Association between impacted third molars and external root resorption on adjacent second molars based on cone-beam computed tomography: a systematic review and meta-analysis Authors: Wenxuan Liu, Ailin Liang, Yanling Zhou, Chenxi Liao, Xiaoyan Zhou, Yihua Huang, Qimei Gong Published: 2026 ABSTRACT AbstractBackgroundExternal root resorption (ERR) on second molars (M2s) is an undesirable type of pathological damage associated with adjacent impacted third molars (iM3s). Owing to its limited detectability on conventional radiographs, ERR may remain undiagnosed until irreversible damage has occurred. This systematic review aimed to quantitatively evaluate the association between iM3s and ERR on M2s via cone-beam computed tomography (CBCT).MethodsSix databases, including PubMed, Scopus, Web of Science, the Cochrane Library, Embase and ProQuest, were searched for relevant articles published prior to June 2025. Only human studies using CBCT to investigate iM3-related ERR on M2s were included. Methodological quality was assessed via the Joanna Briggs Institute Critical Appraisal Checklist. Meta-analyses were performed using random-effects model with heterogeneity assessed via RevMan software. Meta-regression analysis and the evaluation of publication bias were conducted using Stata software.ResultsTwenty-six articles were included in this meta-analysis: Twenty-four had a low risk of bias. The quantitative synthesis demonstrated a 31% (95% confidence interval: 26% to 36%) prevalence of iM3-associated ERR on M2s. A higher prevalence of ERR was associated with mesioangular, horizontal or transverse inclination of iM3s and direct contact with the apical third of the M2 roots.ConclusionsFindings indicate that ERR on M2s is related to the contact between iM3s and the apical third of M2 roots. Mesioangular, horizontal, or transverse iM3 inclinations are significantly associated with higher ERR prevalence on adjacent M2s. This study identified key associated factors for ERR on M2s with iM3s, providing evidence-based suggestions for treatment plans of asymptomatic iM3s.RegistrationPROSPERO: CRD42022332920.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12903-026-08820-7.Keywords: External root resorption, impacted third molars, cone-beam computed tomography, meta-analysis Background Tooth impaction refers to a condition in which a tooth fails to erupt into its normal functional position due to factors such as malposition, inadequate space, or physical obstruction along its eruption path [1]. While any tooth may be impacted, third molars (M3s) are particularly susceptible owing to their late eruption timing. Reported rates of M3 impaction vary considerably across different populations and study designs, ranging from 16.67% to 68.6% [1–7]. Impacted third molars (iM3s) are associated with various pathological conditions, including periodontitis, dental caries, cyst formation, and tumor development. Among these, external root resorption (ERR) of the adjacent second molar (M2) represents a particularly prevalent and serious outcome, which may cause endodontic diseases or even tooth extraction [8]. ERR is a localized pathological process characterized by the loss of cementum and dentin beginning from the external root surface. This resorption is mediated by host immune responses and odontoclastic activity triggered by mechanical or chemical stimulation [9, 10]. Based on etiology and clinical manifestations, ERR can be categorized into several forms [10, 11]. This study focuses specifically on external surface resorption, which is primarily caused by pressure and frequently associated with impacted teeth, orthodontic forces, cysts, or tumors. Sustained mechanical pressure from an iM3 may inflict injury on the root surface of the adjacent M2, leading to periodontal ligament damage, compromised vascular supply, and ultimately cementoblast necrosis [9]. Clinically, this form of resorption is often asymptomatic and undetectable until it approaches the pulp tissue. Without timely diagnosis and intervention, it can progress to endodontic involvement, tooth mobility, or tooth loss [12]. Hence, early detection of ERR is critical for preventing irreversible damage. Given that ERR is not readily visible during clinical examination, radiographic imaging serves as the cornerstone of diagnosis. Compared with intraoral radiographs and panoramic radiography (PAN), cone-beam computed tomography (CBCT) overcomes the inherent limitations of two-dimensional modalities, such as structural superimposition, geometric distortion, and magnification errors, demonstrating superior sensitivity and accuracy in detecting ERR on M2s [8, 13]. Oenning et al. reported that CBCT enabled the detection of ERR at a rate 4.3 times higher than that of PAN [1], establishing it as the imaging modality of choice for diagnostic confidence. Despite the precision of CBCT in identifying minor ERR, CBCT-based findings on the clinical features of iM3-related ERR on M2s are both scarce and contradictory. Therefore, this systematic review and meta-analysis aims to quantitatively synthesize the existing CBCT-based evidence to evaluate the strength of association between ERR on adjacent M2s with iM3s, and identify the key associated factors influencing the prevalence and severity of ERR. Methods Protocol and registration This meta-analysis was performed in accordance with the Cochrane Handbook for Systematic Reviews of Interventions and reported following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [14]. The study protocol was registered a priori with the International Prospective Register of Systematic Reviews (PROSPERO; registration number: CRD42022332920). The research question was structured via the PECOS framework: Population (patients with iM3s), Exposure (iM3s), Comparison (absence of iM3s or specific characteristics of iM3s), Outcome (ERR of adjacent M2s), Study design (observational study design). Search strategy A comprehensive search was systematically conducted across multiple electronic databases, including PubMed, Scopus, Web of Science, the Cochrane Library and Embase, from inception until June 23, 2025. The search strategy incorporated Medical Subject Headings (MeSH) terms and free-text keywords related to impacted or unerupted teeth, ERR and M3s, combined