Development and validation of a comprehensive clinical–radiographic difficulty scale for impacted mandibular third molar extractions: a prospective study

Abstract

Background

Most existing difficulty indices for impacted mandibular third molar (IMTM) surgery rely predominantly on radiographic criteria and often underrepresent clinical and systemic contributors to surgical difficulty. We aimed to develop and internally assess a multidimensional clinical–radiographic difficulty score and to examine its association with operation time as a pragmatic criterion measure.

Methods

This prospective observational study was conducted in a single oral and maxillofacial surgery unit using consecutive sampling. The difficulty instrument comprised systemic (3 binary items), radiographic (9 ordinal/binary items), and clinical (6 ordinal/binary items) components, summed to yield a total difficulty score. The primary outcome was operation time (minutes). Continuous variables are reported as mean ± SD and median (IQR). Operation time was compared across predefined difficulty groups using Kruskal–Wallis testing with Bonferroni-adjusted Mann–Whitney U post-hoc comparisons. Criterion-related validity was examined using Spearman correlations with bootstrap confidence intervals. Multivariable linear regression with heteroskedasticity-robust standard errors was used to adjust for age, sex, tooth side, and surgeon experience. Discrimination for prolonged surgery (≥ 75th percentile of operation time) was assessed with ROC analysis.

Results

A total of 215 cases were analyzed. Mean operation time was 13.08 ± 6.64 min, and mean total difficulty score was 16.16 ± 2.24. Operation time differed across difficulty categories (Kruskal–Wallis H = 14.29, p < 0.001). Total score showed a low but significant association with operation time (Spearman ρ = 0.266, p < 0.001). In multivariable regression, total score remained independently associated with operation time (β = 0.811 min per 1-point increase, p < 0.001), while surgeon experience was associated with shorter operation time (β=−0.450 min per year, p < 0.001). ROC analysis showed modest discrimination for prolonged surgery (AUC = 0.661).

Conclusions

In this single-center cohort, the proposed multidomain score showed preliminary internal evidence of criterion-related and known-groups validity and may support preoperative risk stratification and planning. External validation and objective reliability testing (inter-rater/test–retest) are warranted.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12903-026-07970-y.

Background

Teeth that could not be located at the occlusal level although the eruption time has come, and which have been partially or completely covered by hard or soft tissue are defined as impacted teeth. Problems that may be caused by impacted third molars constitute many prophylactic and therapeutic indications for the removal of these teeth. Third molars which remain impacted due to various reasons may remain within the maxilla and mandible for a lifetime without causing any complications. However, these teeth can cause odontogenic cysts and tumors, migration or root resorption of adjacent teeth, temporomandibular joint problems, and pain of unknown origin in the maxillofacial region. For these reasons, these teeth may have to be surgically removed [1–6].

American Association of Oral and Maxillofacial Surgeons (AAOMS) published a Position Paper in order to clarify the existing knowledge on the treatment of impacted third molars. Accordingly:

• Considering the best data based on evidence, third molars associated with a disease or at high risk of developing a disease have been recommended to be treated surgically.

• In the absence of disease or when there is no risk of a significant disease, active clinical and radiographic observation has been recommended [7].

Various scales have been developed to identify the level of difficulty of a surgical intervention of impacted third molars. It is reported in the literature that the first model about the difficulty of surgical removal of an impacted third molar was described by MacGregor in 1979 [8]. WHARFE classification developed by MacGregor was based on the image in the panoramic radiography, and accordingly, angulation of the impacted tooth according to Winter classification, angulation of the second molar tooth, root shape and development of the impacted tooth, follicle morphology and route of eruption are factors which are associated with prolonged operation time and difficulty level [9]. Another scale taken as a basis for identifying the level of difficulty is based on Pell-Gregory classification. According to this scale, it was aimed to find the level of difficulty based on the relation of impacted teeth with the mandibular ramus and occlusal plane. Pell-Gregory scale is not sufficient alone to accurately determine the level of difficulty [10]. On the other hand, Pederson scale is a modification of Pell-Gregory scale. Considering that only the relation of an impacted third molar with the mandibular ramus and occlusal plane may not show the level of difficulty of the surgical intervention, the position of the impacted third molar was also added to the scale [11, 12]. According to the modified Parant scale, the level of difficulty is determined postoperatively according to the type of surgical intervention [13, 14]. The studies conducted attempted to grade the surgical difficulty based on radiological variables [15, 16]. However, there are also clinical and systemic factors that affect surgical difficulty.

