Impact of cone beam CT on diagnosis of external cervical resorption: the severity of resorption assessed in periapical radiographs and cone beam CT. A prospective clinical study

Abstract

Objectives:

To compare the severity of external cervical resorption (ECR) observed in periapical (PA) images and cone beam CT (CBCT) using the Heithersay classification system and pulp involvement; and to assess inter- and intraobserver reproducibility for three observers.

Methods:

CBCT examination was performed in 245 teeth (in 190 patients, mean age 40 years, range 12–82) with ECR diagnosed in PA images. Three observers scored the severity of ECR using the Heithersay classification system (severity class 1–4) and pulp involvement (yes/no) in both PA images and CBCT. Percentage concordance and κ-statistics described observer variation in PA images and CBCT for both inter- and intraobserver reproducibility.

Results:

For all three observers, the ECR score was the same in the two modalities in more than half of cases (average 59%; obs1: 54%, obs2: 63%, obs3: 61%). However, in 38% (obs1: 44%, obs2: 33%, obs3: 36%) of the cases, the observers scored more severe ECR in CBCT than in PA images (p < 0.001). The ECR score changed to a less severe score in CBCT only in 3% (obs1: 1%, obs2: 4%, obs3: 4%). For pulp involvement, 14% (obs1: 7%, obs2: 20%, obs3: 15%) of the cases changed from “no” in PA images to “yes” in CBCT. In general, κ values were higher for CBCT than for PA images for both the Heithersay classification score and pulp involvement.

Conclusions:

ECR was generally scored as more severe in CBCT than PA images using the Heithersay classification and also more cases had pulp involvement in CBCT.

Introduction

External cervical resorption (ECR) is characterized by damage of dental hard tissues due to pathologic odontoclastic activity.¹ ECR is often asymptomatic without visible clinical appearance and is mostly detected in radiographic images by coincidence. The radiographic appearance can vary depending on the localization and extent of the resorption. The entry point is located below the gingival epithelial attachment,² which is often in the cervical area, hence the name. ECR appears as radiolucent, irregular striae that may extend into the crown and/or the root of the tooth. In case of severe destruction, ECR may eventually perforate to the periodontal ligament along the vertical extent of the resorptive lesion and thereby more surface lesions are present. ECR does not penetrate the pulp until late in the process because the layer of pre-dentine and odontoblasts protect the pulp.³ This layer is called the pericanalar resorption-resistant sheet (PRRS).² The PRRS is visualized as a radiopaque line parallel to the pulp, which is very characteristic for ECR. ECR also spreads horizontally around the pulp of the tooth.

In 1999, Heithersay developed a classification system to categorize ECR.⁴ The system classifies ECR into four categories with increasing disease severity (class 1 to 4). Class 1 describes a small resorption in the cervical area of the tooth with only superficial perforation of the dentine. Class 2 is a more invasive lesion, closer to the coronal pulp and with the possibility of slight penetration into the radicular dentin. Class 3 denotes an even larger resorption penetrating the radicular dentin in up to one-third of the root (coronal part). Class 4 is the most severe degree with a dominating resorption in both the coronal and radicular part of the tooth, most likely penetrating the pulp. The classification system is based on periapical (PA) radiographic images supported by histopathologic features.⁵ With the development of cone beam CT (CBCT), it became possible to assess ECR in all dimensions and add information on the extent of the lesion as well as the location, size and number of surface entries. This extra information is suspected to influence the treatment decision.⁶ Recently, other classification systems have been suggested based on CBCT.^6–8 These systems all focus on the advantages of three-dimensional (3D) information. The additional information from CBCT included in recent classification systems is the circumferential spread of ECR around the pulp⁷ and the height of the ECR in addition to its relation to the pulp.⁸ Recently, it has been suggested to include the size and location of the entry point in a new classification system⁶; however, this system is based on very few teeth (n = 23), which is also the case for a study by Goodell et al. (n = 30).⁷ The classification system developed by Patel et al⁸ has been tested in a recent study from 2020 on 14 teeth.⁹ In this study, it was concluded that there was a weak level of agreement (Cohen’s κ 0.40) for the “Patel classification” compared with a moderate level for the Heithersay classification (Cohen’s κ 0.69). A review from 2018 concluded that original basic scientific research on the subject of ECR is lacking.¹⁰ The review showed that the greater part of papers were case reports and case series. There is a deficiency of prospective studies with a high number of teeth and a standardized radiographic protocol, reflecting the clinical situation. This calls for a study to test the diagnostic efficacy of ECR in PA images compared with CBCT before introducing new classification systems.

