Marginal and internal fit of CAD/CAM restorations with CAD-defined cement space: a meta-analysis of micro-CT studies

Abstract

Statement of problem

In fixed restorations produced with a fully digital workflow, there is limited meta-analytic evidence examining the relationship between predefined cement gap and marginal/ internal fit.

Aim

The purpose of this meta-analysis is to examine the pooled marginal and internal fit of CAD/CAM restorations in studies in which cement space was predefined numerically during the CAD design stage and fit was assessed using micro-computed tomography (Micro-CT).

Materials and methods

The search was conducted between 2013 and 2024 across the PubMed, Scopus, Science Direct and Google Scholar databases by following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines and predefined eligibility criteria. Based on CAD/CAM fixed restorations and Micro-CT analysis, 2 groups were categorized in terms of marginal and internal fit, and 4 subgroups in terms of cementation and scanner. A meta-analysis was conducted across multiple studies using a random-effects model (p < 0.05). Expected heterogeneity was assessed among the studies and subgroup analysis was also performed.

Results

41 articles were included for marginal, and 39 articles for internal fit. All included studies were assessed as having a low risk of bias. The mean marginal fit was 105.66 μm (95% CI: 91.55–119.78), and the overall mean internal fit was 138.57 μm (95% CI: 122.56–154.58). High heterogeneity was seen in both main groups. In the cemented subgroup, the mean marginal fit was 137.67 μm, and the mean internal fit was 181.87 μm. In the non-cemented subgroup, the mean marginal fit was 103.19 μm, and the mean internal fit was 131.67 μm. In the EOS subgroup, the marginal fit was 102.79 μm, and the internal fit was 132.04 μm. In IOS, marginal and internal fit were 109.83 μm and 148.02 μm, respectively.

Conclusions

Within the limitations of this meta-analysis, pooled marginal and internal fit values varied across studies despite predefined cement space during the design stage. Subgroup analyses suggested that cementation status and scanner type may influence the reported fit values. However, these findings should be interpreted cautiously due to substantial heterogeneity and variability in study conditions.

Clinical Implications

In CAD/CAM restorations with predefined cement space, clinicians should be aware that reported marginal and internal fit values may vary depending on material, workflow, and measurement conditions. Therefore, clinical interpretation should be made cautiously, particularly for internal fit, where universally accepted thresholds remain unclear.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12903-026-08541-x.

Introduction

In recent years, computer-aided design and manufacturing (CAD/CAM) systems have been widely used in the production of prosthetic restorations. In CAD/CAM systems, almost all processes from tooth preparation to cementation can be performed rapidly, with high precision and without human intervention, offering advantages over traditional workflows [1]. However, similar to conventional restorations, the long-term biomechanical and clinical success of CAD/CAM restorations largely depends on their marginal and internal adaptation [2]. Internal and marginal adaptation is influenced by many factors. These include preparation, finish line, impression method, design, selected material, restoration type, cement gap, intraoral adjustment procedures, and cementation [2].

A distinct advantage of CAD/CAM workflows is the ability to digitally define the cement within the known optimal ranges [3]. The cement gap is a pre-determined virtual space designed to allow appropriate seating of the restoration and adequate cement flow [4]. In CAD/CAM systems, restorations are designed with a specific cement gap value in the digital environment, but it remains uncertain how consistently this parameter is reflected in the final restoration. Studies have reported that the internal and marginal fit of restorations may be affected by factors such as the type of material used, the accuracy of the milling or printing system, and sintering shrinkage [4–6]. Differences in scanner resolution and digitization methods may also contribute to variations in fit [5], as has been demonstrated in both full-arch and quadrant restorations [6]. While several fabrication variables are known to influence fit, the relationship between the predefined cement space and the final fit remains incompletely understood. Internal and marginal adaptation have been evaluated in multiple micro-computed tomography (micro-CT) studies using polymer-infiltrated ceramics [7, 8], modified endocrowns [9], and implant-supported resin restorations manufactured via 3D printing [10]. Restoration design, fabrication technique, and the selected CAD software may all impact the final adaptation [11–13].

Dental restorations’ marginal and internal fit is a critical factor influencing their long-term success, clinical performance, and patient outcomes. Precise adaptation of restorations helps minimize cement dissolution, microleakage, secondary caries, and mechanical complications. However, the reported fit values vary considerably across studies due to differences in materials, fabrication techniques, and measurement methods. Micro-CT has become the gold standard in measuring marginal and internal fit due to its high-resolution, non-destructive capabilities [14, 15]. Its ability to provide precise three-dimensional evaluation of marginal and internal fit enhances measurement reliability and comparability across studies. This method has been used to assess adaptation in various materials, including feldspathic ceramics, lithium disilicate, and high-performance polymers [16–19]. The finish line design and material properties have also been shown to significantly affect adaptation in both provisional and definitive restorations [20, 21].

