Fit accuracy in the rest region of RPDs fabricated by digital technologies and conventional lost-wax casting: a systematic review and meta-analysis

Background Digital technologies have recently been introduced into the fabrication of removable partial dentures (RPDs). However, it is still unclear whether the digitally fabricated RPDs fit better than conventionally cast ones in the rest region. The aim of this study was to evaluate the fit accuracy in the rest region of RPDs fabricated by digital technologies and compare it to those made by conventional lost-wax (CLW) technique. Methods A comprehensive search was conducted in Cochrane Library, PubMed, EMbase, Web of Science and SpringerLink. Studies published up to August 2022 were collected. Two authors analyzed the studies independently and assessed the risk of bias on the modified methodological index for non-randomized studies (MINORS) scale. The mean values of gap distance between rests and corresponding rest seats of each study were extracted as outcome. A random-effects model at a significance level of P < 0.05 was used in the global comparison and subgroup analysis was carried out. Results Overall, 11 articles out of 1214 complied with the inclusion criteria and were selected, including 2 randomized controlled trials (RCTs), 1 non-randomized clinical trial and 8 in vitro studies. Quantitative data from Meta-analysis revealed that fit accuracy in the rest region of RPDs fabricated with CLW showed no statistically significant difference with digital techniques (SMD = 0.33, 95%CI (-0.18, 0.83), P = 0.21). Subgroup analysis revealed a significantly better fit accuracy of CLW-fabricated RPDs in the rest region than either additive manufacturing (AM) groups or indirect groups (P = 0.03, P = 0.00), in which wax or resin patterns are milled or printed before conventional casting. While milled RPDs fit significantly better than cast ones in the rest region (P = 0.00). With digital relief and heat treatment, hybrid manufactured (HM) clasps obtained better fit accuracy in the rest region (P < 0.05). In addition, finishing and polishing procedure had no significant influence in the fit accuracy in all groups (P = 0.83). Conclusions RPDs fabricated by digital technologies exhibit comparable fit accuracy in rest region with those made by CLW. Digital technologies may be a promising alternative to CLW for the fabrication of RPDs and additional studies are recommended to provide stronger evidence. Trial registration CRD42020201313. Supplementary Information The online version contains supplementary material available at 10.1186/s12903-023-03348-6.