The discussion on when to perform a CBCT of ECR is ongoing, and due to the lack of high evidence studies, no strict guidelines exist. The European guidelines on when to perform a CBCT regarding ECR were updated in 2018,¹¹ adding to The European Society of Endodontology position statement from 2014.¹² Both focused on using CBCT for potentially treatable ECR lesions only. The prognosis of ECR depends on early diagnosis, i.e. before ECR has become severe.¹³ The farther the destruction of the tooth, the more challenging the treatment will be, if possible at all. It may be speculated if the 3D nature of CBCT will provide the dentist with more information and thus allow him or her to establish a more precise severity diagnosis and treatment plan.

The aim of this study was to compare the severity of external cervical resorption observed in periapical images and CBCT using the Heithersay classification system and pulp involvement; and to assess inter- and intraobserver reproducibility for three observers using both radiographic modalities.

Methods and materials

Radiographic examination

This prospective clinical study included a total of 190 patients (91 females and 99 males) with a mean age of 40 years (ranging from 12 to 82) who had a total of 245 teeth with ECR as assessed from PA images. The distribution of tooth type is shown in Figure 1. 169 patients (89%) contributed with one tooth, 9 patients with two teeth, 5 patients with three teeth, and 7 patients with more than three teeth. During 2014–2019, the patients were referred to Section for Oral Radiology, Department of Dentistry and Oral Health, Aarhus University, Denmark either from general dental practice (GP), community dental healthcare clinics or the student clinics at the Department of Dentistry and Oral Health. The criterion for project inclusion was a diagnosis of ECR in an intraoral image as assessed by the referring dentist and confirmed by a radiologist. Only teeth without endodontic treatment were included. At the Section for Oral Radiology, all patients had two radiographic examinations performed: the first examination of the tooth consisted of two intraoral periapical projections: an orthogonal in addition with an eccentric exposure (Digora PSP system, Soredex, Finland). The second examination was CBCT (Cranex3D (197 cases), Soredex, Finland; 5 × 5 cm field of view (FOV), 0.085 mm voxel resolution or Scanora 3D (48 cases), Soredex, Finland; 6 × 6 FOV and 0.13 mm voxel resolution). Both examinations were preferably performed on the same day. FOV was centered at the tooth in question. Approval by an ethical committee was not required for the present study, because it is considered a quality control of already implemented radiographic examinations. The patients gave their informed consent, which was noted by their general dentist in the patient’s record, to have the radiographic examinations performed at the dental school.

Figure 1.

Distribution of tooth type.

Radiographic assessment

Three trained observers (one dentist with specialist radiology experience, one dentist with specialist endodontic experience and one dentist in radiology training) individually scored the PA images and afterwards the CBCT volumes. The observers had access to the PA images during the CBCT assessment like in the clinical situation. The three observers were calibrated, discussing 25 cases (not included in the study) before initiating the individual registrations. Radiographic assessment was performed on high-quality monitors 22’’ HD [Samsung, Tianjin, China (TSEC)] in a room with dimmed lighting. The PA images of a particular tooth were assessed simultaneously on the monitor. For CBCT, the full volume was assessed using OnDemand software (CyberMed; Daejeon, Korea Inc. Irvine, CA). The observers had the possibility to use enhancement filters and change the windowing parameters.