with appropriate Boolean operators (AND, OR). The detailed search strategies for each database are provided in Table S1 in the supplementary material. Gray literature was identified via ProQuest. To enhance relevance, filters were applied to exclude records related to children, pediatrics, case reports, literature reviews and systematic reviews. No restrictions were placed on publication date or language. Eligibility criteria Studies were included according to the following criteria: Clinical studies investigating the prevalence of ERR on M2s adjacent to iM3s; Clinical studies investigating the prevalence of ERR on M2s adjacent to iM3s; Use CBCT as the imaging modality; Use CBCT as the imaging modality; Observational study design; Observational study design; Availability of prevalence data for ERR on adjacent M2s, either directly or calculable from raw data. Availability of prevalence data for ERR on adjacent M2s, either directly or calculable from raw data. The exclusion criteria were as follows: Studies involving patients with craniofacial anomalies, genetic syndromes or primary dentition; Studies involving patients with craniofacial anomalies, genetic syndromes or primary dentition; Studies where three or more independent evaluators reported discrepant ERR diagnostic results without a predefined consensus resolution method, leading to unextractable or unreliable prevalence data; Studies where three or more independent evaluators reported discrepant ERR diagnostic results without a predefined consensus resolution method, leading to unextractable or unreliable prevalence data; Unavailable full text; Unavailable full text; Non-eligible publication types, including case reports, case series, reviews, conference abstracts, opinion pieces, letters, posters, or in vitro studies. Non-eligible publication types, including case reports, case series, reviews, conference abstracts, opinion pieces, letters, posters, or in vitro studies. Literature screening and data extraction All identified records were imported into EndNote X9 (Thomson Reuters, New York, NY, USA) for management. Duplicates were removed via both automated tools and manual inspection. Two reviewers independently screened titles and abstracts, followed by a full-text assessment against eligibility criteria. Inter-reviewer agreement was assessed, and any disagreements were resolved through discussion or consultation with a third reviewer. Data extraction was performed independently by two authors: one performed the initial extraction, and the second verified the extracted data for accuracy. The extracted information included study characteristics (authors, publication year, design), sample demographics (sample size, age, sex), features of iM3s (position, inclination, impacted level, contact with M2s, associated dental follicle, impaction type), ERR characteristics (location, severity, prevalence), CBCT acquisitions and main findings. As the included studies employed slightly different diagnostic criteria for ERR, prevalence data were extracted according to the original definitions used in each study to preserve the validity of the reported outcomes. For the subgroup analysis of the relationship between impaction depth and ERR, only studies using the Pell and Gregory system to classify the impaction level of iM3s were included [2, 5]. Quality appraisal Methodological quality of the included studies was assessed via the Joanna Briggs Institute (JBI) critical appraisal tool for analytical cross-sectional studies [15, 16]. This tool evaluates the risk of bias across several domains, such as the clarity of inclusion criteria, appropriateness of subject selection, methods for measuring exposure and outcome, control of confounding management, and validity of statistical analysis. Two reviewers independently rated each study using “yes”, “no”, or “unclear” responses. Discrepancies were discussed with a senior endodontist. Studies were categorized as having low (≥ 70% “yes”), moderate (50–69% “yes”), or high (≤ 49% “yes”) risk of bias based on total scores. Statistical analysis Analyses were conducted using Review Manager 5.4 (RevMan, The Nordic Cochrane Centre, Copenhagen, Denmark). The risk difference (RD) was calculated if the data were close to a normal distribution; otherwise, the odds ratio (OR) should be employed. Pooled estimates were reported with 95% confidence intervals (CIs). All meta-analyses were performed via random-effects models, weighed by the inverse variance method on the basis of the DerSimonian and Laird method [17]. Heterogeneity was evaluated through the Q test and I² statistic, with I² > 50% or p < 0.10 indicating significant heterogeneity. Subgroup analyses were conducted to evaluate the impact of categorical variables, such as the impacted inclination and impacted depth on ERR. Forest plots were constructed to graphically represent the overall effect and for subsequent visual inspection analyses (p < 0.05 was considered significant). Sensitivity analysis was performed by sequentially excluding each study to assess its influence on the overall results. Meta-regression analysis was performed to assess the relationship between the prevalence of ERR and mean age of patients or the voxel size and field of view (FOV) of CBCT to seek possible source of heterogeneity. Publication bias was evaluated via funnel plots and Egger’s regression test if ten or more studies were included, implemented in Stata software (version 17, Stata Crop, USA). Rosenthal’s Fail-Safe N (FSN) analysis was also performed as a supplement. Results Study selection A total of 1550 records were initially identified through database searching (Fig. 1). After duplicate removal, 994 articles underwent title and abstract screening. Subsequently, 54 full-text articles were assessed for eligibility. Following a detailed evaluation, 28 studies were excluded for the specific reasons documented in Table S2. Ultimately, 26 studies [1–7, 9, 12, 18–34] met the inclusion criteria and were included in the systematic review and meta-analysis (inter-reviewer agreement, kappa = 0.86). Manual screening of reference lists from the included studies yielded no additional eligible articles. Flow diagram of the literature search and selection criteria Study characteristics A total of 6285 patients with 