Even within the same CAD/CAM system, adaptation can vary depending on the scanner, material, and restoration type [22, 23]. Virtual cement space and restorative material selection jointly influence the precision of endocrowns [24], with significant differences observed between materials such as lithium disilicate and PEEK [25]. Milled and pressed restorations exhibit adaptation variation due to differences in CAM workflows [26], while zirconia copings display system-dependent discrepancies [27].

Comparative studies show that the choice of measurement technique (e.g., replica technique vs. micro-CT) affects perceived fit accuracy [28, 29], and even within micro-CT, two-dimensional vs. three-dimensional assessments can yield different results [30]. Evaluation methods significantly influence comparative studies between zirconia and cobalt-chromium prostheses [31]. CAD software type [32], and restorative system [33, 34] are all known to affect fit. Provisional materials also present fit variation across systems and materials [35], and digital cement space settings directly impact occlusal veneer adaptation [36, 37]. The selection of the CAM system further influences the internal and marginal accuracy of Y-TZP copings [38]. Measurement technique remains a major factor in determining fit quality, with different protocols yielding divergent outcomes [39]. Scanner resolution, powdering, and scanning strategy also modulate final adaptation [40].

Comparisons between digital and conventional impressions for lithium disilicate restorations continue to yield conflicting data on fit accuracy [41], and studies have even shown that operator magnification (e.g., loupes vs. microscope) during tooth preparation can significantly affect the final marginal fit [42]. Scanning time and system choice also influence the efficiency and accuracy of intraoral scans [43].

Measurement methodology affects interim restorations [44], and laminate veneer studies highlight the effect of material choice and measurement sensitivity [45]. The historically accepted marginal fit threshold of 120 μm was established by McLean and von Fraunhofer [46], though contemporary reviews continue to debate what constitutes “clinically acceptable” fit [47]. CAD/CAM fabrication technique and restorative material interact to influence adaptation [48], and increasing the cement space can sometimes improve fit in monolithic zirconia restorations—though only up to a point [49]. Finally, meta-analyses comparing digital and conventional workflows reinforce the importance of the cement space parameter as a determinant of fit and clinical performance [50].

While individual studies provide valuable insights, their findings may be limited by small sample sizes, variations in study design, and methodological inconsistencies. A meta-analysis was conducted to synthesize data from multiple studies evaluating the marginal and internal fit of dental restorations, addressing these limitations and providing a more comprehensive understanding. Given the expected variability across studies (due to differences in restorative materials, digital and conventional fabrication methods, and measurement techniques), a random-effects model was selected. This approach accounts for both within-study and between-study variability, ensuring more generalizable conclusions. Given the variability across individual studies, this meta-analysis aimed to synthesize Micro-CT-based evidence on the marginal and internal fit of CAD/CAM restorations in which the cement space was predefined during the design stage. By pooling quantitative fit values, this study seeks to provide a clearer understanding of the reported adaptation outcomes under standardized measurement conditions. In digital workflows, CAD software enables precise numerical definition of cement space during design phase; however, the final restoration fit may still vary due to manufacturing and material-related factors. This discrepancy between digital design parameters and final restoration fit presents a clinically relevant aspect of digital workflow accuracy. It should be emphasized that this meta-analysis synthesizes reported values from studies with predefined cement space, rather than directly quantifying changes between predefined cement space values and final measured gaps.

Materials and methods

This meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (a completed PRISMA 2020 checklist is provided as Supplementary Material), and the PICOS question was “What are the reported marginal and internal fit values of CAD/CAM restorations with predefined cement space assessed using Micro-CT ”? Accordingly, the present meta-analysis focuses on reported fit measurements rather than direct comparisons between predefined cement space values and final measured gaps.

• P (population): Fixed prosthetic restorations produced with CAD/CAM

• I (intervention): Studies in which the cement gap is determined at the design stage, and the value is given in numbers

• C (comparison): Comparison of reported marginal and internal fit values across studies with predefined cement space under different experimental conditions (i.e., scanner type, cementation status).

• O (outcome): marginal and internal fit

• S (study design): Studies where only micro-CT is used for fit analysis.

The PICOS question above was used for the search strategy. The search strategy was developed to capture studies reporting marginal and internal fit values in CAD/CAM restorations with predefined cement space, and was adapted for each database accordingly. The search strategy included combinations of keywords and Boolean operators such as (‘CAD/CAM’ OR ‘computer-aided design’) AND (‘cement gap’ OR ‘cement space’) AND (‘marginal fit’ OR ‘internal fit’) AND (‘micro-CT’ OR ‘micro computed tomography’). English-language literature indexed in PubMed, Scopus, ScienceDirect, and Google Scholar was searched by 2 independent researchers (ET, BKY). Afterward, the manual search was completed by consulting the references of the selected studies. Between 2013 and 2024, studies published in peer-reviewed journals within the scope of the Science Citation Index were searched. After the preliminary search, duplicate studies were removed, and the remaining studies were sorted by eligibility. The inclusion and exclusion criteria are listed in Table 1. Forty studies [4–44] that met the criteria were included (Fig. 1). Since two different types of restorations (inlays and all-crowns) were examined separately in one study [22], the results of that study were evaluated as 2 separate studies. The full electronic search strategy for at least one database is provided in the Supplementary Material.