The observers assessed ECR in the PA images according to the Heithersay classification system (class 1, 2, 3, or 4); and further registered whether or not the ECR had involved the pulp observed as no distinct PRRS present (yes/no). After assessing the PA images, they assessed the CBCT full volumes according to the same variables. After 18 months, the observers rescored 10% of randomly selected images from both modalities to assess intraobserver reproducibility.

Data treatment and statistical analysis

Data was registered in Excel (Microsoft Office 2010, Microsoft Corp., Redmond, WA) and imported to Stata 15.1 (StataCorp, College Station, TX; release 15) for statistical analysis.

Descriptive statistics was used to determine the relationship between the Heithersay classification score and pulp involvement observed in PA images and CBCT for each observer. To test whether the scores obtained by the two radiologic methods differed, Stuart-Maxwell analyses (marginal homogeneity) and Bowker’s test were used.

Inter- and intraobserver reproducibility were estimated by two outcome variables: percentage of observer accordance and weighted κ analysis for the two imaging modalities separately. κ values exceeding 0.75 were defined as “excellent” reproducibility, those between 0.40 and 0.75 as “fair to good” reproducibility and those lower than 0.40 as “moderate to poor” reproducibility.¹⁴

Results

Comparison between Heithersay classification score in PA images and CBCT

Table 1 shows the Heithersay scores in PA images and CBCT. For all three observers, the assessments were the same in the two modalities in the majority of the cases. However, for each observer a more severe score was recorded in CBCT than in PA images in a significant number of cases (p < 0.001). Observer 1 recorded the same score in 54% (133 cases), a more severe score in 44% (109 cases), and a less severe score in 1% (3 cases). Observer 2 recorded the same score in 63% (154 cases), a more severe score in 33% (82 cases), and a less severe score in 4% (9 cases). Observer 3 recorded the same score in 61% (149 cases), a more severe score in 36% (87 cases), and a less severe score in 4% (9 cases). Figure 2 shows a case in which the Heithersay classification score changed from the PA images to the CBCT.

Table 1.

Relationship between Heithersay classes in PA images and CBCT (1–4 = increasing severity) scored by the three observers

OBS1	PA class 1	PA class 2	PA class 3	PA class 4	Total
CBCT class 1	4	0	0	0	4
CBCT class 2	3	11	2	0	16
CBCT class 3	4	21	78	1	104
CBCT class 4	2	3	76	40	121
Total	13	35	156	41	245
OBS2	PA class 1	PA class 2	PA class 3	PA class 4	Total
CBCT class 1	7	0	0	0	7
CBCT class 2	4	15	6	0	25
CBCT class 3	2	13	51	3	69
CBCT class 4	1	2	60	81	144
Total	14	30	117	84	245
OBS3	PA class 1	PA class 2	PA class 3	PA class 4	Total
CBCT 0	1	1	0	0	2
CBCT class 1	1	3	1	1	6
CBCT class 2	2	20	2	0	24
CBCT class 3	0	20	74	0	94
CBCT class 4	0	1	64	54	119
Total	4	45	141	55	245

CBCT, cone beam CT; PA, periapical.

Figure 2.

A case with ECR in tooth 21, in which Heithersay class 3 was scored based on PA images (a, b) by two observers, while one observer scored class 4. (c, d) show CBCT sections in the coronal and sagittal plane. Based on CBCT, two observers changed to score 4, which was in accordance with the last observer. All observers scored “yes” for pulp involvement in both PA images and CBCT. CBCT, cone beam CT; ECR, external cervical resorption; PA, periapical.