10,344 iM3s were included across the studies. The number of iM3s per study ranged from 73 to 2156, with a mean of 398. All included articles were cross-sectional in design. One study, Şahin et al. was in a self-paired design, comparing M2s adjacent to iM3s with contralateral M2s adjacent to erupted M3s [19]. Four studies directly compared the diagnostic performance of CBCT and PAN, confirming the superior accuracy of CBCT in detecting ERR [1, 7, 20, 31]. The relationship between ERR and iM3 location was evaluated in 22 studies. The angulation of iM3s, classified primarily via Winter’s system, was analyzed in 19 studies for its influence on the prevalence of ERR. Among the 14 studies assessing impaction level, most applied the Pell and Gregory classification; three studies focusing on maxillary iM3s used the modified Archer classification [2, 29, 30]. Proximity between M2/M3 (yes/no) was defined as a distance of ≤ 0.5 mm [4, 25]. There were 6 studies assessing the influence of contact with iM3s on the prevalence of ERR [4, 5, 24, 25, 29, 32], among which 3 studies further analyzed the association between contact location on M2s and ERR prevalence [4, 25, 32]. The association between ERR prevalence and impaction type (fully bony vs. soft tissue impaction) was examined in four studies [4, 24, 25, 28], as was the relationship between the dental follicle size of iM3s and ERR on adjacent M2s. The detailed demographic and study characteristics are summarized in Table 1. Table 1Summary of the descriptive characteristics of the studies included (n = 26)Study IDStudy DesignSampleAge (year)GroupsCBCT acquisition parametersOutcomesPrevalence of ERR5Severity of ERROenning et al. (2014) [1]Cross-sectional66 patients (25 M1, 41 F2)188 iM33sMean: 22.8Range: 15–76Teeth were assessed byPanoramic radiography and CBCT4 (self-control)A Classic i-CAT CBCT unit (Imaging Sciences International, Hatfield, PA)Scanning parameters: 120 kvp, 8 mAVoxel size: 0.25 mmFOV6: 13 × 17 cmM27s: 188CBCT: None:145 (77.12%)ERR: 43 (22.88%)Panoramic radiography:None: 178 (94.69%)ERR: 10 (5.31%)NROenning et al. (2015) [22]Cross-sectional116 patients (46 M, 70 F)174 iM3sMean: 23.7Range: 14–62Not listedA Classic i-CAT CBCT unit (Imaging Sciences International, Hatfield, PA)Scanning parameters: 120 kvp, 8 mAVoxel size: 0.25 mmFOV: 13 × 17 cmM2s: 174i-CAT Classic unit: None: 49 (56.32%)ERR: 38 (43.68%)Picasso-Trio unit: None: 39 (44.82%)ERR: 48 (55.18%)Total: None: 88 (50.57%)ERR: 86 (49.43%)Slight: 65Moderate: 19Severe:2Wang et al. (2017) [12]Cross-sectional216 patients (111 M, 105 F)362 mandibular iM3sMean: 30.45Range: 16–71NRA CBCT unit (New Tom, Verona, Italy)Operating parameters: 110 kV, 7.3 mAFixed focal spot: 0.5 mmFOV: 18 × 16 cmM2s: 362None: 289 (79.83%)ERR: 73 (20.17%)Slight: 55Moderate: 9Severe: 9Hermann et al. (2019) [26]Cross-sectional86 patients (26 M, 60 F)111 maxillary iM3sMean: 26Range: 15–55NRA Scanora 3D8 unit (Soredex, Helsinki, Finland)Operating parameters: 90 kV, 10 mAVoxel resolution: 0.13 mmFOV: 6 × 6 cmExposure time: 23 sBase projuctions: 451M2s: 111None: 69 (62.16%)ERR: 42 (37.84%)NRLi et al. (2019) [21]Cross-sectional276 patients (124 M, 142 F)507 iM3s16–2525–35≥ 36NRDCTPRO, VATECH, Yongin-Si, Republic of KoreaOperating parameters: 90 kVp, 9.0 mAVoxel size: 0.2 mmFOV: 16 × 7 cmScanning time: 24 sM2s: 507None: 276 (54.4%)ERR: 231 (45.6%)Slight: 166Moderate: 37Severe: 28Smailienė et al. (2019) [5]Cross-sectional109 patients (41 M, 68 F)254 iM3s< 23, mean: 20.96> 23, mean: 32.03median: 23NRWhiteFox machine (de Götzen S.r.l., ACTEON Group, Olgiate Olona, Italy)Scanning parameters: 105 kV, 16 mAVoxel size: 0.5 mmFOV: 150 × 130 mmSize of slices: 0.2 mmM2s: 254None: 152 (59.8%)ERR: 102 (40.2%)Slight: 76Moderate: 20Severe: 6Şahin et al. (2019) [19]Cross-sectional200 patients200 mandibular iM3sMean: 28.19 ± 8.3Range: 23–35Treatment: 200 M2s with iM3sControl: 200 M2s with erupted M3sNRTreatment: 200 M2sNone: 163 (81.5%)ERR: 37 (18.5%)Control: 200 M2sNone: 200 (100%)ERR: 0NRSuter et al. (2019) [25]Cross-sectional433 patients (198 M, 235 F)640 mandibular iM3sMean: 28.2Range: 15–80 (234 (54.0%) < 25)NR3D Accuitomo 170 scanner (Morita Corp.,Kyoto,Japan) using a full scan rotation (360) of 17.5 s(standard mode) or 10.5 s (high-speed mode)Exposure settings: 90 kV, 5.0 mAFOV/Voxel size: 4 × 4 cm/80 µm, 6 × 6 cm/125 µm, 8 × 8 cm/160 µm, 10 × 10 cm/160 µmM2s: 640None: 436 (68.1%)ERR: 204 (31.9%)Slight: 193Moderate: 9Severe: 2Tassoker et al. (2019) [6]Cross-sectional200 patients (82 M, 118 F)200 iM3sMean: 27.26 ± 8.6Range: 18–50M: mean: 29.1 ± 9.27F: mean: 25.97 ± 7.90NR3D Accuitomo 170 machine (Morita, Kyoto, Japan)M2s: 200None: 158 (79%)ERR: 42 (21%)Slight: 30Moderate: 4Severe: 8Velickovic et al. (2019) [23]Cross-sectional46 patients (27 M, 19 F)96 iM3sMean: 38.3Range: 19–75NROrthophos XG 3D (Sirona Dental Systems GmbH,Bensheim, Germany)Field size: 8 × 8 cmM2s: 98None: 89 (91.67%)ERR: 9 (8.33%)Slight: 8Moderate: 0Severe: 1Orhan et al. (2020) [27]Cross-sectional351 patients (143 M, 208 F)351 mandibular iM3smean: 28M: mean: 29.8, range: 15–70F: mean: 26.7, range: 15–71NR1. CBCT: Morita 3D Accuitomo 170 (J Morita, Kyoto Japan)FOV/Voxel size: 40 × 40 mm/0.08 mm; 50 × 50/0.2 mm;2. CBCT: Planmeca Promax 3D Max (Helsinki, Finland)Voxel size: 0.2 mmFOV: 55 × 50 mm, 100 × 50 mmM2s: 351None: 235 (66.05%)ERR: 116 (33.05%)NRSchriber et al. (2020) [4]Cross-sectional84 patients (48 M, 36 F)84 iM3sMean: 34.6Median: 28< 25: 33 (39.3%)25–35: 20 (23.8%)> 35: 31 (36.9%)NR3D Accuitomo 170 scanner (Morita Corp.,Kyoto,Japan) using a full scan rotation (360) of 17.5 s(standard mode) or 10.5 s (high-speed mode)Exposure ssettings: 90 kV, 5.0 mAFOV/Voxel size: 4 × 4 cm/80 µm, 6 × 6 cm/125 µm, 8 × 8 cm/160 µm, 10 × 10 cm/160 µmM2s: 84None: 42 (50%)ERR: 42 (50%)Slight: 31Moderate: 5Severe: 6Tunç et al. (2020) [24]Cross-sectional167 patients (73 M, 94 F)250 iM3sMean: 26.08 ± 4Range: 18–40NRA KaVo 3D eXam device (KaVo Dental, Biberach, Germany)Operating parameters: 120 kv (peak), 10.11–18.54 mAsVoxel size: 0.25–0.4 mmFOV: 160 × 60 to 130Scanning time: 4.8–8.9 sM2s: 250None: 166 (66.4%)ERR: 84 (33.6%)Slight: 44Moderate: 24Severe: 16Sakhdari et al. (2021) [20]Cross-sectional320 patients (130 M, 190 F)320 iM3sMean: 28.25 ± 7.66Range: 16–62Treatment: diagnoses based on findings in