Table 1.

Inclusion and exclusion criteria

Inclusion Criteria	Exclusion Criteria
In vitro studies published in peer-reviewed journals indexed in SCI, with full-text accessible	Abstracts, case reports, case series, conference proceedings, or in vitro or in vivo studies published in peer-reviewed journals not indexed in SCI
Studies in English	Studies in languages ​​other than English
Studies published between 2013–2024	Studies published before 2013
Studies using digital impression and CAD/CAM production systems	Studies where traditional impression and production systems are used, but no digital workflow is used
Studies using micro-CT for fit analysis	Studies using methods other than micro-CT (silicon replica, cross-sectional method, etc.) for fit analysis
Studies that evaluate internal and marginal gap values ​​micrometrical and provide mean, standard deviation, or median	Studies that evaluate internal and marginal gap values ​​qualitatively or do not provide micrometric results
Studies in which the cement gap is determined at the design stage and this is clearly stated numerically	Studies in which the cement gap was not determined at the design stage, or if it was determined, it was not given numerically

Fig. 1.

Preferred reporting items for meta- analysis flow diagram of search strategy and study selection

The risk of bias in the in vitro studies included in the study was evaluated by 2 independent researchers (ET, BKY) using the criteria from a previously published review [45]. Questions 4 and 5 were revised in relation to our own meta-analysis. A “YES” answer was coded to indicate that the question asked in the publication was reported, and a “NO” answer was coded to indicate that there was no information about that question in the study. If the number of yes answers is 1–4, it indicates high risk; 5–7, medium risk; and 8–9, low risk. In the risk analysis of the 10 questions in Table 2, the total number of yes responses was counted. Each criterion was scored as Yes, No, or Unclear. As the yes answers increased, the risk decreased. The answers to the first 2 questions were no, and the remaining 8 were yes, indicating a low risk of bias for all included studies (Table 2). Although all included studied were categorized as low risk based on the applied criteria, it should be acknowledged that in vitro study designs inherently involve methodological variability, which may not be fully captured by the scoring system used.

Table 2.

Evaluation questions in terms of quality analysis and risk of bias

Questions	Answers
1) Teeth Randomization
2) Blinding of the Operator 3) Control Group
4) Has the cement gap been determined and its value clearly reported? 5) Number of the Measurement Points (≥ 4) for per Specimen? 6) Was the Outcome Accurately Measured to Minimize Bias? 7) Statistical Analysis Carried Out 8) Gap Measurements by a Single Operator 9) Luting Space Parameter 10) Was the Study Conducted in an Acceptable Way?	Yes, no, unclear or not applicable

Data extraction was performed independently by two researchers. The following variables were collected from each study: author and year, material type, restoration type, supporting structure, sample size, predefined cement space, scanner type, cementation status, and reported marginal and/or internal fit values (mean ± standard deviation or median where applicable). In studies reporting multiple restoration types or conditions, each dataset was treated as an independent entry. Disagreements were resolved through discussion.

A meta-analysis was conducted to examine the marginal and internal fit of dental restorations across multiple studies using a random-effects model. The random-effects model was selected due to expected heterogeneity across studies, as different materials, techniques, and outcome measures were used. The primary outcomes were the mean marginal fit (in micrometers) and mean internal fit (in micrometers). Subgroup analyses were planned by the scanner type [intraoral (IOS) or extraoral (EOS) scanners] and fit measurements with or without cementation. Studies [11, 22, 41] in which both IOS and EOS were used together were excluded from the scanner subgroup analysis (Table 3). The statistical significance of the results was determined using a two-tailed p-value threshold of 0.05. Heterogeneity was assessed using the I² statistic, and sensitivity analyses were conducted to assess the robustness of the results. Publication bias was not formally assessed due to the expected heterogeneity and variability of the in vitro study designs.

Table 3.