Comparison between score of pulp involvement in PA images and CBCT

As seen in Table 2, observer 1 recorded identical scores for pulp involvement in 88% (216 cases) in the two modalities. In 7% (18 cases), observer 1 changed score for pulp involvement from “no” (PA images) to “yes” (CBCT), whereas the score was changed from “yes” to “no” in 4% (11 cases). For observer 2, agreement was established in 77% (189 cases). Observer two changed score from “no” (PA images) to ”yes” (CBCT) in 20% (50 cases), and from “yes” to “no” in 2% (6 cases). For observer 3, agreement was established in 82% (202 cases). Observer 3 changed score from “no” (PA images) to ”yes” (CBCT) in 15% (37 cases) and from “yes” to “no” in 2% (6 cases). For observers 2 and 3, the change to the more severe diagnosis, i.e. pulp involvement in CBCT, was statistically significant (p < 0.001). For observer 1, no statistical difference was seen between the two modalities (p < 0.19).

Table 2.

Relationship between score “involvement of the pulp” in PA images and CBCT scored by the three observers

Obs1	PA No	PA Yes
CBCT No	29	11
CBCT Yes	18	187
Obs2	PA No	PA Yes
CBCT No	61	6
CBCT Yes	50	128
Obs3	PA No	PA Yes
CBCT No	29	6
CBCT Yes	37	173

CBCT, cone beam CT; PA, periapical.

Interobserver reproducibility

The overall interobserver reproducibility was fair to good when classifying ECR in both PA images and CBCT (Table 3). κ values were, however, overall higher for CBCT. In general, the score “involvement of the pulp” also had a higher agreement among the observers when assessed in CBCT than in PA images.

Table 3.

Interobserver agreement in percent and weighted κ value for Heithersay class and pulp involvement (non-weighted κ value)

	OBS1 vs OBS2	OBS1 vs OBS3	OBS2 vs OBS3
Heithersay class
PA images	94.7% 0.62	95.8% 0.62	94.4% 0.57
CBCT	96.9% 0.75	98.2% 0.74	98.2% 0.78
Pulp involvement
PA images	68.0% 0.32	79.2% 0.42	72.7% 0.43
CBCT	85.7% 0.59	88.2% 0.54	84.5% 0.54

CBCT, cone beam CT; PA, periapical.

Intraobserver reproducibility

Especially, observer 2 obtained a consistently excellent intraobserver reproducibility as seen in Table 4. For the other two observers, it was fair to good. κ values were higher for CBCT than PA images for all observers, except for observer 1 when scoring Heithersay class in PA images, in which the same κ value was obtained as in CBCT; and for observer 3 in the same category, where PA images obtained a higher κ value than CBCT.

Table 4.

Intraobserver agreement in percent and weighted κ value for Heithersay class and pulp involvement (non-weighted κ value)

	OBS1	OBS2	OBS3
Heithersay class
PA images	96.4% 0.65	97.3% 0.73	98.2% 0.82
CBCT	94.0% 0.65	98.0% 0.92	94.7% 0.51
Pulp involvement
PA images	88.0% 0.50	92.0% 0.83	76.0% 0.50
CBCT	96.0% 0.65	96.0% 0.90	92.0% 0.63

CBCT, cone beam CT; PA, periapical.

Discussion

In this prospective cohort study, patient cases were collected over a period of 6 years. The patients had a tentative diagnosis of ECR observed in PA images by their usual dentist and were referred for radiologic examination including both new PA images and CBCT. The PA images recorded by the patient’s own dentist were of different quality and date, and it was therefore decided to acquire standardized PA images for the study. Also, the PA images and CBCT examination were performed on the same day (with very few exceptions) for optimal comparison of the two modalities. We believe that the selected patients to a high degree reflected the distribution of diagnosed ECR in the population, since they were referred from various sectors of dental practice and from an area mirroring the average demographic constellation of the Danish population. However, the sample did not have an even distribution of the four severity classes in the Heithersay classification system, since Heithersay class 1 and 2 were underrepresented in the referred patients, which is to be expected because of the difficulty in diagnosing small lesions.¹⁵ The study is seen as a level three study in the Fryback and Thornbury hierarchical model for evidence-based guidelines and efficacy assessments,¹⁶ since it evaluated whether an additional radiographic examination (CBCT) influenced the diagnose determined from a traditional imaging method (2D).