CBCTControl: initial treatment plan based on the clinical examination and the findings in the panoramic imageRotograph Evo 3D (Villa, Italy) and Carestream9300 (Kodak, France)Operating parameters: 80–90 kVp, 9–10 mAVoxel size: 0.180–0.185 mmFOV: 8.5 × 8.5 cm or 10 × 10 cmM2s: 320None: 237 (74.06%)ERR: 83 (25.94%)Slight: 41Moderate: 23Severe: 19Yesiltepe et al. (2022) [2]Cross-sectional121 patients189 maxillary iM3s≥ 18NRPlanmeca Promax 3D Mid (Planmeca Oy, Helsinki, Finland)M2s: 189None: 125 (66.1%)ERR: 64 (33.9%)NRNaik et al. (2022) [7]Cross-sectional73 patients73 mandibular iM3sRange: 18–40Teeth were assessed byPanoramic radiography and CBCT (self-control)Soredex Cranex 3DXScanning parameters: 89 kV, 7 mAVoxel size: 0.124 mmFOV: 5 × 5 cmExposure time: 15–17 sResolution: Auto WL‑556 and WW‑3307M2s: 73None: 40 (54.79%)ERR: 33 (45.21%)NRZou et al. (2023) [9]Cross-sectional261 patients (76 M, 185 F)360 mandibular iM3sMean: 27.78 ± 7.92NRNEWtom VG device (Verona, Italy)Operating parameters: 70–110 kVp, 6.8–10 mAsVoxel size: 200 μmFOV: 8 × 8 cm cm2 (height×diameter) to 6 × 6 cm cm2Axial pitch: 0.125 mm; axial thickness: 0.400 mmM2s: 360None: 235 (65.28%)ERR: 125 (34.72%)Slight: 56Moderate: 30Severe: 39Gürses et al. (2023) [28]Cross-sectional129 patients (63 M, 66 F)437 iM3sRange: 16–55Mean: 24.7NR832.32 mGy x cm2 using an 8 cm x 15 cm FOV areaOperating parameters: 90 kV, 5.0 mAVoxel size: 0.250 mmExposure time: 8.14M2s: 381None: 174 (45.67%)ERR: 207 (54.33%)Slight: 144Moderate: 49Severe: 14Lacerda-Santos et al. (2023) [18]Cross-sectional97 patients (26 M, 71 F)107 iM3sNot listedNR1. CBCT: an Orthophos XG 3D scanner (Dentsply Sirona)Operating parameters: 85 kV, 5 mAVoxel size: 0.1 mmFOV: 5 × 5 cmSample: 33 scans2. CBCT: an Orthopantomograph OP300 scanner(Instrumentarium Dental)Operatiing parameters: 95 kV, 5 mAVoxel size: 0.1 mmFOV: 5 × 5 cmSample: 74 scansM2s: 107None: 56 (52.34%)ERR: 51 (47.66%)Slight: 20Moderate: 11Severe: 20Zhou et al. (2023) [30]Cross-sectional88 patients (30 M, 63 F)93 maxillay iM3s≤ 20: 321–30: 3731–40: 25> 40: 28NRKaVo3DeXam, GermanyM2s: 93None: 52 (55.91%)ERR: 41 (44.09%)NRAkkitap et al. (2023) [29]Cross-sectional348 patients (168 M, 180 F)640 iM3sRange: 25–73Mean: 35.42NROperating parameters: 90 kVp and 10 mAVoxel size: 0.2 mmFOV: 16 × 9 cmM2s: 640None: 396 (61.88%)ERR: 244 (38.13%)Slight: 90Moderate: 42Severe: 112Baensch et al. (2024) [31]Cross-sectional367 patients (174 M, 193 F)914 iM3sRange: 11–87Mean: 26.97 ± 16.31Group 1: there were signs of potential ERR in the PAN9 (overlapping of the iM3 crown and the M2 root (s) and/or visible resorption of the M2)Group 2: there were no signs of ERR in the PANNewTom5G XL, Quantitative Radiology, Verona, Italy;Voxel size: 0.2 mmFOV: 12 × 8 cmM2s: 914None: 728 (79.65%)ERR: 186 (20.35%)Slight: 146Moderate: 22Severe: 18Cui et al. (2024) [3]Cross-sectional327 patients (166 M, 161 F)578 mandibular iM3sRange: 18–63Mean: 28.67 ± 9.23NRKaVo 3D eXam, GermanyM2s: 578None: 440 (76.12%)ERR: 138 (23.88%)Slight: 91Moderate: 32Severe: 15Lima et al. (2024) [34]Cross-sectional200 petients310 iM3sRange: 18–80NRNewTom5G XL, Quantitative Radiology, Verona, ItalyVoxel size: 0.2 mmFOV: 12 × 8 cmM2s: 310None: 275 (88.71%)ERR: 35 (11.29%)NRZhang et al. (2024) [33]Cross-sectional806 patients806 iM3sRange: ≥ 21NRNRM2s: 806None: 735 (91.19%)ERR: 71 (8.81%)NRKou et al. (2025) [32]Cross-sectional798 patients2156 iM3sMean: 36.7 ± 12.3NRA HiRes3D device (LARGEV)Operating parameters:100 kVp, 2 to 10 mAVoxel size: 0.075–0.25 mmFOV: 12 × 8 cmM2s: 2156None: 1810 (83.95%)ERR: 346 (16.05%)Slight: 180Moderate: 116Severe: 50(1) M: Male; (2) F: Female; (3) iM3: Impacted third molar; (4) CBCT: Cone-beam computed tomography; (5) ERR: External root resorption; (6) FOV: field of view; (7) M2: Second molar; (8) 3D: three-dimensional; (9) PAN: Panoramic film Summary of the descriptive characteristics of the studies included (n = 26) 66 patients (25 M1, 41 F2) 188 iM33s Mean: 22.8 Range: 15–76 Teeth were assessed by Panoramic radiography and CBCT4 (self-control) A Classic i-CAT CBCT unit (Imaging Sciences International, Hatfield, PA) Scanning parameters: 120 kvp, 8 mA Voxel size: 0.25 mm FOV6: 13 × 17 cm M27s: 188 CBCT: None:145 (77.12%) ERR: 43 (22.88%) Panoramic radiography: None: 178 (94.69%) ERR: 10 (5.31%) 116 patients (46 M, 70 F) 174 iM3s Mean: 23.7 Range: 14–62 A Classic i-CAT CBCT unit (Imaging Sciences International, Hatfield, PA) Scanning parameters: 120 kvp, 8 mA Voxel size: 0.25 mm FOV: 13 × 17 cm M2s: 174 i-CAT Classic unit: None: 49 (56.32%) ERR: 38 (43.68%) Picasso-Trio unit: None: 39 (44.82%) ERR: 48 (55.18%) Total: None: 88 (50.57%) ERR: 86 (49.43%) Slight: 65 Moderate: 19 Severe:2 216 patients (111 M, 105 F) 362 mandibular iM3s Mean: 30.45 Range: 16–71 A CBCT unit (New Tom, Verona, Italy) Operating parameters: 110 kV, 7.3 mA Fixed focal spot: 0.5 mm FOV: 18 × 16 cm M2s: 362 None: 289 (79.83%) ERR: 73 (20.17%) Slight: 55 Moderate: 9 Severe: 9 86 patients (26 M, 60 F) 111 maxillary iM3s Mean: 26 Range: 15–55 A Scanora 3D8 unit (Soredex, Helsinki, Finland) Operating parameters: 90 kV, 10 mA Voxel resolution: 0.13 mm FOV: 6 × 6 cm Exposure time: 23 s Base projuctions: 451 M2s: 111 None: 69 (62.16%) ERR: 42 (37.84%) 276 patients (124 M, 142 F) 507 iM3s 16–25 25–35 ≥ 36 DCTPRO, VATECH, Yongin-Si, Republic of Korea Operating parameters: 90 kVp, 9.0 mA Voxel size: 0.2 mm FOV: 16 × 7 cm Scanning time: 24 s M2s: 507 None: 276 (54.4%) ERR: 231 (45.6%) Slight: 166 Moderate: 37 Severe: 28 109 patients (41 M, 68 F) 254 iM3s < 23, mean: 20.96 > 23, mean: 32.03 median: 23 WhiteFox machine (de Götzen S.r.l., ACTEON Group, Olgiate Olona, Italy) Scanning parameters: 105 kV, 16 mA Voxel size: 0.5 mm FOV: 150 × 130 mm Size of slices: 0.2 mm M2s: 254 None: 152 (59.8%) ERR: 102 (40.2%) Slight: 76 Moderate: 20 Severe: 6 200 patients 200 mandibular iM3s Mean: 28.19 ± 8.3 Range: 23–35 Treatment: 200 M2s with iM3s Control: 200 M2s with erupted M3s Treatment: 200 M2s None: 163 (81.5%) ERR: 37 (18.5%) Control: 200 M2s None: 200 (100%) ERR: 0 433 patients (198 M, 235 F) 640 mandibular iM3s Mean: 28.2 Range: 15–80 (234 (54.0%) < 25) 3D Accuitomo 170 scanner (Morita Corp.,Kyoto, Japan) using a full scan rotation (360) of 17.5 s (standard mode) or 10.5 s (high-speed mode) Exposure settings: 90 kV, 5.0 mA FOV/Voxel