Characteristics of included studies

Author and Year	Material	Sample size	Type of support tooth	Restoration type	Points Measured per Specimen	Adjusted cement space	Scanner type	Fit analysis performed cemented or not cemented
Prudenta et al. 2018	Lithium disilicate block	n:10	Extracted tooth	Full crown	6	80 μm	IOS	Not cemented
Kim et al. 2016	PMMA Resin block Lithium disilicate	n:5	Typodont	Full crown	2	60 μm	Both of IOS and EOS	Cemented
Nagi et al. 2023	PEEK, Lithium disilicate block	n:8	Extracted tooth	Full crown	16	50 μm	IOS	Not cemented
Attia et al. 2023	Milled and pressed PEEK	n:10	Stainless- steel die	Full crown	12	30 μm	EOS	Not cemented
Silveira et al. 2017	Lithium disilicate block Composite resin block	n:10	Extracted tooth	Full crown	12	80 μm	IOS	Not cemented
Tabata et al. 2020	Lithium disilicate block Resin infiltrated glass ceramic	n:16	Extracted tooth	Full crown	12	80 μm	EOS	Not cemented
Rizonaki et al. 2022	Lithium disilicate block	n:10	Typodont	Full crown	9	80 μm	IOS	Cemented
Topkara et al. 2022	Feldspathic ceramic block	n:10	Extracted tooth	Partial crown	9	120 μm	IOS	Cemented
Tamam et al. 2023	Feldspathic ceramic block	n:7	Typodont	Full crown	7	90 μm	EOS	Not cemented
Rezende et al. 2017	High-strength zirconia block	n:12	Zirconia die	Full crown	8	80 μm	EOS	Not cemented
Abduljavad et al. 2022	Lithium disilicate block	n:10	Extracted tooth	Partial crown	11	Marginal: 0 Internal: 80 μm	Both of IOS and EOS	Not cemented
Akat et al. 2022	Feldspathic Ceramic block	n:11	Typodont	Full crown	9	Marginal:80 μm Internal: 120 μm	IOS	Not cemented
Alharbi et al. 2018	Hybrid Composite Resin Material	n:40	Typodont	Full crown	51	Marginal: 60 μm Internal : 100 μm	EOS	Not cemented
Aldahian et al. 2021	Cercon base PMMA block	n:10	Typodont	Full crown	12	100 μm	EOS	Not cemented
Atlas et al., 2024	High- strength glass-ceramic Zirconia Reinforced Lithium Silicate block	n:18	Extracted tooth	Full crown	24	25 μm	IOS	Cemented
Bandiaky et al. 2023	High- strength zirconia- based material	n:9	Typodont	Five- unit Zirconia based FDP’s	756	50 μm	EOS	Not cemented
Borba et al. 2013	Yttria-partially-stabilized tetragonal zirconia polycristals, Glass-infiltrated zirconiareinforced Alumina-based ceramic block	n:10	Stainless- steel die	Three- unit FDPs	18	20 μm	EOS	Not cemented
ArRejaie et al. 2018	Zirconia	n = 10	Stainless- steel die	Full crown	11	30 μm	EOS	Not cemented
Bayrak et al. 2021	Felshpathic ceramic block	n:11	Typodont	Partial crown	11	Marginal : 25 μm Internal: 80 μm	IOS	Not cemented
Dauti et al. 2018	Polymer infiltrated ceramic block	n:10	Typodont	Full crown	4	Marginal: 0 μm Internal gap: 50 μm	EOS	Cemented
Dauti et al. 2020	Polymer infiltrated ceramic block	n:50	Typodont	Full crown	4	Marginal: 0 μm Internal gap: 80 μm	IOS	Not cemented
Freitas et al. 2020	Lithium disilicate block	n:8	Typodont	Full crown	9	40 μm	EOS	Not cemented
Daou et al. 2024	Polychromatic, süper-high translucency zirconia, Co- Cr sintered metal	n:12	Typodont	Three-units-FDPs	18	30 μm	EOS	Not cemented
Elbadawy et al. 2021	Lithium disilicate block	n:30	Typodont	Partial crown	9	50 μm	EOS	Not cemented
Elbadawy et al. 2022	Hybrid all ceramic, Zirconia reinforced lithium silicate glass ceramic	n:30	Typodont	Partial crown	9	50 μm	EOS	Not cemented
Ferrairo et al. 2021	Monolithic lithium disilicate block	n:10	Typodont	Full crown	84	50 μm	EOS	Not cemented
Hassouneh et al. 2023	Resin matrix ceramic, Partially crystallized lithium disilicate glass ceramic, Fully crystallized LDGC	n:10	Extracted tooth	Partial crown	8	80 μm	IOS	Not cemented
Son et al. 2019	Pre- sintered yttria stabilized tetragonal zirconia polycrystalline block	n:40	Extracted tooth	Full crown	10	40 μm	EOS	Not cemented
Park et al. 2019	PMMA Resin block	n:10	Plastic supports simulating implant abutments	Three- unit implant supported prothesis	11	0 μm	EOS	Not cemented
Oğuz et al. 2021	Feldspathic ceramic block	n:12	Typodont	Full crown	9	100 μm	EOS	Not cemented
Peng et al. 2020	Bis- acrylic composite resin, CAD/CAM PMMA Resin, 3D Printed methacrylic olygomers	n:12	Typodont	Full crown	9	75 μm	EOS	Cemented
Toma et al. 2023	Hıgh performance polymer (PEEK and PEKK)	n:8	Typodont	Partial crown	13	30 μm	EOS	Cemented
Saab et al. 2018	Yttrium oxide partially stabilized tetragonal zirconia polycrystal ceramic block	n:10	Stainless- steel die	Full crown (coping)	4	Marginal: 30 μm İnternal: 70 μm	IOS	Not cemented
Pereira et al. 2024	Lithium disilicate (LS2) reinforced glass ceramic CAD/CAM block	n:10	Typodont	Full crown	7	80 μm	IOS	Not cemented
Zheng et al. 2022	Pre- sintered zirconia- reinforced lithium silicate glass ceramic, Fully- sintered zirconia- reinforced lithium silicate glass ceramic, Resin composite	n:10	Typodont	Partial crown	58	120 μm	IOS	Not cemented
Alajaji et al. 2017	Lithium disilicate block	n:15	Typodont	Partial crown	120	Marginal: 25 μm Internal: 100 μm	IOS	Not cemented
Yildirim et al. 2017	Hybrid-ceramic and resin nanoceramic block	n:12	Typodont	Full crown	160	40 μm	IOS	Not cemented
Cunali et al. 2017	Polycrystallşne zirconia copings	n:10	Metal die	Full crown	4	Marginal: 20 μm İnternal: 70 μm	EOS	Not cemented
Jalalian et al. 2023	PMMA Block	n:7	Metal die	Full crown	4	30 μm	IOS	Not cemented
Ekici et al. 2021	Feldspathic ceramic block	n:12	Typodont	Full crown	18 for crown, 7 for inlay	All ceramic crowns: 100 μm, Inlay ceramic crown: 120 μm	BOTH OF EOS and IOS	Not cemented
Vag et al. 2020	Lithium disilicate block	n:8	Extracted tooth (Cadaver)	Full crown	10	70 μm	IOS	Not cemented