In the present study, two CBCT units were used, Scanora 3D and Cranex 3D units. CBCV units may vary in voxel resolution, FOV, etc. However, in a previous study comparing these two units focusing on reproducibility of assessment of mandibular third molars, no differences were found between observer concordance and interobserver κ agreement for these units.¹⁷ In the present study, the same CBCT units were used. CBCT was therefore treated as one modality in the analyses.

In the present study, use of the Heithersay classification yielded more severe ECR scores in CBCT than in PA images in on average 38% of the cases. In less than 3% of the cases, the Heithersay class changed to a less severe score in CBCT. The change to a more severe score was statistically significant. The severity of the lesion therefore seems to be underestimated in PA images, which may not fully display the difficulties of potential treatment. CBCT may therefore be a valuable diagnostic tool in the treatment planning process, since the extent of ECR has been shown to affect the prognosis of treatment.¹³ Only a small number of clinical studies have included a reasonable number of teeth in the comparison of diagnostic efficacy between PA images and CBCT.^3,18 The same tendency as observed in the present study appeared in a study where 115 teeth with ECR were included.³ Six examiners assessed Heithersay class 3 and 4 in 64.3% of the cases in PA images compared with 91.9% of the cases in CBCT. The scoring in CBCT was performed by a consensus panel, which was treated as a reference standard. It is unfortunate that observer variation was not disclosed in this study³ since it is obvious that CBCT is also a diagnostic method with inherent observer variation, and therefore it cannot be seen as a reference standard for ECR or other similar diagnostic tasks. Another study compared the diagnostic efficacy in PA images and CBCT in 13 teeth ex vivo with fabricated resorption-like defects that may not resemble the characteristics of this complex disease entity.¹⁹ Moreover, images (both PA and CBCT) were presented in a PowerPoint presentation, i.e. the CBCT sections to be evaluated were prefabricated, and the full volumes were not assessed by the observers. This setting differs immensely from the clinical situation.¹⁹ An ex vivo study comparing both fabricated root resorptions (n = 42), natural external root resorptions (n = 42), and control teeth (n = 42) concluded that the fabricated lesions were easier detected in both PA images and CBCT.¹⁸ For the natural external root resorptions in this study, CBCT imaging had higher sensitivity and specificity for detecting root resorptions than PA images as compared to micro-CT imaging that served as reference standard. An in vivo study also comparing PA images and CBCT concluded that CBCT was superior in assessing ECR.²⁰ This study included only five teeth with ECR though, and also used prefixed CBCT image sections displayed in a PowerPoint presentation although access to the raw data was available. In general, failure to assess the full CBCT data set involves a risk of selection bias, and this set-up does not resemble the clinical situation, in which general practitioners are to make decisions. Clinical studies with only 5 and 13 teeth moreover do not provide a high power to determine whether to use PA images or CBCT. In the present study, a high number of teeth was included (245), and it was designed to mirror a normal clinical situation with both PA images and a full CBCT volume diagnosed by the clinician. Furthermore, both PA images and CBCT were scored by the same three observers in the present study contrary to a consensus decision for CBCT. Having the same three observers assessing both radiographic modalities moreover makes an inter- and intraobserver reproducibility assessment possible.

As for pulp involvement, a severe resorptive state was observed in CBCT more often than in PA images. For observers 2 and 3, the change to a more severe score was statistically significant. Pulp involvement may be a parameter of most interest, when establishing the prognosis of a tooth. In a recently suggested classification system based on 3D imaging,⁸ the pulpal relation was scored, i.e. whether the lesion was confined to dentin or had pulp involvement. In another study testing this system on 14 teeth, it was stated that scoring the proximity to the root canal in CBCT was difficult.⁹ Involvement of the pulp raises the need for endodontic treatment along with removing the resorbing tissue. This may be an important factor to consider when deciding on a treatment plan for ECR.