It is necessary to work according to many criteria and standards in the stages of development, adaptation and use of a new scale. Otherwise, the validity and reliability level of the scale would be low. Researchers may develop scales based on certain criteria. If a new scale is to be developed, the first step is to conduct a literature review regarding the issue. While doing this, it is important to consider which questions/topics should be evaluated for the subject of the scale. A format should be determined for the method of measurement, and a pool of items/questions should be created. This format may be in the form of arranging, classifying, equal interval and proportional scale according to mathematical properties of data, or based on certain special measurement methods [17]. The next step is to resort to an expert’s opinion for the pool of items that has been created. Experts evaluate the necessity of the questions as well as their clarity and specificity. They may thus suggest removing or modifying certain statements. However, accepting or rejecting these suggestions should be left to the discretion of the individual preparing the scale. Content validity ratio (CVR) is assessed with the expert’s opinion. The means of these ratios are found to determine the content validity index (CVI). The scale that is shaped according to the experts’ opinions is applied to the draft sample group. Trial application is important for the performance of the scale. The sample that is selected should represent the target group. After draft application, the scale is evaluated by using validity and reliability analyses and finalized accordingly [18].

The duration of the third molar removal operation may vary depending on factors such as the position of the tooth, its depth within the bone, presence of a systemic disease in the patient, and success of cooperation [19, 20]. Prolonged operation time may increase potential complications, and may also cause the complications to become more severe [21]. Previously, visual analog scale (VAS) was being used to assess the difficulty of surgical procedures. Currently, operation time is regarded and used as an adequate difficulty criterion since it is objective and clinically suitable. In addition, operation time is used for determining surgical difficulty since it also allows for comparisons [10, 22]. Some researchers have stated that it would be more effective to accept operation time alone as a criterion for determining the level of difficulty of impacted third molar surgeries [23, 24]. Therefore, operation time was used as an evaluation criterion for determining difficulty in the present study as well.

Existing difficulty indices for impacted mandibular third molar surgery tend to focus predominantly on radiographic parameters, such as tooth angulation, depth of impaction, and the relationship with adjacent anatomical structures. Although these indices provide useful information in clinical practice, the assessment of clinical and systemic factors that may influence surgical difficulty often remains limited, particularly for less experienced practitioners.

The scale proposed in the present study conceptually differs from previous indices by adopting a multidimensional clinical–radiographic approach that simultaneously incorporates systemic, clinical, and radiological components. While conventional indices have primarily been described as tools to support surgical planning among experienced surgeons, the proposed scale is intended to contribute to preoperative risk assessment, especially for general dentists working in primary care settings and surgically trained assistants. In this context, it may help reduce the risk of complications by supporting appropriate case selection and referral decisions.

Within the existing literature, there appears to be a limited number of scales that integrate systemic conditions, clinical findings, and radiographic features into a single composite score and relate this score to intraoperative indicators such as operation time. The present study is therefore expected to offer a comprehensive and clinically informative assessment tool that may support clinical decision-making, referral planning, and educational applications in third molar surgery.

Methods

Study design and ethics

This study was designed as a prospective, observational validation study conducted at the Department of Oral and Maxillofacial Surgery, Kırıkkale University. The study was carried out in accordance with the ethical principles outlined in the Declaration of Helsinki. Ethical approval was granted by the Regional Ethical Review Board of Kırıkkale University, Clinical Research Ethics Committee (Approval No: 08/11). Written informed consent was obtained from all participants before enrollment.