Results

The results are presented as pooled mean values (µm) rather than standardized effect sizes. Two separate meta-analyses were performed, each analyzing the mean fit of dental restorations. The first meta-analysis (Fig. 2), based on 41 datasets from 40 studies [4–43], revealed significant heterogeneity across the studies (I² =94.19%, tau² =1355.99). The estimated mean marginal fit was 105.66 μm (95% CI: 91.55–119.78), with a highly significant p-value (< 0.0001). The tau² value suggests a substantial amount of variability in the pooled mean values across the studies. The Q-test (Q(df = 40) = 826.93, p < 0.0001) indicated strong evidence of heterogeneity, confirming that the variability among studies is greater than would be expected by chance alone.

Fig. 2.

Overall forest plot of pooled mean marginal fit values (µm) with 95% confidence intervals

The second meta-analysis (Fig. 3), based on 37 datasets from 36 studies [4–9, 11–39, 44], also demonstrated considerable heterogeneity (I² =94.09%, tau² =1547.52), with an overall mean internal fit estimate of 138.57 μm (95% CI: 122.56–154.58). Similar to the first analysis, this result was highly significant (p < 0.0001), indicating that the effect is robust across studies. The Q-test (Q(df = 36) = 1172.31, p < 0.0001) again confirmed the presence of substantial heterogeneity between studies. The tau² value further emphasized the considerable variability in the pooled mean values.

Fig. 3.

Overall forest plot of pooled mean internal fit values (µm) with 95% confidence intervals

In both analyses, the I² values (94.19% and 94.09%) were extremely high, suggesting that the majority of the observed variability in pooled mean values was due to heterogeneity rather than sampling error. The H² values (17.21 and 16.93) further indicated that the variability between studies was much larger than what could be attributed to sampling error. Given the high degree of heterogeneity, we focused on factors contributing to the observed variability, such as whether or not cementation is performed before fit analysis and scanner type (Fig. 4), and conducted subgroup analyses.

Fig. 4.

Schematic representation of subgroup classification

Subgroup analyses by cementation (cemented or non-cemented conditions)

Marginal fit/ cemented subgroup

A random-effects meta-analysis was conducted to assess the marginal fit of dental restorations across six studies (k = 6) with fit analysis with cemented samples (Fig. 5). The estimated total heterogeneity was substantial (tau² = 2973.86, SE = 3204.38), with a corresponding tau value of 54.53. The I² statistic indicated that 72.92% of the total variability across studies was attributable to between-study heterogeneity. The H² statistic was 3.69, further supporting the presence of significant heterogeneity. The test for heterogeneity was significant, Q(5) = 14.84, p = 0.0111, suggesting substantial variation among the studies. The overall pooled estimate was 137.67 μm (SE = 30.44), and this result was highly significant, z = 4.52, p < 0.0001. The 95% confidence interval for the pooled estimate ranged from 78.00 to 197.33, indicating a robust effect.