Introduction
In traditional dental practice, removable partial dentures (RPDs) are fabricated by lost-wax casting [1].By this collaborative process between dentists and dental technicians, high-quality RPD frameworks can be produced.Nevertheless, it requires a great deal of experience and remains to be labor-intensive [2].In 1970s, computeraided design/computer-aided manufacturing (CAD/ CAM) was applied into dentistry by Duret and Preston [3].Since then, digital technology began its dental life.
A CAM system can be categorised into two types: subtractive manufacturing (SM) and additive manufacturing (AM).The most common SM technology used in dentistry is computer numerical controlled (CNC) milling.This method uses a milling machine to produce the object by removing bulk material from solid blocks with all the steps controlled by a computer program [4,5].While AM, also known as three-dimensional printing or rapid prototyping (RP), includes a range of different technologies such as stereolithography (SLA), selective laser melting (SLM), selective laser sintering (SLS), direct metal laser-sintering (DMLS), fused deposition modeling (FDM), selective electron beam melting (SEBM) and inkjet printing [4,5].SLA is basically used in the manufacture of resin-based structures, for instance temporary crowns, acrylic teeth, dentures, mouth guard and bite plane appliances through deposition of consecutive layers of photosensitive material that is readily polymerized [6,7].SLM, SLS and DMLS are laser powder forming techniques that use a high-energy laser beam to fuse material in its powder form and construct 3D objects layer by layer [6].When processing polymers and ceramic the industry generally refers to this as SLS whereas for metals the terms used are SLM or DMLS [5].FDM is a filament extrusion-based process that a plastic filament is heated to a semiliquid state and then extruded through a nozzle to deposit on to a platform to create 3D parts directly from a CAD model [5,8].SEBM is generally used for forming near-net shaped components of metals by melting metal powder layer per layer with an electron beam in a high vacuum [5,9].Inkjet printing works by propelling individual small ink drops toward a substrate and is capable of printing objects using two materials with quite distinctively different properties [5].
Milling(MI) manufacturing is superior in creating a smooth surface, while AM technologies overcome the limitations of subtractive methods, producing complex small shapes layer by layer directly from a computer model without limitations of the size of the smallest cutting tool [5,10].And thus hybrid manufacturing(HM) emerged, combining the advantages of the two techniques.Nakata et al. developed a one-process molding machine which integrated repeated laser sintering and MI into a single platform [11,12].Due to economic considerations, the indirect digital method consists of milling or printing wax/resin patterns that are then converted into cast-metal frameworks through conventional lost-wax technique (CLW) [13].All these technologies mentioned above are collectively referred to as digital technologies in this review.
It should be a primary quality of CAD/CAM systems that they can produce accurate fitting prosthetic components [14].The Aker's clasp commonly used in RPDs is composed of three parts: clasp arm, counter arm and the rest.Rests affords efficient resistance to functional chewing forces, which are transmitted vertically to the abutment teeth and conducted along the long axes of the teeth.To avoid independent movement or slippage of RPDs under occlusal loading, the rests and the teeth must remain in stable contact.Considering this, it is important for the rests to not only be rigid but also fit accurately to the rest seats [15].Stern et al. evaluated the adaptation between the occlusal rests and their corresponding rest seats in order to investigate the clinically acceptable in the fit accuracy of RPDs [16].Fit accuracy of digitally fabricated RPD rests have been evaluated and described in several studies [11][12][13][17][18][19][20][21][22][23][24], with inconsistent conclusions.In a study by Pelletier et al., frameworks made with SLS were less accurate at rest region than those produced with CLW [23], while Soltanzadeh et al. found that compared to 3D-printed groups, the cast RPD group showed better overall fit and accuracy [24].Therefore, it is still unknown whether the digital technologies could provide acceptable fit accuracy for the rests in RPDs.
The purpose of this study is to systematically review in vitro and clinical studies comparing the fit accuracy in the rest region of RPDs fabricated by digital technologies and conventional lost-wax technique.The null hypothesis was that no differences would be found between CLW and digital technologies.

Search methods
This systematic review was registered in the International Prospective Register of Systematic Reviews (PROS-PERO: CRD42020201313).A systematic approach was followed according to the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) Statement [25] and the Cochrane Handbook [26] No publication language restrictions were taken into account (Table 1).We used Medical Subject Headings (MeSH) terms and EMTREE, along with free-words to target the PICOS.In addition to the electronic search, relevant reviews and references lists of included fulltext articles were manually checked as well.

Screening and selection criteria
Studies that reported outcome data for both digitally and conventionally fabricated RPDs or clasp samples were included.All related studies with an English abstract were included in this review.For control group, only RPDs fabricated by CLW were included.RPDs fabricated by digital technologies, including AM, SM, HM and indirect digital technologies were included as intervention group.
The main outcome for this review was fit accuracy in the rest region, which was defined as the gap distance in micrometers between the rests and their corresponding rest seats.Randomized controlled trials (RCT), nonrandomized clinical studies and in vitro studies were included.Case reports, case series, expert opinions, commentaries, editorials, reviews, and conference abstracts were excluded.A study that was not accessible to read in full or was not available in the databases was also excluded (Table 2).
A reference manager software program (EndNote v.X9.3.1) was used and the duplicates were discarded electronically.The remaining articles derived from the extensive search were screened through title and abstract by two reviewers (JQ, DW) independently.The fulltext was checked if title and abstract provided insufficient information with regards to the inclusion criteria.Finally, articles selected from the inclusion and exclusion criteria were further screened in full-text and doublechecked by both reviewers (JQ, DW).Any disagreements at the above stages between reviewers were resolved by