Study aim and setting

This prospective observational study aimed to develop and internally assess a multidimensional clinical–radiographic difficulty scale for impacted mandibular third molar (IMTM) extractions and to examine its association with operation duration as a pragmatic criterion measure. The study was conducted in a single oral and maxillofacial surgery unit using consecutive sampling of eligible cases.

A structured questionnaire was specifically developed for this study to evaluate clinical and radiographic factors associated with impacted mandibular third molar extractions. The English version of the questionnaire is provided in Supplementary File 1.

Sample size and study population

The sample size for this study was determined to ensure adequate statistical reliability of the results. The required sample size was calculated using a dedicated statistical power analysis software, GPower* (version 3.1, Universität Düsseldorf, Germany). Based on an a priori power analysis with a 95% confidence interval, an effect size of 0.3, a significance level (α) of 0.05, and a statistical power (1 − β) of 0.80, a minimum of 191 participants was determined to be required. To maintain the confidence level and to compensate for potential data loss, a total of 215 patients were included in the study.

Participants

Consecutive patients undergoing IMTM extraction were eligible if the preoperative clinical examination and radiographic assessment permitted completion of all scale items and if operation duration could be recorded. Cases with missing core scale items or missing operation time were excluded from analytic models (complete-case analysis).

Scale development and content validity

Candidate items were generated based on literature review and clinical consensus to capture systemic, radiographic, and clinical contributors to surgical difficulty, in line with standard scale/index development principles [25]. An expert panel (n = 9) reviewed all candidate items for necessity and relevance. Content validity ratio (CVR) was calculated for each item using Lawshe’s approach [26], and the scale-level content validity index (CVI) was summarized according to established recommendations for reporting CVI [27]. Items that did not meet the predefined content validity threshold or were judged to be infeasible or redundant were removed or revised. A pilot application was conducted to assess feasibility, clarity, and completion burden before final data collection [25].

Difficulty scale items and scoring

The final scoring system followed the Surgical Difficulty Evaluation Form (presented in the manuscript table). The instrument includes three components: systemic (3 items; binary; 0–3 points), radiographic (9 items; ordinal/binary; higher scores indicate greater difficulty), and clinical (6 items; ordinal/binary).

For binary items originally coded as 1 = present and 2 = absent, responses were recoded as 1 for presence and 0 otherwise; for the “no neighbouring tooth” item, a value of 1 was assigned when the adjacent tooth was absent and 0 when present. The total difficulty score was computed as the sum of systemic + radiographic + clinical component scores. For known-groups analyses, cases were categorized a priori as Easy (8–17 points), Moderately difficult (18–20 points), and Very difficult (21–30 points). The new scale is provided in Table 1.

Table 1.

Criteria of new scale and postoperative evaluation

Surgical difficulty evaluation form
Systemic evaluation	Score
Systemic Diseases	1
Diagnosed Psychological Disease	1
Psychological Drug Usage	1

Radiological evaluation
Depth
Level A: High Occlusal Level	1
Level B: Medium Occlusal Level	2
Level C: Deep Occlusal Level	3
Ramus Relationship/Space Available
Class 1: Sufficient Space	1
Class 2: Reduced Space	2
Class 3: No Space	3
Angulation
Mesioangular	1
Horizontal/transverse	2
Vertical	3
Distoangular	4
Buccolingual	5
Invert	6
Roots Form
Convergent	1
Divergent	1
Bulbous	1
Number of Roots
Single Root	1
Double Root	2
> 2	3
Relationship to Inferior Alveolar Canal
Apex of the Root 2 mm away from the alveolar canal	1
Apex of the root less than 2 mm from the root canal	2
Apex of the root in contact with mandibular canal without sign	3
Apex of the root in contact with mandibular canal with signs	4
No Neighboring Teeth	1
Total Edentulous	1
The Presence of Pathology Around the Tooth	1