Fig. 5.

Forest plot of mean marginal fit/ Cementation subgroups

Marginal fit/ non-cemented subgroup

A random-effects meta-analysis was conducted to examine the subgroup effects among studies (k = 35) without cemented samples (Fig. 5). The model’s tau² estimator (REML) was calculated to be 1331.46, with a standard error of 446.18, indicating a high level of heterogeneity across the studies (tau = 36.49). The I² value was 94.66%, suggesting that 94.66% of the total variability was due to heterogeneity rather than sampling error. The H² value was 18.72, indicating a large amount of total variability relative to sampling variability. The test for heterogeneity was significant, Q(df = 34) = 796.85, p < 0.0001, confirming substantial variability across studies. The overall effect estimate was 103.19 μm (SE = 7.52), which was statistically significant, z = 13.71, p < 0.0001. The 95% confidence interval for the effect estimate ranged from 88.44 to 117.93.

Internal fit/ cemented subgroup

A random-effects meta-analysis was conducted to examine the pooled mean value for internal fit across five studies (k = 5) with fit analysis with cemented samples (Fig. 6). The estimated total heterogeneity was tau² =823.66 (SE = 1130.59), and the square root of the estimated heterogeneity was tau = 28.70. The I² statistic indicated that 59.34% of the total variability was due to heterogeneity, and the H² statistic suggested that the total variability was 2.46 times greater than the sampling variability. A test for heterogeneity was significant, Q(df = 4) = 9.62, p = 0.0473. The model results showed a significant effect with an estimated pooled mean value of 181.87 μm (SE = 18.48), z = 9.84, p < 0.0001, and a 95% confidence interval for the pooled mean value ranging from 145.65 to 218.09.

Fig. 6.

Forest plot of mean internal fit/ Cementation subgroups

Internal fit/ non-cemented subgroup

A random-effects model was used to analyze the data from 32 studies (k = 32) without cemented samples to assess the internal fit of dental restorations (Fig. 6). The model estimated total heterogeneity as tau² =1398.05 (SE = 521.40), with a square root of tau² (tau) of 37.39. The I² statistic indicated substantial heterogeneity (I² =93.55%), and the H² statistic suggested that the total variability was 15.5 times greater than the sampling variability. The heterogeneity test revealed significant variability across studies, Q(df = 31) = 1142.9983, p < 0.0001. The model estimate for the outcome measure was 131.67 μm (SE = 8.40), with a z-value of 15.68, indicating a highly significant effect (p < 0.0001). The 95% confidence interval for the estimate ranged from 115.21 to 148.12.

Subgroup analyses by scanner type (IOS or EOS)

Marginal fit/ EOS subgroup

A random-effects model was used to conduct the subgroup meta-analysis, with 20 studies (k = 20) with extraoral scanner type included (Fig. 7). The estimated total heterogeneity (tau²) was 2571.62 (SE = 1031.77), with a square root of tau² (tau) equal to 50.71. The total heterogeneity as a percentage of total variability (I²) was 95.71%, indicating high variability across studies. The H² statistic, which reflects the ratio of total variability to sampling variability, was 23.31. The test for heterogeneity was significant, Q(df = 19) = 248.52, p < 0.0001. The estimated pooled mean value was 102.79 μm (SE = 12.92), with a z-value of 7.95, which was statistically significant (p < 0.0001). The 95% confidence interval for the pooled mean value ranged from 77.46 to 128.12.

Fig. 7.

Forest plot of mean marginal fit/ Scanner subgroups

Marginal fit/ IOS subgroup

A random-effects model was used to analyze the data from 17 studies (k = 17) with intraoral scanner type (Fig. 7). The model estimated a total heterogeneity (tau²) of 369.51, with a standard error of 273.33. The square root of the heterogeneity (tau) was 19.22. The total variability, as measured by the I² statistic, was 63.70%, indicating moderate to high heterogeneity, and the H² value was 2.76, suggesting that the variability in the studies was approximately 2.76 times greater than what would be expected due to sampling error alone. The test for heterogeneity was statistically significant (Q(df = 16) = 52.21, p < 0.0001). The model estimate for the overall effect was 109.83 μm (SE = 7.26), with a z-value of 15.12 and a p-value less than 0.0001. The 95% confidence interval for the pooled estimate ranged from 95.59 to 124.06, indicating a significant effect of the variable being analyzed.

Internal fit/ EOS subgroup

A random-effects model (k = 20) was used to analyze the internal fit of dental restorations (Fig. 8), with heterogeneity assessed using the estimated tau² of 1285.25 (SE = 615.07). The model showed a high level of heterogeneity, with an I² value of 89.09%, indicating substantial variability between studies. The Q-test for heterogeneity was significant, Q(df = 19) = 117.54, p < 0.0001. The estimated overall pooled mean valuewas 132.04 μm (SE = 10.22), z = 12.92, p < 0.0001, with a 95% confidence interval of [112.01, 152.07]. These results suggest a significant positive effect across the studies, with substantial heterogeneity.