Data extraction
Two authors (JQ and DW) conducted the data extraction as well as risk of bias assessments independently, and any disagreements were resolved through consensus.The following information was extracted: 1) Author and year of publication; 2) Study design; 3) Groups; 4) Tooth die or model type; 5) Sample type; 6) Method used for evaluating the fit accuracy of the rest); 7) Sample size; 8) Main outcomes; 9) Scanning information; 10) CAD software; 11) Manufacturing machine; 12) Finishing and polishing.We contacted the corresponding authors of individual studies for missing data or additional study information.And those with no respondence after three contact attempts were excluded from meta-analysis and included in the qualitative aspect of this review.For studies that reported the gap distance values before and after polishing of the samples, the data after polishing was selected for the global meta-analysis.And for studies that evaluated vertical and horizontal distances between the RPD rests and rest seats, only the horizontal data (distances between the bottom of rests and rest seats on the occlusal surface of the tooth) was extracted.

Critical appraisal
The Cochrane Risk of Bias Assessment Tool for Randomized Controlled Trials [26] were used with the software RevMan version 5.3 (The Cochrane Collaboration, Copenhagen, Denmark) to assess the risk of bias for two included RCTs [17,23].And to assess the risk of bias of other included in vitro experiments [11-13, 19-22, 24] and a non-randomized clinical study [18], we developed a modified version of Methodological Index for Non-Randomized Studies (MINORS) scale based on the original one [27].A total of 13 items were included in the adapted scale, with an additional item proposed for clinical studies (Table 3).The items are scored 0 (not reported), 1 (reported but inadequate) or 2 (reported and adequate) [27].Discrepancies of opinion during the assessment were resolved through discussion until a consensus was finally reached between the 2 reviewers (JQ and DW).
And finally the overall score was calculated.The ideal global score would be 24 for the in vitro studies and 26 for the clinical studies.

Statistical analysis
A software program StataMP17.0 was used for data processing and meta-analysis.The number of rests was considered as a statistical unit.Standardized mean difference (SMD) with 95% confidence interval (95% CI) was used to compare digital technologies and CLW fabricated RPDs on fit accuracy in the rest region.The Dersimonian-Laird method was used in the random effects model and the Inverse-variance method was used in the fixed effects model to account for differences between studies.Using Cochrane Q test and I 2 test (25-50% slight, 50-75% moderate, and > 75% high heterogeneity), heterogeneity among the pooled studies was tested [28,29].The P < 0.05 was considered statistically significant.According to different digital technologies adopted by each experimental group (AM, MI and indirect digital technologies), the included studies were assigned to three subgroups and subgroup analysis was conducted to investigate possible causes of heterogeneity among the results.The final results were presented by forest maps.And to assess robustness of the synthesised results, sensitivity analyses were conducted by excluding the remaining articles into the literature one by one, conducting meta-analysis again and comparing them with the overall results before exclusion.Potential publication bias among studies included in the meta-analysis were assessed and presented by Funnel plots.In order to reduce the risk of bias in our reference list and avoid any risk of autocitation read, the fi-index tool was used [30,31].For the research results that cannot be integrated, a comprehensive description and separate analysis was carried out.

Search and selection
The final electronic search identified 1214 database articles, 337 from Cochrane Library, 389 from PubMed, 214 from EMbase, 114 from Web of Science and 160 from SpringerLink.After removal of duplicates, 956 records remained, from which 905 were excluded through screening on the basis of titles and abstracts.And the remained 51 articles were read in full.Forty publications were further excluded as they did not meet inclusion criteria or lack of available data, leaving 11 as eligible studies for this systematic review.Details of the selection process are presented in Fig. 1.
Nine studies had RPD frameworks as the unit of analysis, while two studies used Akers clasp assemblies [11,12].Regarding the fabrication materials, cobalt-chromium (Co-Cr) was most commonly used, while one study made one-piece RPDs from polyetheretherketone (PEEK) [19] and another one cast clasp samples of CP titanium Grade 3 [11].
All studies evaluated the horizontal gap between the rests and rest seats, while one addressed the gap distance from both horizontal and vertical dimensions [13].There's no standard method for the quantitative measurement of fit accuracy of RPD rests.Among the included studies, 6 used the silicone film method and 2 applied silicone film method combined with digital superimposition to evaluate the fit accuracy [19,22].One study scanned the intaglio surface of each RPD framework and superimposed the STL file onto that of its master model [24] and another 1 made the measurement directly under a light microscopy at × 560 magnification [13].In addition, clinical observation including visual inspection and pressing test were also conducted by three studies [17][18][19].Detailed information of individual studies is presented in Tables 4, 5 and 6.