Clinical evaluation
Age (years)
0–29	1
30–39	2
> 40	3
Maximum Interincisal Distance (mm)
30–39	3
40–49	2
> 50	1
Pericoronitis	1
Narrow Distance Between Lip Corners ( According to the perpendicular line from the pupils)	1
Degree of Eruption
Partial	1
Unerupted	2
Body Mass Index (BMI) > 25.0	1

Total score
Diffculty index
Easy	8–17
Moderately difficult	18–20
Very difficult	21–30

Statistical analysis

Data were screened for plausibility and distributional form. Continuous variables were summarized as mean ± SD and median (IQR), and categorical variables as counts (%). Operation time comparisons across difficulty categories were performed using the Kruskal–Wallis test [28]. Where overall differences were detected, post-hoc pairwise comparisons were conducted using the Mann–Whitney U test [29] with Bonferroni adjustment for multiple testing. Effect sizes were reported as epsilon-squared (ε²) for Kruskal–Wallis and rank-biserial correlation (r_rb) for pairwise comparisons [30, 31]. Convergent/criterion-related validity was assessed using Spearman rank correlations (ρ) [32]. Bootstrap resampling was used to estimate 95% confidence intervals (CIs) for correlation coefficients and for ROC AUC estimates [33]. To address potential confounding, multivariable linear regression was fitted with operation time as the dependent variable and total score, age, sex, tooth side, and surgeon experience as covariates. Heteroskedasticity-consistent (HC) robust standard errors were used (HC3), following recommendations for improved finite-sample performance [34, 35]. Discrimination for prolonged surgery was assessed with ROC analysis; AUC inference followed a nonparametric approach for correlated ROC curves [36]. The optimal cut-off was selected using Youden’s J statistic [37]. Internal consistency was explored with Cronbach’s alpha [38]. Because the instrument is multidimensional and conceptualized as an index with formative indicators rather than a unidimensional reflective scale, alpha values were interpreted descriptively and not as strict accept/reject criteria [39, 40].

Results

Study sample and score distributions

A total of 215 cases were analyzed (Table 1). Mean patient age was 24.08 ± 6.43 years (median 23.0; IQR 19.0–27.0). Female patients comprised 61.4% of the cohort. Mean surgeon experience was 3.11 ± 3.84 years (median 3.0; IQR 1.0–3.0). Mean operation time was 13.08 ± 6.64 min (median 13.0; IQR 8.0–17.0) (Table 2).

Table 2.

Sample characteristics and score distributions

Variable	Value
Cases (n)	215
Patient age (years)	24.08 ± 6.43 (median 23.0; IQR 19.0–27.0)
Female, n (%)	132 (61.4%)
Male, n (%)	83 (38.6%)
Tooth side: Right, n (%)	120 (55.8%)
Tooth side: Left, n (%)	95 (44.2%)
Surgeon experience (years)	3.11 ± 3.84 (median 3.0; IQR 1.0–3.0)
Operation time (min)	13.08 ± 6.64 (median 13.0; IQR 8.0–17.0)
Systemic component score	0.15 ± 0.39 (median 0.0; IQR 0.0–0.0)
Radiographic component score	10.81 ± 1.72 (median 11.0; IQR 10.0–12.0)
Clinical component score	5.20 ± 1.18 (median 5.0; IQR 4.5–6.0)
Total difficulty score	16.16 ± 2.24 (median 16.0; IQR 15.0–17.0)
Complication field completed, n (%)	5 (2.3%)
Complication present (among completed), n (%)	1 (20.0%)

Component score distributions are summarized in Table 1. The systemic component showed a pronounced floor effect (median 0.0; IQR 0.0–0.0), while radiographic and clinical component scores demonstrated broader dispersion. The mean total difficulty score was 16.16 ± 2.24 (median 16.0; IQR 15.0–17.0) (Table 2).

Difficulty categories and operation time (known-groups validity)

Based on predefined cut-offs, 166 (77.2%) cases were classified as Easy, 37 (17.2%) as Moderately difficult, and 12 (5.6%) as Very difficult (Table 3). Operation time differed significantly across difficulty categories (Kruskal–Wallis H = 14.29, df = 2, p < 0.001; ε² = 0.058) (Table 4).