Fig. 8.

Forest plot of mean internal fit/ Scanner subgroups

Internal fit/ IOS subgroup

A random-effects meta-analysis was conducted across 15 (k = 15) studies (Fig. 8). The results of the model indicated significant heterogeneity among the studies, with an estimated total heterogeneity of tau² =1853.63 (SE = 973.44). The I² statistic was 91.86%, suggesting substantial variability in the pooled mean values beyond sampling error. The H² value of 12.29 further indicated considerable total variability. The test for heterogeneity was highly significant, Q(df = 14) = 187.15, p < 0.0001, confirming the presence of substantial variability across studies. The model estimate for the effect was 148.02 μm (SE = 13.50), with a z-value of 10.97 (p < 0.0001). The 95% confidence interval for the pooled estimate ranged from 121.56 to 174.48, further supporting a robust and statistically significant effect.

Discussion

In this meta-analysis, marginal and internal fit of CAD/CAM restorations with predefined cement space were evaluated based on studies in which measurements were performed using Micro-CT. It should be noted that the present analysis does not directly quantify the deviation between the predefined cement space and the final measured values, due to inconsistencies in reporting across studies. Therefore, the findings should be interpreted as pooled fit measurements rather than direct deviations from initial design parameters.

41 studies were included in this meta-analysis (Table 3). Full crowns were examined in 26 of them [4–8, 12, 14, 15, 17, 18, 20, 23, 25–27, 30, 32, 33, 35, 38–44], partial crowns in 9 [9, 11, 13, 16, 19, 21, 24, 36, 37], both full and partial crowns in 1 [22], and bridges in 4 of them [10, 29, 31, 34]. In 23 of the studies [4, 5, 7, 8, 12, 13, 15, 16, 18–20, 22, 24, 29–33, 36, 37, 41, 43, 44], typodonts were used as support teeth, extracted human teeth were used in 9 studies [9, 11, 14, 17, 21, 25, 39, 40, 42], metal dies in 6 studies [26–28, 34, 35, 38], and in 2 studies, cadaver teeth [6], zirconia die [23] were used, respectively. In 1 study [10], plastic supports simulating implant abutments were produced. In these studies, the mean cement gap determined at the design stage is 67.8 μm (between 0 and 120 μm), (Table 3).

According to classical knowledge, the ideal marginal fit is 120 μm [46]. A recent systematic review reported that this threshold has become smaller (shown to be less than 100 μm, 90 μm, and even 75 μm), the marginal fit varies between 7.5 μm and 206.3 μm, and there is no consensus on the ideal amount [47]. The clinical acceptability may vary depending on restoration type, material and clinical conditions [46, 47]. For internal fit, no universally accepted clinical threshold exists, which limits direct clinical interpretation of the pooled values. In the current meta-analysis, the mean marginal fit after manufacturing was found to be 105.66 μm (95% CI: 91.55–119.78). This value is higher than that reported in approximately 90% of the included studies. As for internal fit, this value, with an average of 138.57 μm (95% CI: 122.56–154.58), is higher than all 39 studies included. These findings are consistent with previous systematic reviews reporting wide variability in marginal and internal fit values across different CAD/CAM systems and materials, further supporting the need for cautious clinical interpretation [47, 50].

Due to the high heterogeneity seen in the meta-analysis of the main groups (I²= over 94% for both marginal and internal fit), subgroup analyses were planned. This heterogeneity is likely attributable to restoration types, restorative materials, and different techniques used. Variations in predefined cement space values, scanner types, and cementation conditions may have further contributed to the observed variability [3–6]. As can be seen in Table 3, there is high diversity in terms of restoration type and material in studies that meet the inclusion criteria. At this point, it may be questioned how the subgroups were determined or why a subgrouping is not made according to restoration type or restorative materials. However, it is difficult to identify studies conducted under identical conditions. Therefore, a more realistic subgrouping question was tried to be asked, based on the inclusion criteria of this meta-analysis. Since digital impression, virtual design, CAD/CAM manufacturing, and fit analysis with Micro-CT were performed in all included studies, an answer to the question “What could be different between studies?” was sought in determining the subgroups. And so, the section from design to fit analysis (Fig. 4) was considered similar (almost like a constant). Thus, the stage before design (type of digital impression) and the stage before fit measurement (analysis of whether cemented or not) were selected for subgroups. Additionally, differences in manufacturing techniques (i.e. milling versus additive manufacturing), sintering protocols, and variations in measurement protocols within Micro-CT analyses may also have contributed to the observed heterogeneity. These variables were selected because they represent clinically and technically distinct stages in the digital workflow that may directly influence the final fit outcomes.