size: 4 × 4 cm/80 µm, 6 × 6 cm/125 µm, 8 × 8 cm/160 µm, 10 × 10 cm/160 µm M2s: 640 None: 436 (68.1%) ERR: 204 (31.9%) Slight: 193 Moderate: 9 Severe: 2 200 patients (82 M, 118 F) 200 iM3s Mean: 27.26 ± 8.6 Range: 18–50 M: mean: 29.1 ± 9.27 F: mean: 25.97 ± 7.90 M2s: 200 None: 158 (79%) ERR: 42 (21%) Slight: 30 Moderate: 4 Severe: 8 46 patients (27 M, 19 F) 96 iM3s Mean: 38.3 Range: 19–75 Orthophos XG 3D (Sirona Dental Systems GmbH, Bensheim, Germany) Field size: 8 × 8 cm M2s: 98 None: 89 (91.67%) ERR: 9 (8.33%) Slight: 8 Moderate: 0 Severe: 1 351 patients (143 M, 208 F) 351 mandibular iM3s mean: 28 M: mean: 29.8, range: 15–70 F: mean: 26.7, range: 15–71 1. CBCT: Morita 3D Accuitomo 170 (J Morita, Kyoto Japan) FOV/Voxel size: 40 × 40 mm/0.08 mm; 50 × 50/0.2 mm; 2. CBCT: Planmeca Promax 3D Max (Helsinki, Finland) Voxel size: 0.2 mm FOV: 55 × 50 mm, 100 × 50 mm M2s: 351 None: 235 (66.05%) ERR: 116 (33.05%) 84 patients (48 M, 36 F) 84 iM3s Mean: 34.6 Median: 28 < 25: 33 (39.3%) 25–35: 20 (23.8%) > 35: 31 (36.9%) 3D Accuitomo 170 scanner (Morita Corp.,Kyoto, Japan) using a full scan rotation (360) of 17.5 s (standard mode) or 10.5 s (high-speed mode) Exposure ssettings: 90 kV, 5.0 mA FOV/Voxel size: 4 × 4 cm/80 µm, 6 × 6 cm/125 µm, 8 × 8 cm/160 µm, 10 × 10 cm/160 µm M2s: 84 None: 42 (50%) ERR: 42 (50%) Slight: 31 Moderate: 5 Severe: 6 167 patients (73 M, 94 F) 250 iM3s Mean: 26.08 ± 4 Range: 18–40 A KaVo 3D eXam device (KaVo Dental, Biberach, Germany) Operating parameters: 120 kv (peak), 10.11–18.54 mAs Voxel size: 0.25–0.4 mm FOV: 160 × 60 to 130 Scanning time: 4.8–8.9 s M2s: 250 None: 166 (66.4%) ERR: 84 (33.6%) Slight: 44 Moderate: 24 Severe: 16 320 patients (130 M, 190 F) 320 iM3s Mean: 28.25 ± 7.66 Range: 16–62 Treatment: diagnoses based on findings in CBCT Control: initial treatment plan based on the clinical examination and the findings in the panoramic image Rotograph Evo 3D (Villa, Italy) and Carestream 9300 (Kodak, France) Operating parameters: 80–90 kVp, 9–10 mA Voxel size: 0.180–0.185 mm FOV: 8.5 × 8.5 cm or 10 × 10 cm M2s: 320 None: 237 (74.06%) ERR: 83 (25.94%) Slight: 41 Moderate: 23 Severe: 19 121 patients 189 maxillary iM3s M2s: 189 None: 125 (66.1%) ERR: 64 (33.9%) 73 patients 73 mandibular iM3s Teeth were assessed by Panoramic radiography and CBCT (self-control) Soredex Cranex 3DX Scanning parameters: 89 kV, 7 mA Voxel size: 0.124 mm FOV: 5 × 5 cm Exposure time: 15–17 s Resolution: Auto WL‑556 and WW‑3307 M2s: 73 None: 40 (54.79%) ERR: 33 (45.21%) 261 patients (76 M, 185 F) 360 mandibular iM3s NEWtom VG device (Verona, Italy) Operating parameters: 70–110 kVp, 6.8–10 mAs Voxel size: 200 μm FOV: 8 × 8 cm cm2 (height×diameter) to 6 × 6 cm cm2 Axial pitch: 0.125 mm; axial thickness: 0.400 mm M2s: 360 None: 235 (65.28%) ERR: 125 (34.72%) Slight: 56 Moderate: 30 Severe: 39 129 patients (63 M, 66 F) 437 iM3s Range: 16–55 Mean: 24.7 832.32 mGy x cm2 using an 8 cm x 15 cm FOV area Operating parameters: 90 kV, 5.0 mA Voxel size: 0.250 mm Exposure time: 8.14 M2s: 381 None: 174 (45.67%) ERR: 207 (54.33%) Slight: 144 Moderate: 49 Severe: 14 97 patients (26 M, 71 F) 107 iM3s 1. CBCT: an Orthophos XG 3D scanner (Dentsply Sirona) Operating parameters: 85 kV, 5 mA Voxel size: 0.1 mm FOV: 5 × 5 cm Sample: 33 scans 2. CBCT: an Orthopantomograph OP300 scanner (Instrumentarium Dental) Operatiing parameters: 95 kV, 5 mA Voxel size: 0.1 mm FOV: 5 × 5 cm Sample: 74 scans M2s: 107 None: 56 (52.34%) ERR: 51 (47.66%) Slight: 20 Moderate: 11 Severe: 20 88 patients (30 M, 63 F) 93 maxillay iM3s ≤ 20: 3 21–30: 37 31–40: 25 > 40: 28 M2s: 93 None: 52 (55.91%) ERR: 41 (44.09%) 348 patients (168 M, 180 F) 640 iM3s Range: 25–73 Mean: 35.42 Operating parameters: 90 kVp and 10 mA Voxel size: 0.2 mm FOV: 16 × 9 cm M2s: 640 None: 396 (61.88%) ERR: 244 (38.13%) Slight: 90 Moderate: 42 Severe: 112 367 patients (174 M, 193 F) 914 iM3s Range: 11–87 Mean: 26.97 ± 16.31 Group 1: there were signs of potential ERR in the PAN9 (overlapping of the iM3 crown and the M2 root (s) and/or visible resorption of the M2) Group 2: there were no signs of ERR in the PAN NewTom 5G XL, Quantitative Radiology, Verona, Italy; Voxel size: 0.2 mm FOV: 12 × 8 cm M2s: 914 None: 728 (79.65%) ERR: 186 (20.35%) Slight: 146 Moderate: 22 Severe: 18 327 patients (166 M, 161 F) 578 mandibular iM3s Range: 18–63 Mean: 28.67 ± 9.23 M2s: 578 None: 440 (76.12%) ERR: 138 (23.88%) Slight: 91 Moderate: 32 Severe: 15 200 petients 310 iM3s NewTom 5G XL, Quantitative Radiology, Verona, Italy Voxel size: 0.2 mm FOV: 12 × 8 cm M2s: 310 None: 275 (88.71%) ERR: 35 (11.29%) 806 patients 806 iM3s M2s: 806 None: 735 (91.19%) ERR: 71 (8.81%) 798 patients 2156 iM3s A HiRes3D device (LARGEV) Operating parameters: 100 kVp, 2 to 10 mA Voxel size: 0.075–0.25 mm FOV: 12 × 8 cm M2s: 2156 None: 1810 (83.95%) ERR: 346 (16.05%) Slight: 180 Moderate: 116 Severe: 50 (1) M: Male; (2) F: Female; (3) iM3: Impacted third molar; (4) CBCT: Cone-beam computed tomography; (5) ERR: External root resorption; (6) FOV: field of view; (7) M2: Second molar; (8) 3D: three-dimensional; (9) PAN: Panoramic film Methodological quality of the included studies Among the 26 included studies, the majority were assessed as having a low risk of bias, with two exceptions: one study was rated as high risk [27], and the other was rated as moderate risk [33]. The overall average quality score was 79.5% (Figs. 2 and 3). It should be noted, however, that none of the studies satisfied all of the methodological quality criteria. A detailed breakdown of the risk of bias assessments is provided in Table S3. Risk of bias summary. (+) = yes; (?) = unclear; (-) = no Risk of bias graph Synthesis of the results Overall prevalence The prevalence of ERR on M2s adjacent to iM3s varied considerably across the included studies, ranging from 8.81% to 54.33%. A random-effects model was used to pool data from all 26 studies with a total of 8188 teeth, yielding an overall pooled prevalence of 31% (95% CI: 26% to 36%; n = 26) (Fig. 4). Significant heterogeneity was observed among the studies (I²= 97%, df = 25, p < 0.00001), prompting to the conduct of meta-regression analysis and several subgroup analyses to explore potential sources of variation. Sensitivity analysis, performed by sequentially excluding individual studies, indicated that no single