Table 3.

Operation time by difficulty category

Difficulty group	n (%)	Operation time, mean ± SD	Operation time, median (IQR)
Easy	166 (77.2%)	12.13 ± 6.23	11.0 (7.0–16.0)
Moderately difficult	37 (17.2%)	16.41 ± 7.01	15.0 (13.0–20.0)
Very difficult	12 (5.6%)	16.00 ± 7.29	17.0 (10.8–20.0)

Table 4.

Known-groups validity: Kruskal–Wallis test

Test	H statistic	df	p	Effect size (epsilon-squared)
Kruskal–Wallis	14.290	2	< 0.001	0.058

Post-hoc pairwise tests indicated significantly longer operation time in the Moderately difficult group compared with the Easy group (Bonferroni-adjusted p = 0.0019; r_rb = 0.359) (Table 6). The Easy vs. Very difficult comparison did not remain significant after Bonferroni adjustment (adjusted p = 0.1439), and Moderately difficult vs. Very difficult showed no difference (adjusted p = 1.0000), likely reflecting the relatively small number of very difficult cases (Table 5).

Table 6.

Post-hoc pairwise comparisons (Mann–Whitney U; Bonferroni)

Group 1	Group 2	n1	n2	U	p	p (Bonferroni)	rank-biserial r
Easy	Moderately difficult	166	37	1969.0	< 0.001	0.0019	0.359
Easy	Very difficult	166	12	655.5	0.0480	0.1439	0.342
Moderately difficult	Very difficult	37	12	218.0	0.9349	1.0000	0.018

Table 5.

Internal consistency (Cronbach’s alpha)

Scale/Subscale	Cronbach’s alpha
Systemic (3 items)	0.149
Radiographic (9 items)	-0.033
Clinical (6 items)	0.145
Total (18 items)	0.130

Reliability (internal consistency)

Internal consistency estimates are presented in Table 3. Cronbach’s alpha values were 0.149 for the systemic component, − 0.033 for the radiographic component, and 0.145 for the clinical component; the alpha for the total scale was 0.130 (Table 6). Given the multidimensional/formative structure of the index, these alpha values were interpreted descriptively rather than as strict thresholds of reliability.

Correlations (convergent/criterion-related validity)

Correlation results are summarized in Table 5. The total difficulty score showed a low but statistically significant association with operation time (Spearman ρ = 0.266; 95% CI 0.145 to 0.381; p < 0.001). Radiographic and clinical component scores were also significantly correlated with operation time (ρ = 0.192 and ρ = 0.239, respectively) (Table 7). Surgeon experience was inversely correlated with operation time (ρ = −0.199; 95% CI − 0.333 to − 0.067; p = 0.003) (Table 4).

Table 7.

Spearman correlations (bootstrap 95% CI)

Variable 1	Variable 2	N	Spearman rho	95% CI	p
Total score	Operation time (min)	215	0.266	0.145 to 0.381	< 0.001
Radiographic score	Operation time (min)	215	0.192	0.068 to 0.314	0.005
Clinical score	Operation time (min)	215	0.239	0.108 to 0.367	< 0.001
Systemic score	Operation time (min)	215	-0.058	-0.179 to 0.074	0.398
Surgeon experience (years)	Operation time (min)	215	-0.199	-0.333 to -0.067	0.003
Patient age (years)	Operation time (min)	215	0.051	-0.093 to 0.182	0.455
Total score	Patient age (years)	215	0.210	0.072 to 0.345	0.002
Total score	Surgeon experience (years)	215	-0.024	-0.151 to 0.100	0.728

Multivariable regression

In multivariable regression (R² = 0.161; adjusted R² = 0.141), each 1-point increase in total difficulty score was independently associated with a 0.81-minute increase in operation time (β = 0.811; 95% CI 0.398 to 1.225; p < 0.001) after adjustment for age, sex, tooth side, and surgeon experience (Table 7). Surgeon experience remained independently associated with shorter operation time (β = −0.450; 95% CI − 0.614 to − 0.286; p < 0.001) (Table 8).

