevalence of ERR and MBL in PAN and CBCT images does not mean the diagnostic accuracy of these imaging modalities for such tasks. Thus, the prevalence reported is imaging modality and observer-dependent. As none of the studies assessed observer experience, future studies should be directed to elucidate the relationship between the imaging modalities (PAN and CBCT) and observer experience (juniors vs seniors, or different specialists) in the prevalence of ERR and MBL of second molars adjacent to impacted third molars.

CBCT images may vary significantly depending on CBCT machine, technical specifications, and acquisition parameters; thus, it has been recommended to not just extrapolate results from one CBCT unit to another.²⁵ For the detection of fine small, detailed structures (or pathologies), high-resolution scans are recommended¹⁰ which are, not only but also, related to the voxel size of the scanning. The voxel size used for scanning in the include studied varied from 0.13 to 0.3 mm, which may be adequate for ERR detection.²⁶ From a clinical perspective, however, superficial (very small) resorptions, which would demand high resolution CBCT protocols, in general, do not impact the management of an impacted third molar. In other words, the treatment planning is potentially changeable when the resorptive lesion compromises the dentine. The same reasoning can be applied for bone loss extension, with the difference that the clinical impact is more related to the prognosis definition than the treatment plan. On this matter, we understand that standard protocols and restricted fields of view can represent the optimal strategies to balance dose and image quality for this diagnostic task, given that higher resolution scans usually require higher radiation doses delivered to the patient.

Conclusion

It seems reasonable to affirm that a two-dimensional exam is always the first choice to assess third molars. However, CBCT may be indicated if there is suspicion of ERR in the adjacent tooth, as there is a significant increase in the detection of this condition in CBCT. Regarding MBL, although CBCT also provides greater detection, it is not too expressive. Therefore, after PAN examination, whether ERR is suspected in a second molar adjacent to a mesioangular or horizontal third molar, it is recommended to acquire a CBCT for better evaluation.

Moreover, there is a considerable agreement between PAN and CBCT on the assessment of ERR and MBL, though those are usually related to the agreement in the absence of the pathology rather its presence. Still, studies aiming at the higher levels of diagnostic efficacy (e.g. diagnostic thinking and treatment planning) of CBCT in comparison to PAN related to the detection of ERR and MBL of second molar adjacent to impacted third molar must be carried out to further elucidate this research topic.