study substantially influenced the overall results, supporting the robustness of the pooled estimate (Table S4). The results of meta-regression with a random-effects model showed no significant relationship between ERR and the mean age or the parameter of CBCT (p > 0.05) (Fig. S1-S3). Forest plot of the overall prevalence of ERR on M2s adjacent to iM3s Gender Twelve studies with 2318 teeth were included to assess sex as a potential factor associated with ERR on M2s adjacent to iM3s [3, 4, 6, 12, 20–22, 28–30, 32]. The pooled prevalence of ERR was 32% (95% CI: 23% to 41%; n = 11) in males and 34% (95% CI: 26% to 42%; n = 11) in females. No significant difference was observed (p = 0.72), indicating that sex is not associated with the occurrence of ERR (Fig. S4). Age The subgroup analysis of the association between patients’ age and ERR was conducted based on seven studies with a total of 3870 teeth [6, 12, 21–23, 32, 33]. Samples in each study were divided into three age groups: ≤ 25 years old, 26–35 years old, and ≥ 36 years old. There was no statistically significant difference in the pooled prevalence of ERR among each group (≤ 25: 20.63%, 95% CI: 11.50% to 32.89%, n = 7; 26–35: 25.93%, 95% CI: 16.67% to 37.89%, n = 7; ≥ 36: 21.88%, 95% CI: 12.28% to 35.48%, n = 7) (p = 0.76) (Fig. S5). Position of iM3s A subgroup analysis of 22 studies evaluated the influence of the iM3 position (maxilla vs. mandible) on the prevalence of ERR [1–7, 9, 12, 18, 20, 21, 23–32]. The total sample comprised 8856 teeth. The pooled prevalence was 28.57% (95% CI: 20% to 38.27%; n = 14) in the maxilla and 31.51% (95% CI: 26.47% to 37.5%; n = 18) in the mandible, with no significant difference (p = 0.56) (Fig. S6). The frequency of severe ERR was greater in the maxilla (6.54%; 95% CI: 2.91% to 10.71%; n = 8) than in the mandible (3.85%; 95% CI: 1.96% to 7.41%; n = 10), although this difference was not statistically significant, either (p = 0.35) (Fig. S14). Inclination of iM3s Nineteen studies examining the effect of iM3 angulation on ERR were included in this analysis [1–6, 9, 18, 20, 21, 23–25, 27–30, 32, 33]. The sample size was 8178 teeth. There was a statistically significant association between the iM3 inclination and the prevalence of ERR (p < 0.00001) (Fig. S7). Horizontal iM3s were associated with the highest prevalence of ERR on adjacent M2s (45.05%; 95% CI: 34.64% to 56.14%; n = 19), followed by mesioangular (41.86%; 95% CI: 35.06% to 49.24%; n = 19) and transverse impaction (33.77%; 95% CI: 20.63% to 49.49%; n = 9). No significant differences were detected among these angulation categories (p = 0.5) (Fig. S8). The lowest ERR prevalence was found adjacent to vertical (14.53%; 95% CI: 9.09% to 22.48%; n = 19) and distoangular (23.08%; 95% CI: 16.67% to 31.03%; n = 17) iM3s. Inverted impaction cases were excluded because of generally small sample sizes and highly variable ERR prevalence rates (0–62.5%). No significant association was found between the inclination and ERR severity (p = 0.11) (Fig. S15). Impacted level of iM3s According to the Pell and Gregory’s classification, another subgroup analysis was conducted on 11 studies with a total of 5137 samples [3, 5, 12, 18–23, 28, 32]. The highest prevalence of ERR occurred adjacent to class C impaction (42.53%; 95% CI: 28.57% to 54.95%; n = 11), followed by class B impaction (30.56%; 95% CI: 21.26% to 41.86%; n = 11) and class A impaction (24.24%; 95% CI: 12.28% to 42.86%; n = 10). However, there was no significant difference among the groups (p = 0.25) (Fig. S9). No significant association was found between impacted level and ERR severity (p = 0.12) (Fig. S16). Contact between iM3s and M2s Six studies with 4024 teeth evaluated the effect of direct contact between iM3s and M2s [4, 5, 24, 25, 29, 32]. A substantially greater prevalence of ERR was observed in the contact group (39.76%; 95% CI: 25.93% to 55.17%; n = 6) than in the noncontact group (2.91%; 95% CI: 0.99% to 9.09%; n = 6) (Fig. S10). This pronounced contrast strongly aligns with the established mechanical pressure mechanism of ERR [35]. Further analysis showed that when the iM3 was in contact with the apical third of the M2 root, the prevalence of ERR was significantly higher than when it was in contact with the cervical or the middle third of the root (p < 0.00001) (Fig. S11). Impacted type of iM3s Three studies [4, 24, 28] were included to analyze the association between ERR on M2s and the type of iM3 impaction in relation to the bone. The results revealed that there was no significant difference between the influence of full bony impaction and soft tissue coverage on ERR (p = 0.68) (Fig. S12). Although not included in the meta-analysis because raw data were unavailable, Suter et al. [25] reported no significant association, either. Dental follicle size of iM3s The size of the dental follicle was regarded as absent (< 1 mm), regular (1–3 mm), enlarged (3–5 mm), or cystic (> 5 mm; excluded from the included studies). The subgroup analysis of 3 articles [4, 24, 28] revealed that the dental follicle size of iM3s had no influence on ERR (p = 0.87) (Fig. S13). Consistent with this, Suter et al. [25] reported no significant association between dental follicle size and ERR. Publication bias assessment The funnel plot exhibited asymmetry (Fig. S17), and Egger’s regression test indicated significant publication bias in the pooled prevalence estimates of ERR (p < 0.05) (Fig. S18). The FSN value was 29.79, which is below the threshold of 5k + 10 (k = 26). This indicates that the pooled result might be unstable, supporting the presence of publication bias as indicated by Egger’s test. However, the nonparametric Trim-and-Fill method did not identify any potentially missing unpublished studies (Fig. S19). This discrepancy is more likely attributed to the inherent characteristics of single-group proportion meta-analyses and substantial between-study heterogeneity (I² = 97%), rather than classic publication bias, as prevalence studies do not define dichotomous positive or negative outcomes [36]. Discussion This systematic review and meta-analysis aimed to quantitatively evaluate the association between iM3s and ERR on M2s and to identify potential factors influencing ERR prevalence, thereby assisting clinicians in decision-making