Table 8.

Multivariable linear regression (HC3 robust SE)

Predictor	Beta	95% CI	p
Intercept	-0.675	-6.418 to 5.068	0.818
Total difficulty score (per 1 point)	0.811	0.398 to 1.225	< 0.001
Patient age (per 1 year)	0.061	-0.136 to 0.258	0.542
Male (vs. female)	1.594	-0.222 to 3.410	0.085
Surgeon experience (per 1 year)	-0.450	-0.614 to -0.286	< 0.001
Left side (vs. right)	-0.111	-1.779 to 1.557	0.896

ROC analysis

Prolonged surgery was defined as operation time ≥ 17 min (75th percentile). The total score demonstrated modest discrimination for prolonged surgery (AUC = 0.661; bootstrap 95% CI 0.578 to 0.742) (Table 9). The optimal cut-off according to Youden’s J was total score ≥ 17, yielding sensitivity 59.7% and specificity 66.7% (Table 8).

Table 9.

ROC analysis for prolonged surgery

Metric	Value
Prolonged surgery definition	Operation time ≥ 17 min (75th percentile)
AUC (total score)	0.661
AUC 95% CI (bootstrap)	0.578 to 0.742
Optimal cut-off (Youden J)	Total score ≥ 17
Sensitivity at cut-off	59.7%
Specificity at cut-off	66.7%

Discussion

Previous studies have been insufficient especially in investigating the contribution of systemic and clinical conditions to the surgical difficulty of third molar surgery. Researchers usually focused on radiological difficulty level. There is no difficulty scale which would allow for simultaneously evaluating all factors that affect surgical difficulty. Therefore, a new scale was developed in the present study based on the assessments in other difficulty scales from the literature. In the present cohort, the systemic component demonstrated a pronounced floor effect (median 0.0; IQR 0.0–0.0), while radiographic and clinical components had broader dispersion (Table 2). This distribution likely reflects the young and largely medically healthy case-mix and has implications for how the systemic domain performs when operation time is used as the criterion outcome.

In the literature, some researchers aimed to evaluate surgical difficulty by using radiological data, and attempted to develop a new difficulty scale which they named as Kharma. As a result, they reported that the scale they developed to preoperatively predict the surgical difficulty of removal of an impacted third molar is not reliable in evaluating surgical difficulty, and that clinical data should also be considered in addition to radiological data [41]. The present study allowed for developing a new and more comprehensive surgical difficulty scale considering systemic, radiological and clinical data. Consistent with this multidomain rationale, both radiographic and clinical component scores were significantly associated with operation time (Table 6), supporting that difficulty is not solely radiographic in nature.

In a prospective cohort study, it was reported that in addition to anatomic and radiological factors for estimating surgical difficulty factors before removal of impacted third molars, certain clinical variables are also important factors for determining the surgical difficulty of removal of lower third molars, and that they should also be evaluated. The researchers reported that especially teeth with atypical root shape that need to be divided, teeth which are located close to the inferior alveolar canal, teeth which have knobby root anatomy, and teeth which are prone to excessive bleeding render the surgical procedure more complicated, and that information should be given to the patient preoperatively regarding the details of the surgical procedure [42]. This study is similar to the hypothesis of the present study. However, systemic situation assessment was not included. In our dataset, the systemic component was not significantly correlated with operation time (Table 7), which is plausibly explained by the limited variability in systemic items (Table 2). This suggests that the systemic domain may show greater utility in older or medically complex cohorts, where dispersion in systemic risk is expected to be higher.

According to a recent systematic review, on highly difficult or complex cases can be counted as the variables found to be effective on: such as elderly patient and overweight (patient variables), surgeons with little experience, and the use of complex surgical techniques need due to improper root form (operative variables) and radiological factors such as deep impaction, improper root form, and morphology, and close relationship with the second molar, maxillary sinus or inferior alveolar nerve canal (radiological variables) [43]. In the present scale, we created, all the variables listed above are included. In line with the literature emphasizing operator effects, surgeon experience was inversely correlated with operation time (Table 7), underscoring that operative duration is influenced not only by anatomic complexity but also by operator-related factors.