Even though the Heithersay classification system is simpler than the suggested CBCT classifications systems,^6–8 it remains complex and can invite different interpretations, which is also reflected in the results of this study. Other classification systems include more parameters, like entry point and circumferential spread. These extra parameters contribute to more detailed information on the ECR,^6–8 however to include many parameters also makes it more difficult for observers to reproduce the score, and it is more time consuming. Previously published classification systems have been tested only on a few number of teeth with ECR.^6–9 ECR is a complex disease, both to diagnose and to treat.^10,21 To benefit from additional information suggested in other classifications systems, thorough testing is required before these systems should be included in clinical routines.

Even though the interobserver reproducibility was higher for CBCT than for PA images, some variation among observers was noted. κ values ranged from 0.74 to 0.78 in assessment of Heithersay class in CBCT, and even though a value exceeding 0.75 is defined as excellent, scores in CBCT are still not 100% reproducible. CBCT contributes with valuable information of the disease, however, it does not satisfy the criteria for a reference standard method.²² Former studies have speculated that this method will represent the “true extent” of the resorption.^{1,3,8,10,19–21,23} However, the true extent of the lesion can be established only through tooth histology, and no study has compared CBCT severity assessment of ECR to histology, for instance. In a previous study, 27 teeth were examined with PA images and CBCT and then extracted and further analyzed using nanofocus-CT imaging, hard tissue histology and scanning electron microscopy.² It was concluded that ECR is a dynamic and complex condition, but no conclusion was made that hard tissue histology and CBCT provided similar results. Another study, which also comprised a multimodular combination of investigation techniques on a single tooth (clinical examination, PA images, CBCT, nano-CT, microfocus CT and hard tissue histology) concluded that a combined approach was an ideal setup in future studies, but still no equality between CBCT and hard tissue histology was displayed.²⁴ For intraobserver reproducibility, especially observer two had excellent κ values in almost every assessment. For observer 1, the κ values were either the same or higher in CBCT than in PA images. Observer 3, on the other hand, obtained a higher reproducibility in the Heithersay classification score in PA images than in CBCT. Observer 2 was the most experienced observer in handling and interpreting CBCT volumes, which can explain the different outcomes.

The past years have seen a stronger focus on ECR.^11,15 It can be speculated whether this resorptive condition therefore appears with a higher prevalence or whether dentists have become more aware of the disease since recently introduced 3D radiographic modalities may have facilitated establishing the diagnosis. Existing guidelines suggest that CBCT examination is indicated in the treatment planning of potentially treatable ECR lesions only.^11,12 The literature on treatment decision and prognosis of treatment in relation to ECR lesions is sparse though.²⁵ In a follow-up clinical study with 101 teeth, Heithersay showed that ECR lesions of class 1 and 2 severity obtained complete treatment success after 5 years.¹³ For class 3 lesions, the overall success rate was 77.8% and for class 4 lesions 12.5% after 5 years. The different classes in the Heithersay system may thus predict the treatment prognosis. Correct classification is hence important to inform the patient about treatment options and prognosis. However, because of its complexity, treatment of ECR appears to be a specialist job. Heithersay thus has a specialist background in endodontics and may be considered more competent than the average clinician for this type of treatment. Another study with 14 teeth focused on the surgical management of ECR and concluded that all teeth with Heithersay class 1–3 survived after 20 months and with a clinical success rate of 79% (survival defined as tooth in situ with no subjective symptoms, whereas clinical success equaled no pathology).⁹ No class 4 teeth were included in this study. Future studies will assess treatment outcome of ECR lesions of various severity.

Conclusions

On average, more than one-third of ECR lesions were scored as more severe using the Heithersay classification system in CBCT volumes compared with periapical images recorded with an orthogonal and eccentric projection. Furthermore, involvement of the pulp also changed from PA images to CBCT from “no” involvement to “yes” in 14%. Inter- and intraobserver reproducibility was in general better for CBCT than PA images.