regarding prophylactic extraction of iM3s. The management of asymptomatic, pathology-free iM3s remains controversial. Some clinicians advocate prophylactic extraction to prevent potential pathologies [37, 38], whereas others emphasize that many iM3s remain asymptomatic throughout life and highlight the risks of postoperative complications [39]. Due to the higher radiation exposure and cost compared with two-dimensional radiography, routine CBCT screening for early ERR detection prior to symptom development is impractical. Therefore, clarifying the degree of association between iM3s and ERR and identifying specific associated factors are essential for the targeted management of asymptomatic iM3s, where prophylactic extraction may be considered for cases with high-risk features. Summary of the main findings The pooled prevalence of ERR on M2s adjacent to iM3s was 31%. Subgroup analyses indicated that impacted angulation significantly influenced the prevalence of ERR. Specifically, horizontal, mesioangular and transverse impaction were associated with the highest prevalence of ERR, while distoangular and vertical impaction posed the lowest. Direct contact between iM3s and M2s markedly increased the likelihood of ERR, especially when the contact occurred at the apical third of M2 roots. No significant correlations were found with sex, age, arch position (maxilla vs. mandible), impaction depth, impaction type (bony vs. soft tissue), or dental follicle size. It should be noted that the current evidence regarding the impaction type and dental follicle size is still limited (only 3 for each subgroup analysis), so the non-significant results should be interpreted cautiously due to the small number of included studies and low statistical power. Meta-regression analysis implied that the age of patients and the voxel size and FOV of CBCT may not have significant impact on ERR prevalence, either. Agreements and disagreements with other studies ERR in maxilla vs. mandible No significant difference in ERR prevalence was found between maxillary and mandibular M2s in our analysis, although the literature reports conflicting results. Some studies reported a higher prevalence of ERR in the maxilla [24, 28], whereas more researchers reported higher rates in the mandible [1, 6, 20, 21, 23, 31]. This discrepancy may be attributable to anatomical variations: maxillary M3s tend toward vertical or distoangular impaction due to continued maxillary tuberosity growth and mesial tooth drift, whereas mandibular M3s are commonly impacted mesioangularly or horizontally owing to space constraints [40, 41]. Although not statistically significant, the pooled prevalence in our study was numerically greater in the mandible (maxilla: 28.57%, n = 14; mandible: 31.51%, n = 18), which is consistent with several previous reports. Contact between iM3s and M2s Since contact is almost a biological prerequisite for pressure-related ERR, which is supported by our meta-analysis, it’s important to clarify the specific contact characteristics (type/location) and proximity threshold. However, only 3 studies included have analyzed the relationship between contact location and ERR, and the subgroup analysis revealed significantly heightened ERR prevalence when iM3s contact the apical third of M2 roots. This finding was consistent with existing evidence that the apical region exhibits greater susceptibility to resorption [32, 42]. More related researches are needed and future studies should also investigate quantitative proximity thresholds (e.g., < 1 mm vs. ≥1 mm) to further refine risk stratification. Angulation of iM3s In our study, horizontal, mesioangular, and transverse inclinations were related to a higher ERR prevalence (horizontal: 45.05%, 95% CI: 34.64% to 56.14%, n = 19; mesioangular: 41.86%; 95% CI: 35.06% to 49.24%, n = 19; transverse: 33.77%; 95% CI: 20.63% to 49.49%, n = 9), although there was no statistically difference among this high-risk cluster of iM3 angulation. This finding was consistent with previous reports. For example, Matzen et al. reported that the vast majority of resorption cases occurred in relation to mesioangulated or horizontal iM3s [43, 44]. Similarly, Butt et al. investigated that M2s adjacent to horizontal iM3s exhibited the highest frequency of ERR [45]. The increased prevalence of ERR associated with these angulation is likely attributable to the greater likelihood of direct contact with iM3s, as well as the sustained mechanical pressure induced by the impacted teeth [46]. Impacted level of iM3s The relationship between the impacted level and ERR remains debated. More studies reported that class C impaction pose the highest prevalence of ERR on M2s [2, 5], which was consistent with our subgroup analysis highlighting the role of direct contact between iM3s and M2 roots. However, Oenning’s study reported contradictory findings [21]. These discrepancies may stem from variations in sample composition, such as the pooling of maxillary and mandibular cases. In the maxilla, M3s are more frequently impacted in class C, whereas mandibular iM3s most commonly present as class B, followed by class A [21, 47]. In our study, the results revealed no statistically significant difference in association between the impacted level and ERR, though the prevalence of ERR on M2s was highest when iM3s were impacted in class C (class A: 24.24%, 95% CI: 12.28% to 42.86%, n = 10; class B: 30.56%, 95% CI: 21.26% to 41.86%, n = 11; class C: 42.53%, 95% CI: 28.57% to 54.95%, n = 11). More high-quality studies with larger sample sizes and site-specific analyses are warranted to clarify these relationships. Association between ERR severity and iM3s Notably, our subgroup analyses revealed no significant association between ERR severity and arch position (p = 0.35), impaction angulation (p = 0.11) or impaction depth (p = 0.12), suggesting that these factors may not necessarily affect the progression of ERR. However, some researches have reported that there was a statistically significant association between severity and arch position of iM3s, location of ERR or the age of patients [3, 6, 12, 18, 20, 21, 32]. Possible explanations include: (1) The relatively limited number of studies that have analyzed the severity of ERR; (2) The smaller sample sizes of severe ERR