In a recent study, it was aimed to develop a new scale to assess the difficulty level of impacted third molar removal surgery in the Asian population. They indicated that degree of bone impaction had the greatest contribution and the number of roots had the least contribution to the level of difficulty. They reported that a limitation of the study was the absence of additional factors that may affect surgical difficulty such as maximum mouth opening and cheek flexibility [44]. Our scale allows for a more comprehensive assessment of surgical difficulty. The contribution of cheek flexibility and maximum mouth opening to surgical difficulty has been evaluated. Supporting this, the clinical component—which captures access-related variables—showed a significant correlation with operation time (Table 7).

In another study that aimed to develop a new scale for the purpose of assessing the difficulty of impacted third molar surgery, the researchers compared their own scale with the Pederson index. In addition to the radiological data in the Pederson index, they added certain clinical data including depth from point of elevation, mouth opening, tongue size, angulation of external oblique ridge, cheek flexibility, width of the root, and curvature of roots to their scale. The researchers have reported that clinical data are at least as important as radiological data in estimating surgical difficulty [45]. The above-mentioned clinical data are included in the present study as well. The present study has also evaluated the status of infection around the impacted third molar, presence of pathological development, presence of an adjacent tooth, the age of the patient, and the level of the eruption of the tooth in the mouth. In addition, total score increased modestly with age (Table 7), consistent with age being a patient-level contributor to difficulty reported in prior work.

Sanmaritoni et al. suggested a new difficulty scale concerning the removal of mandibular third molars and scored variables in a manner similar to our study. The researchers evaluated variables such as angulation, available space, depth, bone density, relation with the mandibular canal, bucco-lingual position and dental morphology in order to assess surgical difficulty. Unlike the present study, these researchers focused rather on radiological variables [46]. Our findings indicate that radiographic complexity contributes to longer operative time (Table 7); however, the overall association between total score and operation time was modest (Table 7), supporting the view that radiographs alone do not fully capture determinants of operative duration and perceived difficulty.

In the systematic review, as patients difficulty variables, the patient’s age and increased body mass ındex; as the operating variables, the surgeon has little experience and, the application of complex surgical techniques; as a radiological variable, deep impaction, unfavorable angulation and root morphology, and a close relationship to important anatomic regions were the factors affecting difficulty [43]. All-important variables were evaluated on the current surgical difficulty scale. When categorized using predefined cut-offs, operation time differed significantly across difficulty groups (Table 4), with the most robust separation observed between easy and moderately difficult cases after multiplicity correction (Table 5). This suggests that the tool may be especially useful for distinguishing “easy” versus “non-easy” cases within similar clinical settings.

A recent systematic review highlighted that existing difficulty indices rely predominantly on radiographic criteria, often neglecting important variables such as clinical parameters and the surgeon’s experience [47]. Although modern multivariable models and some artificial intelligence approaches have demonstrated the capacity to predict operative time, their generalizability and clinical applicability remain limited [48]. These findings underscore the ongoing need for a standardized and comprehensive difficulty scale. Therefore, there is a clear necessity to develop validated indices that incorporate both radiographic as well as clinical and systemic parameters. To address confounding statistically, we performed multivariable regression including patient age, sex, tooth side, and surgeon experience. After adjustment, total difficulty score remained independently associated with longer operation time (Table 8), while surgeon experience remained independently associated with shorter operation time (Table 8). This supports criterion-related validity beyond operator experience, while also confirming that operator-related factors substantially influence operative duration.