Risk of bias in studies
As was shown in Fig. 2a, both RCTs [17,23] detailed the generation method of the random sequence and carried out allocation and concealment.The outcome indicators were also reported, and appropriate models were used to process the missing data.The reason for the uncertainty of two bias risks was that Chia et al. did not explain whether to implement the blinding method for the test personnel, while Pelletier et al. did not provide sufficient information to explain whether to implement the blinding method for the outcome evaluators.Figure 2b details the analysis of risk of bias results.
As shown in Table 7, among the 9 non-randomized studies accessed by modified MINORS scale, 5 demonstrated low risk of bias, and 3 were classified as medium risk of bias, with only 1 presented high risk of bias [12].This was mainly caused by a lack of information about the number of measurement sites for each sample, and the outcome data of control group was retrieved from their previous study.All the studies clearly stated their aims and made quantitative evaluation of fit accuracy.An adequate control group was set in both clinical and in vitro studies.One item, viz.prospective calculation of the study size, was of high risk of bias, with none of the included studies reporting on this item (Fig. 3).

Results of syntheses
There are three studies with missing outcome data.We contacted the authors, but only one responded [21], leaving the two remaining studies to be excluded from the meta-analysis [11,12].In the global meta-analysis performed, the SMD was 0.33(95%CI: -0.18, 0.83, P = 0.21) in favor of CLW.But this difference did not show statistical significance (P > 0.05) and high statistical heterogeneity was found (τ 2 = 0.99, I 2 = 91.19%,H 2 = 11.35,Random effects model) (Fig. 4).

Sensitivity analysis
The result of the sensitivity analysis is shown in Fig. 5, indicating that research done by Pelletier had the most effect on heterogeneity, and heterogeneity decreased after removing this study [I 2 = 85.42%,SMD = 0.06, 95%CI (-0.34, 0.47), P = 0.76] (Fig. 6).

Subgroup analysis
Subgroup analysis was conducted for different types of digital technologies.All studies were compared according to three groups (Table 8).A comparison of fit accuracy in the rest region between AM RPDs and CLW ones involving 9 groups from 8 studies was performed, which showed significant difference between AM and CLW (SMD = 0.83, 95%CI (0.10, 1.56), P = 0.03).Significant heterogeneity between analyses was identified (P = 0.00, I 2 = 92.04%) in a randomeffects model (Fig. 7).
The subgroup of Indirect digital technologies vs CLW included 6 groups from 4 studies.Results in a fixed effects model indicated that CLW RPDs obtained a significant better fit accuracy in rest region than RPDs fabricated by indirect digital technologies (SMD = 0.51, 95%CI (0.23, 0.80), P = 0.00).Low heterogeneity between these analyses was identified (P = 0.09, I 2 = 47.22%) (Fig. 9).
One study evaluated the fit accuracy of RPD rest in all groups before and after finishing and polishing [21].To evaluate the potential effect of finishing and polishing procedure to the fit accuracy of RPDs, an additional comparison was made and results were presented in Fig. 10.No significant differences between groups were observed (SMD = 0.09, 95%CI (-0.71, 0.89), P = 0.83 > 0.05) in a fixed effects model.
Hybrid manufacturing was used in two in vitro studies [11,12] and compared with CLW.Both studies fabricated Aker's clasp assemblies by the same oneprocess molding machine.And silicone film method was used to measure the gap distance between the  [7].With digital relief and heat treatment, Torri et al. fabricated HM clasps with better fit accuracy in the rest region [12].

