Diagnostic accuracy of imaging examinations for peri-implant bone defects around titanium and zirconium dioxide implants: A systematic review and meta-analysis

Dental implants are a clinical resource widely used in oral rehabilitation with a high success rate (97%) if the variables that influence their longevity are appropriately considered.1 Despite clinical success, there is still a high prevalence of patients who experience diseases of the peri-implant tissue. After 5 years of follow-up, approximately 8.6% to 9.7% of dental implants exhibit chronic inflammation of the soft and hard tissues around them. Peri-implantitis is characterised by increased probing depth, bleeding, and progressive alveolar bone loss and can result in bone defects such as dehiscence, fenestration and intrabony defects.2, 3, 4

The diagnosis of peri-implantitis and bone defects is based on clinical and imaging examinations. Early detection may lead to a more favourable prognosis. In contrast, an inaccurate diagnosis and unsatisfactory management can lead to gingival recession, aesthetic complications, and loss of implant osseointegration.4, 5, 6

Periapical radiographs (PRs), panoramic radiography (PAN), magnetic resonance imaging (MRI), computed tomography (CT), and cone-beam computed tomography (CBCT) are imaging modalities used to detect peri-implant bone defects. Some authors have suggested the initial acquisition of PRs,7, 8 which allow the visualization of mesial and distal aspects of the implant. However, 2-dimensional radiography can underestimate incipient lesions and intrabony defects.8

CBCT has become an important diagnostic tool to assess the peri-implant region. Although PR allows the visualization of mesial and distal bone areas around the implant, CBCT images prevent overlapping of anatomical structures,9, 10 thus assisting in distinguishing buccal and lingual cortical bone plates in addition to mesial and distal bone tissue. The limitations of this modality include the metallic artifacts generated by the presence of high-atomic-number materials.

Metallic artifacts result from the beam-hardening phenomenon, which occurs due to the distinct absorption of low-energy X-ray photons by materials with high atomic numbers. This creates localised hypodensities or dark voids close to dense objects, such as titanium (Ti) and zirconium (Zr), and impairs assessment of the image.11

Ti has been used for decades to manufacture dental implants due to its biocompatibility and osseointegration, but several other materials have also been used to manufacture dental implants in order to overcome the limitations of Ti. Zirconium dioxide (ZrO₂) implants are usually installed in maxillary or mandibular anterior rehabilitations for aesthetic reasons because they closely match the tooth colour, thereby avoiding darkened areas that are visible upon cervical exposure or through transparent gingival tissue.12, 13 However, both types of implants produce metallic artifacts.11, 14, 15

The demand for ZrO₂ implants in recent years has raised concerns about the presence of metallic artifacts in CBCT images. Several studies14, 15, 16, 17, 18 have reported a greater number of artifacts in images of ZrO₂ implants using some imaging modalities than in images of Ti implants. Based on these reports, it is fundamental to investigate the diagnostic accuracy of the imaging and the degree of artifact interference at the implant-bone interface for both Ti and ZrO₂ implants.

A systematic review is an important pre-specified tool that compiles available evidence to answer a defined question. Through meta-analysis it is possible to statistically analyse results from different individual studies. Both methods improve the synthesis of the best evidence, in order to guide clinicians in the decision-making process.19

Thus, this study aimed to perform a systematic review and meta-analysis to assess the diagnostic accuracy of imaging modalities in the detection of peri-implant bone defects and to compare the diagnostic accuracy between Ti and ZrO₂ implants.

This systematic review was conducted in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Diagnostic Test Accuracy (PRISMA-DTA).20 Studies reporting the accuracy of imaging exams in the evaluation of peri-implant bone defects were extracted from the following databases: MEDLINE (PubMed), EMBASE, Web of Science, ScienceDirect, Scopus, and ProQuest. Keywords related to concepts addressed by the pre-defined question of this study were used to gather all relevant articles. The PubMed search strategy included a combination of Medical Subject Headings (MeSH) terms and text words described as follows: (“accuracy”) OR (“sensitivity”) OR (“specificity”) OR (“physical measurement”) AND (“peri-implantitis” [MeSH Terms]) OR (“peri-implantitis”) OR (“bone”) OR (“bone defect”) OR (“bone loss”) OR (“bone-implant interface” [MeSH Terms] OR “bone-implant interface/diagnosis” [MeSH Terms] OR “bone-implant interface/diagnostic imaging” [MeSH Terms])) OR (“peri-implantitis/diagnostic imaging” [MeSH Terms] OR “peri-implantitis/chemically induced” [MeSH Terms] OR “peri-implantitis/diagnosis” [MeSH Terms] OR “peri-implantitis/classification” [MeSH Terms]) AND (“cone-beam computed tomography” [MeSH Terms]) OR (“spiral cone-beam computed tomography” [MeSH Terms]) OR (“diagnostic imaging” [MeSH Terms]) OR (“dental radiography”) OR (dental radiography [MeSH Terms])) OR (panoramic radiography[MeSH Terms])) OR (“periapical radiography”) OR (“ultrasound”) OR (“magnetic resonance”) OR (“tomography” [MeSH Terms]) OR (“radiography, dental, digital” [MeSH Terms] OR “radiography, dental” [MeSH Terms]) AND (“dental implants” [MeSH Terms] OR “dental implants, single-tooth” [MeSH Terms]).