Some recently developed modern difficulty scales are limited to evaluating radiographic data provided by artificial intelligence–assisted computer programs. In this context, they remain restricted, as they are unable to perform clinical and systemic assessments. Moreover, there is an inherent risk of misdiagnosis when relying on AI [49]. In the future, the radiographic evaluation component of our study could be integrated into an AI-assisted system to contribute to the development of a difficulty scale. However, clinical and systemic assessments still need to be conducted by an experienced clinician. In our ROC analysis, total score demonstrated modest discrimination for prolonged surgery (AUC = 0.661; Table 9), suggesting that the tool is better suited for risk stratification and planning than for stand-alone prediction. The threshold total score ≥ 17 provided balanced sensitivity and specificity (~ 60–67%; Table 9), which may support scheduling and allocation decisions when combined with clinical judgment.

In another study, an AI-supported clinical tool was developed using YOLO11 models based on data from the Pederson index. The researchers emphasized that their program enabled effective radiographic assessment; however, they noted the lack of clinical evaluation [50]. In contrast, our study allows for comprehensive clinical, systemic, and radiographic assessments and could potentially be further enhanced through the use of AI-assisted computer programs.

In order to evaluate factors associated with surgical difficulty in removal of impacted mandibular third molars; clinical, systemic and radiological variables are important for predicting surgical difficulty. Based on the present single-center internal assessment, the proposed score shows statistically significant but modest associations with operation time (Table 7), meaningful separation between easy and moderately difficult categories (Tables 4 and 5), and independent association with operation time after adjustment (Table 8). Therefore, the score may be most appropriately positioned as a structured preoperative stratification and planning tool rather than a definitive predictor of time or complications without external validation.

Internal consistency estimates (Cronbach’s alpha) were low/negative across components and for the total score (Table 6). This should be interpreted in light of the instrument’s multidimensional/formative structure: the domains (systemic, clinical access, radiographic complexity) represent distinct contributors and are not expected to correlate strongly as in unidimensional psychometric scales. Consequently, Cronbach’s alpha is provided descriptively for transparency and should be complemented by inter-rater and test–retest reliability analyses (e.g., ICC/weighted kappa) in future studies with duplicate ratings.

This study has several limitations. All surgical procedures were performed by oral and maxillofacial surgeons—either in training or fully qualified—which may limit the generalizability of the results to general dental practitioners. In addition, as this was a single-center study, center-specific factors may have influenced the findings; therefore, multicenter validation is recommended. The ease of use and applicability of the scale have not yet been systematically tested. Future studies with larger and more diverse samples, including practitioners at various experience levels, are planned to assess usability and to adapt the scale for broader clinical application. Additionally, integration of this scale into artificial intelligence–based systems for radiological data analysis is planned for future research. Moreover, inter-rater and test–retest reliability could not be evaluated in the current dataset because duplicate scoring was unavailable, and operation time may be affected by unmeasured workflow variables (e.g., assistant variability, time of day, operator technique). Although we adjusted for key covariates including surgeon experience (Table 8), residual confounding may remain. Finally, complication recording was limited (Table 2), restricting complication-related inferences in this cohort.

Conclusions

In the present study, operation time was used as a pragmatic criterion measure for surgical difficulty in IMTM removal. Higher difficulty scores were associated with longer operation time, and predefined difficulty categories showed meaningful separation in operation duration, particularly between easy and moderately difficult cases. After adjustment for patient factors and surgeon experience, total difficulty score remained independently associated with operation time, while surgeon experience was associated with shorter operative duration. Overall, these findings suggest that the proposed score may support structured preoperative stratification and surgical planning in similar clinical settings; however, external validation, calibration of cut-offs, and objective reliability assessment (inter-rater/test–retest) are needed before broader generalization.

Authors’ contributions

All authors contributed to the study conception, design, data collection, analysis, and manuscript preparation. All authors read and approved the final version of the manuscript.

Funding

No financial support or funding was received for this study.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethical approval and consent to participate

This study was conducted in accordance with the Declaration of Helsinki. Ethical approval was granted by the Regional Ethical Review Board of Kırıkkale University, Clinical Research Ethics Committee (Approval No: 08/11). Written informed consent was obtained from all participants prior to their inclusion in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.