Reporting biases
A funnel plot was constructed to assess publication bias.As shown in Fig. 11, the funnel plot was visually asymmetric, which indicates that potential publication bias may exist.

Certainty of evidence
As the Galbraith plot shows, most of the points which represent individual studies were within the range of the 95% CI regression line (Fig. 12a) except for two studies: Pelletier et al. [23]【SLS】and Arnold et al. [13]【MId】.
After excluding these two, all of the remaining studies were within the range of the regression line (Fig. 12b), indicating that these two studies may have some impact on the overall effect.

Fi-index tool
This manuscript has been checked with the Fi-index tool and obtained a score of 0 for the first author only on the date 28/07/2023 according to SCOPUS ® [30,31].The fi-index tool aims to ensure the quality of the reference list and limit any auto-citations.

Discussion
The results observed in this study suggested that there were no significant differences in the fit accuracy of the rests of RDPs fabricated by digital technologies and CLW, which means digital technologies can be a viable alternative for the manufacture of RPD frameworks.Subgroup analysis on different types of digital technologies showed that RPDs fabricated by CLW fit significantly better in rest region than those made by AM technologies.Similar result was also found between indirect digital technologies and CLW.
However, MI fabricated RPDs presented a significant better fit accuracy in rest region than CLW RPDs.In regard to the effect of finishing and polishing procedure, RPDs before finishing and polishing presented nominally better but not statistically significant fit accuracy in rest region than those after finishing and polishing.With digital relief and heat treatment, HM clasps also presented significantly better fit accuracy in rest region than cast ones [12].However, this evidence remains to be verified since the HM clasp data was compared to the cast Co-Cr clasp data from their