For other databases, the search strategy was modified using appropriate Boolean strategies. A manual search was also conducted of reference lists in the included studies. The search was carried out in June 2020 and updated in December 2020.

The articles were evaluated by 2 independent and calibrated reviewers (kappa: 0.81). Through the Rayyan website (Qatar Computing Research Institute, URL: http://rayyan.qcri.org), duplicated files were excluded and titles and abstracts were screened according to the eligibility criteria. The initial screening included studies that evaluated the peri-implant region using any diagnostic imaging method. Studies that were not in English or Portuguese, systematic reviews, narrative reviews, letters, and case reports were excluded. The selected articles were then read in full and studies that evaluated the accuracy of diagnostic imaging methods in Ti or ZrO₂ peri-implant bone defects were included. These criteria were applied to full-text readings because titles and abstracts might not exhibit accurate or complete data. Reviewers discussed any disagreements between themselves to reach a consensus or a third reviewer was consulted. Studies that met the inclusion criteria and did not have any of the exclusion criteria were then accepted for data extraction and bias assessment.

The risk of bias and the applicability of the included studies were assessed independently by 2 reviewers using the Quality Assessment Tool for Diagnostic Accuracy Studies 2 (QUADAS-2) and Review Manager version 5.4 (The Cochrane Collaboration, London, UK). Four domains were addressed: patient selection, index test, reference test, and flow and timing. Questions related to the study design could be answered with ‘yes’ if there was a low risk of bias, ‘uncertain’ when it was not possible to determine the risk, and ‘no’ for a high risk of bias.

Data extraction was performed by 2 independent reviewers. The meta-analysis included studies that presented precise numbers or any other data enabling the calculation of diagnostic performance. The extracted data included sensitivity, specificity, true/false and positive/negative results, and the disease prevalence in each sample. The Meta-Disc software (Ramón y Cajal Hospital, Madrid, Spain) was used to perform the random-effect meta-analysis and to identify sources of heterogeneity (the Cochran Q and I² tests).

Summary receiver operating characteristic (sROC) curves were constructed using RStudio software version 1.4.1106 (Integrated Development Inc., Boston, MA, USA) to compare the effect of methodological differences in relation to the variables of each group.

The correct choice of an imaging examination is essential for the dental surgeon to identify, treat, and monitor bone defects and pathologies such as peri-implantitis.

The relevance of this study relates to the current popularity of ZrO₂ implants, primarily due to aesthetic demands. The tooth-like colour, reduced plaque accumulation (when compared to Ti implants), and promising osseointegration results45, 46 have also contributed to the appeal of this material.

Recent studies have demonstrated that, in some imaging modalities, ZrO₂ implants generated more metallic artifacts than Ti implants.14, 15, 16, 17 The presence of more metallic artifacts impacts image quality and the diagnosis of peri-implant bone defects.12, 23, 36, 43

The formation of greater numbers of artifacts in images may be associated with the higher atomic number (Z) of Zr (Z=40) than that of Ti (Z=22). A higher Z correlates with a greater expression of artifacts, modifying the grey scale that affects image contrast, hence decreasing the visibility of anatomical structures.17

Sancho-Puchades et al.17 analyzed the increased formation of metallic artifacts close to ZrO₂ implants when compared to Ti and Ti-Zr in CBCT, which may explain the high number of false-positive diagnoses and lower accuracy.23, 36 Steiger-Ronay et al.13 noticed similar results in terms of peri-implant bone defect measurements in CBCT images. Schriber et al.36 observed measurement overestimations of 1.1 mm and underestimations of 1.28 mm. The imprecisions were attributed to the magnified negative influence of the ZrO₂ metallic artifacts. According to the studies, clinicians should be aware of these discrepancies when assessing ZrO₂ implants in CBCT images.

When assessing study samples, diagnostic studies can have a wide array of variables, resulting in potential heterogeneity. For this study, heterogeneity was assessed by the Q test (Higgins et al.44). The levels of inconsistency were quantified through the I² statistic. The obtained values were considered high for the Ti group (92.5%) and low/moderate for the ZrO₂ group (76%). The meta-analysis outcomes presented significant inconsistency, reinforcing the importance of using this statistical method to consider this divergence among studies and produce more accurate results.

To assess heterogeneity, a meta-analysis of random effects was conducted. For both types of implants, defects were divided into 4-wall and fenestration defects. This subdivision of bone defect types was important for determining the accuracy in various clinical situations.

The sROC curves (Fig. 3) illustrate the data compiled from several studies. The grouping of data through simple pooled weighted averages can produce inaccurate interpretations as the data from individual studies may vary in terms of sample, methodology, and other characteristics. Two previous systematic reviews47, 48 analysed the accuracy of imaging methods in detecting bone defects in Ti implants: Bohner et al.47 used the aforementioned simple grouping of data, not exploring potential sources of heterogeneity, while Pelekos et al.48 performed an analysis of random effects but studied only Ti implants.

