This study aims to establish a systematic approach for predicting RG in MLIs by combining demographic data with morphological characteristics derived from CBCT and intraoral photography. We utilized CBCT to diagnose RG, evaluated morphological traits via intraoral photographs, used Logistic regression to identify the most predictive variables for RG presence, and validated the predictive model with a separate cohort. Unlike previous studies focusing on individual morphological features, our model integrated these variables into a comprehensive nomogram, providing a quantifiable risk assessment tool for RG.
The critical anatomical predictors, Sh, TD, IGP, and IGMed, were carefully determined. The model applies these indicators based on characterization within the ASUDAS framework[6]. Sh involves the development of mesial and distal marginal ridges on the lingual surface. TD is expressed as ridges of varying size and number, usually around the cingulum. IG represents depressions or grooves that interrupt the course of the marginal ridges or basal cingulum. Other characteristics described for MLIs with RG, such as a prominent, pointed tubercle, pronounced cingulum, distal/mesial marginal ridge, and deep fold in the cingulum [1, 4, 10, 15, 24, 25, 27], could not be evaluated using objective criteria. Fortunately, Sh and TD provided a reference scale to assess these characteristics relative to RG. Notably, the nomogram suggests that the visibility of the interruption groove above the gingiva in photographs is significantly associated with the presence of RG, reinforcing the role of crown anatomy in predicting RG. The combined predictors could relatively precisely indicate the existence of RG in the nomogram.
Before deploying the algorithm to build the nomogram, we conducted thorough correlation tests for each anatomical variable. This was done to mitigate overfitting and enhance interpretability. (Fig. 2)
During our tests, we observed a slight but noticeable relationship (r = 0.69) between the presence of Interruption Groove (IGP) and its Median (IGMed). This indicates that visible IGs are usually positioned at the center of the lingual aspect of MLIs, consistent with some literature [27] but differ from previous literature regarding their deviation from the distal aspect [1, 25]. We also found a correlation (r = 0.55) between Interruption Groove Mesial-Distal presence (IGMD) and TD, indicating that a prominent lingual TD is often accompanied by dual grooves around the cingulum above the gingiva. Furthermore, we identified a potential link between normal occlusion (ONorm) and deep overbite occlusion (ODO), with ODO being more prevalent in our sample. The sample also showed a strong correlation (r = 0.79) between size and contour, which aligns with a previous study [28]. After analyzing the anatomical traits, we hypothesized that the crown traits of MLI would reveal the RG in the root. Therefore, we conducted this study. It is crucial to interpret correlations carefully to avoid overfitting. Further validation in diverse populations is necessary to confirm the nomogram's universality.
This study's predictive nomogram for RG in MLIs excluded parameters like Incisal Notching (IN), diminished Size (S), and peg-shaped Contour (C) morphology despite their noted significance in prior studies [4, 10, 24]. The data revealed a negligible correlation between IN and RG, casting doubt on IN as a dependable indicator for RG. Moreover, our analysis found no substantial link between occlusion and the presence of RG, suggesting these two factors may not be interdependent.
Previous research has documented MLIs with RG, noting features such as prominent tubercles, exaggerated cingulums, and pronounced marginal ridges [1, 10, 15, 24, 25, 29]. However, evaluating these characteristics with consistency posed challenges in our study. We adopted the ASUDAS standards to minimize observer bias, utilizing Sh, IG, DS, and TD for assessment [6]. ASUDAS's scoring guide aided in standardizing evaluations, but manual classification still carries the risk of inconsistency. Future studies could employ artificial intelligence to refine our assessment and increase analytical precision, particularly in enhancing image recognition and deploying machine learning algorithms.
When diagnosing RG using CBCT, we followed the existing literature to locate the CEJ accurately [5, 23]. This allowed us to identify the TSN. RG was recognized as a V-shaped defect observed on the radicular side of the TSN, persisting for more than 2mm or approximately six CBCT slices. Defects on the coronal side or those less than 2mm in length were not considered RG. The absence of established diagnostic criteria for RG concerning CEJ and TSN may result in differing interpretations. However, we have carefully defined our methodology to reduce observational variances and align with accepted definitions of RG [24], thereby improving the consistency of our study's findings.
We maintained the accuracy of our results through a strict methodology. However, we must acknowledge potential sample bias, as we excluded cases with complex restorations or prior root canal therapy, which could share an unstudied link with RG. The selected demographic, mainly pre-orthodontic patients averaging 19 years of age, may raise questions about our results' broader relevance; our cohort's unique RG occurrence rates may skew the nomogram's parameters, affecting its more general application. Confirmatory studies across varied patient demographics are crucial to validate our nomogram, ensuring its predictive reliability and utility in various clinical settings.
This study examined the relationship between crown anatomical characteristics and the occurrence of RG, highlighting the critical role of clinical examination. The nomogram we developed offers a pioneering method, revealing potential correlations between evident crown features and RG in MLIs. This predictive tool could increase diagnostic precision in endodontic and periodontic treatments, improve treatment planning, and reduce nonessential radiographic exposure.