In the first subgroup, it was questioned whether the cementation of the samples to the supporting tooth before the fit analysis had an impact on the final result and heterogeneity.

Micro-CT analysis was performed under the conditions where the samples were cemented in 6 studies (Fig. 5) for marginal fit and 5 studies (Fig. 6) for internal fit (Table 3). In the cemented subgroup, the mean marginal fit is 137.67 μm, and the mean internal fit is 181.87 μm. Heterogeneity (I²) decreased to 72.92% and 59.34%, respectively. In the non-cemented subgroup consisting of the remaining studies, the mean marginal fit is 103.19 μm, and the mean internal fit is 131.67 μm. Heterogeneity (I²) remained unchanged (approximately 94%) for both marginal and internal fit. According to the statistically significant subgroup results, marginal and internal fit were within or close to commonly reported clinically acceptable threshold in the uncemented fit analyses and above the cemented ones.

The other subgroup was created according to which type of scanner (IOS or EOS) the digital impression was taken (Figs. 7 and 8). IOS was used in 16 studies, EOS was used in 21 studies, and both were used in 3 studies (Table 3). In studies using EOS, the mean marginal fit was found to be 102.79 μm and the internal fit was 132.04 μm. In IOS, they were 109.83 μm and 148.02 μm, respectively. While heterogeneity (I²) did not change in studies using EOS, it decreased noticeably for marginal fit (I²=63.70%) in studies conducted with IOS. This finding may be associated with differences in scanning approaches; however, no direct causal inference can be made. According to the statistically significant subgroup results, the marginal fit is within or close to commonly reported clinically acceptable limits for both IOS and EOS, while the internal fit is above it for both types of scanners.

Although the included studies defined cement space numerically during the CAD design stage, the pooled Micro-CT measurements revealed variability in the final marginal and internal fit values. This suggests that multiple factors beyond the initial digital settings—such as material properties, manufacturing processes, and measurement protocols—may influence the final adaptation. However, a direct quantitative comparison between the predefined cement space and the final measured values could not be performed within the framework of this meta-analysis. From a clinical and digital workflow perspective, these findings highlight that predefined cement space alone may not fully predict final adaptation, and multiple factors throughout the digital manufacturing process should be considered.

At this point, it is necessary to underline some notable points in fit analysis. It is very important to report the initial value of the cement gap, especially when performing fit analyses with digital systems. Thus, it becomes possible to determine exactly how much change the tested parameter causes and can be useful in improving digital systems. Kale et al. [49] reported that the amount of cement gap affects the marginal fit of monolithic zirconia crowns, and as the amount of initial cement gap increases, the marginal fit decreases.

Including only micro-CT studies in this meta-analysis is a limitation in that it cannot compare fit measurements made with other methods. Additionally, differences between restoration types or restorative materials may have contributed to heterogeneity. Although CAD/CAM systems were used in all included studies, differences between production systems are also limitations. For example, it is known that 5-axle milling provides a better fit than 4-axle milling [50]. Even changes in the materials of the model supporting teeth analyzed in the studies may also have contributed to heterogeneity.

Conclusions

This meta-analysis provides a comprehensive evaluation of the marginal and internal fit of dental restorations, highlighting the significant variability across studies. Both the marginal and internal fit analyses demonstrated substantial heterogeneity, with I² values consistently above 90%, indicating that most of the observed variability was due to differences between studies rather than sampling error. Despite this heterogeneity, the overall pooled mean values for both marginal and internal fit were highly significant, supporting the consistency of the pooled estimates across studies despite heterogeneity.

Subgroup analyses revealed significant effects of cementation and scanner type on both marginal and internal fit. Specifically, fit analysis with cemented samples and extraoral scanners exhibited higher pooled mean values for both marginal and internal fit compared to those with non-cemented samples and intraoral scanners. These findings suggest that both material and technical factors, such as the type of scanner used and the cementation, contribute to the observed variability in fit.

The high degree of heterogeneity observed emphasizes the importance of considering study-level factors when interpreting the results of meta-analyses in the field of dental restorations. The results reflect pooled fit values rather than quantified deviations from the predefined cement space. Future research should aim to explore these sources of variability more thoroughly and standardize measurement techniques to improve the consistency of reported outcomes across studies.

Authors' contributions

B.K. YILMAZ: Collected data, and wrote the manuscript. T.A. HOCAGİL: Performed the statistical analysis. E. TAMAM: Planned the work, interpreted the data, collected the sources, wrote the manuscript, and checked the whole study.

Funding

No funding was used for this study.

Data availability

All data supporting the findings of this study are available in the paper and references.

Declarations

Ethical approval and consent to participate

Not applicable.

Competing interests

The authors declare no competing interests.