Ye et al., 2017 [18]
A clearly stated aim previous study [11] rather than including an in-study control group.Digital technology developed rapidly in dentistry.With the assistance of the computer, previously manual tasks are becoming faster and easier and the processing costs are reduced as well [5].However, Takaichi et al. reported that the fitness of the SLM frameworks and clasps was no better than that of cast ones [32].Moreover, Pordeus et al. reported a similar fit between CAD-CAM technology and the conventional technique [29].The present results in this study are in agreement with these previous findings [29,32].Tan et al. reported that MI is an alternative method of CLW for fabricating titanium RPD clasps [33].Several other studies have also proved that milled RPDs are comparable to or better than CLW RPDs and thus can be recommended for longer-term clinical use [13,33,34].Results of subgroup analysis for MI vs CLW in this study provides further evidence for this.On the contrary, AM group showed significantly worse fit accuracy compared to CLW group in this review.This finding was corroborated by Pelletier et al. [23] who found that SLS frameworks exhibited significantly worse clinical accuracy as well as higher variability at the rest region than CLW frameworks.Arnold et al. also reported distinct fitting irregularities in the fit of RPDs fabricated with RP techniques [13].
In a previous study, Michael Braian found out that among the five AM units namely Arcam ® , Concept laser ® , EOS ® , SLM Solutions ® and EOS ® (Co-Cr), the highest overall fabrication precision was achieved by EOS (CoCr) which was below 0.050 mm, close to that of SM system (Mikron ® ) [4].While the other AM machines presented just acceptable precision (< 0.150 mm) on all axes except for the z-axis, which was even worse (> 0.5 mm) [4].It can be inferred that different AM machines as well as different parameters can affect the fit accuracy of the end product [4], which may also be a possible explanation for the high heterogeneity in the pooled result and in AM subgroup.These results demonstrate that AM techniques should be further improved and standardized in RPD fabrication to make sure that every framework is produced with consistent accuracy [4].Nonetheless, with higher fabrication speed and better accuracy, additive manufacturing will seriously compete with traditional manufacturing in creating good end-use products [1,5].
Except for the 11 included studies, many other studies evaluated the fit accuracy of digitally fabricated RPDs [34][35][36][37][38][39].These studies, because they set no control group [35][36][37] or performed indirect digital technologies as control groups [34,38,39] were excluded after screening.As far as indirect digital technologies are concerned, before conventional investment casting, digital model is obtained by scanning and computer-aided design is performed followed by printing or milling of wax or resin pattern [34,38,40].Therefore, in this study, these indirect technologies were not included within the scope of conventional method and were taken as digital technology for RPD fabrication.A comparison between indirect digital technologies and CLW reflects the difference between digital scanning and conventional method of impression-taking and working cast fabrication, while a comparison between indirect digital technologies and fully digital workflow represents the processing tolerance produced from investment to finishing.However, what is really significant in clinical practice is a summation of the errors involved in all the steps from the  scanning to the post-treatment process and also in all the stages of CLW from impression taking to finishing and polishing.For these reasons, a universal classification of RPD fabrication technologies is suggested, especially for indirect methods.Sensitivity analyses showed that another possible cause of heterogeneity was the measurement method.Similar result was reported by Alabdullah et al., who compared the different approaches to evaluating the fit of RPD frameworks and concluded that the discrepancies in the gap distance values are likely to be caused by different registration methods [41].There is no gold standard for assessing the fit accuracy of RPD.Quantitative methods include silicon film method and surface-matching, and the latter can be carried out whether by matching the surface from the master model and the master model with the silicone registration attached [22,42] or superimposing the intaglio surfaces of RPD frameworks onto the STL file of the master model [24].Besides, the direct optical observation was also used to analyze the fit accuracy by light microscopy [13,43,44].Silicone film method is commonly performed by inserting silicone impression material between the RPD and intraoral dentition or master model under a retentive force that is maintained through the setting time.The the silicone replica of the gap may be sectioned afterwards and its thickness directly revealing the gap was measured with Fig. 7 Forest plot of meta-analysis for fit accuracy of RPD rests-AM vs CLW Fig. 8 Forest plot of meta-analysis for fit accuracy of RPD rests-MI vs CLW stereomicroscope, digital microscope, electronic calipers or profile projector [18,35,45,46].However, Yoon et al. reported that the number of measuring points have effect on the average thickness of the silicon replicas, and that the accuracy of silicone film method was not sufficiently reliable [47].In contrast, three-dimensional surface-matching can be used to assess the fit accuracy of RPDs more comprehensively and effectively than silicone film method [47].Consequently, for silicone film method, the adequate force applied during the setting time, the type of silicone replica, as well as the number and site of measuring points need to be clarified, which is necessary to insure the reliability and reproducibility of the outcomes of individual studies in the future.In addition, up to now there is no consensus about the clinical acceptable gap distance of RPDs.Stern et al. reported that a gap of 0 to 50 μm was deemed to be close contact [16].In a clinical study conducted by Dunham et al., the average gap distance between the rests and the rest seats was 193 ± 203 μm, ranging from 0 to 828 μm [48].Li et al. fabricated 13 one-piece PEEK RPDs and the gap distance was 84.3 ± 23.6 µm in rest region [37].Among the present 11 included studies, the mean average gap distances in rest region ranged from 30 μm to 365 μm in digitally fabricated RPDs, and 20 μm to 279.61 μm in CLW groups.Several studies compared the overall fit accuracy of RPD frameworks fabricated by digital and conventional technologies [19,24,35,42], and some only evaluated the fit accuracy of clasps [11,12,34,40,43].However, the low overall internal discrepancy value is not equivalent to better fit, as it is the compounded result of individual components.The RPD rests could be the ideal reference for the fit evaluation, which is easy for measurement and important for functional loading of the overall framework [45].The fit accuracy of other RPD components, namely the connectors, the clasp arms as well as denture base should be further investigated for both digital and conventional fabrication technologies.
One of the factors that is most influential for fit accuracy is the finishing and polishing procedures on the tissue surface of the frameworks, especially for the rests [20].In other words, to improve fit accuracy, finishing procedures in the laboratory should be well-controlled and excessive removing of metal from the intaglio surface should be avoided [16].Most of the RPD samples included in this review were polished [13,17,18,[20][21][22][23], except in 3 studies [11,12,24].Different methods and extent of manual polishing are likely to affect the interpretation of the final results, while meta-analysis in this review showed no significant influence of finishing and polishing on the fit accuracy in rest region.Hence, further studies considering finishing and polishing procedure with a larger sample size are needed to validate this conclusion.