In the present study, sensitivity and specificity values were higher for the detection of 4-wall defects than for the other subgroups of both implant types. The AUC values (ZrO₂: 0.95 and Ti: 0.80) were also high, although not statistically significant. Hilgenfeld et al.43 evaluated the diagnostic performance of CBCT, PR, and MRI for bone defects at the implant-bone interface in ZrO₂ implants that underwent chemical induction of bone defects. As in our systematic review, CBCT and MRI showed better performance, mainly in large 4-wall defects.

Four-wall bone defects are among the most frequently observed defects in clinical situations.4 Although there is still no consensus in the literature that the morphology of the defect influences the prognosis, it has been demonstrated that some factors related to implant location and patient habits can be associated with morphological characteristics of bone defects, implying a potential relationship with disease severity. It is important to use the imaging modality that best identifies the morphology to better plan the treatment approach.4, 7, 49

Several methodologies were found in the assessed diagnostic studies, either regarding the type of artificially induced bone defect or the different CBCT protocols applied. Schwindling et al.27 analysed the accuracy of PR, low-dose CBCT, and high-dose CBCT for identification, classification, and measurement of peri-implant bone lesions in Ti implants. The accuracy of PR and CBCT were similar in the identification of defects, although low-dose CBCT images provided more accurate identification of defect morphology than PRs. High-dose CBCT slightly increased the diagnostic performance, albeit at the expense of a 14 times higher dose. This study was corroborated by Schriber et al.36 who determined that low- or high-dose protocols did not significantly affect diagnostic accuracy.

Metallic artifact reduction (MAR) algorithms are designed to reduce the expression of metallic artifacts by increasing the imaging quality during reconstruction. Some researchers23, 28, 42 tested the effectiveness of MAR algorithms in detecting peri-implant bone defects, showing no improvement for either Ti or ZrO₂ implants. Comparisons among different acquisition protocols, with variation in terms of voxel size12, 24, 40, 42 and field of view (FOV),26, 34 resulted in similar diagnostic performance in both types of implants.

The use of filters is a post-processing image manipulation that has yielded positive results in improving diagnostic performance in both Ti and ZrO₂ implants.25, 39 For ZrO₂, Bayrak et al.23 found that the adaptive image noise optimiser filter with MAR improved interobserver agreement in the detection of bone defects. Another aspect that positively influenced diagnostic performance was the size of bone defects. In PR, Kühl et al.31 found that major defects (3 mm) demonstrated higher values of sensitivity and specificity. For CBCT, both a larger size of the bone defect and an increased number of affected bone walls resulted in improved diagnostic performance.34, 41, 43

In addition to variability among the studies, another limitation was the in vitro design of all included studies. In experimental conditions, movement artifacts and gingival tissue attenuation are not reproducible.50 Furthermore, fewer studies have investigated Zr implants since these implants were developed more recently than Ti implants. These factors hindered the formulation of a systematically analysed dataset with an acceptable level of evidence.

There were insufficient data to analyse quantitatively the performance of imaging methods (PR, PAN, MRI, and CT) for Ti and ZrO₂ implants. PR has already been extensively studied using Ti implants. However, with the recent advent of Zr in implant dentistry, only 1 study43 used PR with Zr implants, and the accuracy results were lower than with CBCT and MRI.

Hilgenfeld et al.43 emphasised the potential of MRI, which demonstrated a diagnostic accuracy similar to CBCT for ZrO₂ implants and even greater accuracy in small defects, as well as the advantage of non-ionising radiation. However, more studies evaluating the use of this imaging method are required.

No studies were found that evaluated diagnostic performance for PAN and CT in ZrO₂ implants. The radiographic detection of bone defects using both PAN and CT was inferior to other methods (PR and CBCT) in Ti implants.31, 32, 38 This implies that these imaging modalities (PAN and CT) are not appropriate for peri-implantitis assessment.

CBCT diagnostic accuracy in Ti and ZrO₂ implant defects was assessed by applying several methodologies; in most studies, CBCT was superior to PR, at the expense of a higher dose of radiation. According to the American Academy of Oral and Maxillofacial Radiology,51 before the decision to perform a CBCT examination is made, its use should always be weighted and clinically justified by the patient’s clinical signs or symptoms.

In this analysis, high variability was found in imaging studies reporting the detection of bone defects in Ti and ZrO₂ implants. The use of CBCT imaging in the assessment of peri-implant bone defects in ZrO₂ implants seems to impair analysis and decrease accuracy, but statistical significance was observed. Furthermore, for Ti implants, fenestration was more accurately diagnosed than dehiscence. The use of different protocols with modification of the FOV, voxel size, MAR, and the application of filters did not significantly alter the diagnostic performance of the imaging methods for both types of implants.

In order to answer the guiding question of this systematic review with more certainty, further studies are needed using ZrO₂ implants to analyse diagnostic accuracy in different situations.