compared to total ERR in studies reporting severity; (3) The variable prevalence of severe ERR across studies from 0.31% to 18.69%; (4) Sole imaging diagnosis of ERR severity. All of reasons may have reduced the statistical power to significant differences. A plausible biological explanation should also be noted. It’s that factors driving ERR initiation may differ from those governing its progression. The disease progression and severity may be more strongly modulated by systemic host factors (e.g., bone metabolism, immune status) and duration of sustained contact between iM3s and M2s, rather than static anatomical features alone. This finding implies that clinicians should not only focus on the occurrence of ERR but also conduct regular follow-up for patients with iM3s, even in cases of mild initial resorption, to monitor potential progression in severity over time. Future studies with larger cohorts, standardized severity grading and analysis of the association between host factors or contact time and the severity of ERR are warranted to more reliably assess predictors of ERR severity. Strengths and limitations This systematic review summarizes current evidence regarding the association between iM3s and ERR of adjacent M2s via CBCT-based diagnoses. The application of comprehensive database searches, strict inclusion criteria, and independent dual-reviewer screening process enhance the reliability of our findings. Compared with two-dimensional radiography, the exclusive use of CBCT—a highly accurate three-dimensional modality—improves diagnostic confidence. Statistical methods, including random-effects modeling, meta-regression analysis and subgroup analyses, were employed to address heterogeneity. Several limitations should be acknowledged. First, most included studies were cross-sectional studies in design, lacked control groups and were generally of low methodological quality. It is methodologically impossible to establish temporality or causality for this type of studies, so the findings can just indicate risk indicators rather than causal risk factors. Second, variations were observed in CBCT acquisition parameters and diagnostic criteria for ERR across studies, which may affect detection of ERR and judgement of the severity, though it seemed to be an insignificant variable for the resolution of CBCT according to meta-regression analysis. Third, the sample sizes were generally limited in most studies, and the populations varied widely in terms of economic and ethnic backgrounds. These factors contributed to substantial heterogeneity, even within subgroup analyses, with reported ERR prevalence ranging from 8.81% to 54.33%. Moreover, the lack of clinical or histologic validation means that all findings are based solely on CBCT interpretation. The high sensitivity of CBCT may also lead to overestimation of clinically relevant ERR. Therefore, the pooled estimate represents only a general approximation rather than a precise epidemiological value due to the low methological quality of this meta-analysis, and the subgroup-specific findings are likely more clinically meaningful and informative for clinical practice. High-quality cohort studies with controlled designs, longitudinal follow-up, larger sample sizes, and detailed reporting of associated factors are needed to establish temporal sequence and confirm causal relationships between iM3 characteristics and ERR. Conclusions The present study demonstrates that approximately one-third of M2s exhibit ERR associated with iM3s. The contact between iM3s and the apical third of M2 roots increased the prevalence of ERR. Mesioangular, horizontal, and transverse impactions pose a higher prevalence of ERR though the comparisons among these three groups show no statistically significant difference. Since routine CBCT screening for ERR is not advocated, this study may help clinicians to identify patients who may benefit from closer monitoring because of asysmptomatic iM3s. However, it should be emphasized that the research is based on low-quality observational evidence, so it can not give guidance about prophylactic extraction of asysmptomatic iM3s. All clinical decisions should be fully individualized based on patient symptoms and clinical examinations. Supplementary Information Supplementary Material 1 Acknowledgements Not applicable. Clinical trial number Not applicable. Abbreviations External root resorption Second molar Impacted third molar Cone-beam computed tomography Third molar Panoramic radiography Preferred Reporting Items for Systematic Reviews and Meta-Analyses Medical Subject Heading Joanna Briggs Institute Review Manager Risk difference Odds ratio Confidence interval Authors’ contributions All authors have contributed according to the ICMJE criteria. **Wenxuan Liu: ** methodology, soft-ware, validation, formal analysis, investigation, resources, data curation, writing—original draft preparation, writing—review and editing, visualization. **Ailin Liang: ** methodology, soft-ware, validation, investigation, resources, data curation, writing—original draft preparation, writing—review and editing. **Xiaoyan Zhou: ** validation, writing—review and editing. **Chenxi Liao: ** writing—review and editing. **Yanling Zhou: ** writing—review and editing. **Yihua Huang: ** methodology, visualization, supervision, project administration. **Qimei Gong: ** Conceptualization, writing—review and editing, visualization, supervision, project administration, funding acquisition. All the authors have read and agreed to the published version of the manuscript. Funding This work was funded by the Guangdong Basic and Applied Basic Research Foundation (grant number: 2025A1515010884) and the Key Clinical Technique Program of Guangzhou (number: 2023P-ZD07). Data availability All data supporting the findings of this study are available within the paper and its supplementary information file. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Wenxuan Liu and Ailin Liang contributed equally to this work. Contributor Information Yihua Huang, Email: huangyh48@mail.sysu.edu.cn. Qimei Gong, Email: gongqim@mail.sysu.edu.cn. References Associated Data Data Availability Statement All data supporting the findings of this study are available within the paper and its supplementary information file.