Limitation
Only 11 studies were included in this meta-analysis after a comprehensive search in the main databases, indicating that the number of original research in relevant fields is still limited.No consensus has been reached for quality assessment of in vitro studies.The modified version of MINORS scale has not been validated in terms of index content as well as scoring, and the results of assessment can only be referenced conservatively.Moreover, since the findings of this review are mainly based on in vitro studies, caution must be exercised when applying the results into clinical practice.

Conclusions
1. RPDs fabricated by digital technologies exhibit comparable fit accuracy in rest region with those made by CLW.
. The search strategy was based on the PICOS (Population, Intervention, Comparison, Outcomes, Study) format: P (Population): Removable Partial Denture I (Intervention): Digital technologies including AM (3D printing etc.), SM (MI etc.), indirect digital technologies (milling or printing of wax/resin patterns followed by CLW) and HM.C (Comparator): Conventional lost-wax casting technology O (Outcome): Fit accuracy in the rest region, which is represented by the gap distances between the rest seats and the intaglio surfaces of the occlusal rests (μm) S (Study): Clinical studies and in vitro studies An electronic search was performed in Cochrane Library, PubMed, EMbase, Web of Science and SpringerLink on August 2nd, 2022, including articles published from January 1950 until August, 2022.

Fig. 1
Fig. 1 Flow chart of study selection process according to PRISMA diagram

Fig. 2 a
Fig. 2 a Risk of bias summary of the included RCTs.b Risk of bias item presented as percentage across two included RCTs

Fig. 3 Fig. 4 Fig. 5
Fig. 3 Risk of bias by modified MINORS scale.Author's judgments about each item presented as percentage across all in vitro studies and a non-randomized clinical study

Fig. 6
Fig.6 The result of sensitivity analysis (after removing the study ofPelletier)

Fig. 9
Fig. 9 Forest plot of meta-analysis for fit accuracy of RPD rests -Indirect digital technologies vs CLW

Table 2
Inclusion and exclusion criteria Inclusion Criteria 1. Study types: Randomized controlled trials (RCT), non-randomized clinical trials, cohort studies, case control studies and other observational studies as well as in vitro studies 2. Articles published in the period from 1950 until Aug. 2022 3.No publication language limits taken into account Exclusion Criteria 1. Articles that used only qualitative method to evaluate the fit accuracy, such as clinical check and pressing test without available data 2. Articles that studied the RPDs fabricated by indirect digital technology (printing or milling wax/resin patterns before investment casting) rather than CLW as control groups 3. Reviews and studies with only charts and questionnaires 4. Articles unavailable in the databases or articles that are inaccessible to read in full

Table 3
Modified version of MINORS scale

Table 4
Summary of included studies (basic information) RCT Randomized controlled trials, CLW Conventional lost-wax technique, SLM Selective laser melting, RPD Removable partial denture, SLS Selective laser sintering, CAD/RP Computer-aided design/ Rapid prototyping, MI Milling, PEEK Poly ether ether ketone, CAD/RPS Selective laser melting from stone model, LWTR Lost-wax technique from resin model, Rpi Indirect rapid prototyping, Rpd Direct rapid prototyping, MIi Indirect milling, MId Direct milling, MEP Metal 3D printing, RPC Resin printing and subsequent casting, C-M Conventional casting of milled sacrificial patterns, C-P Conventional casting of printed sacrificial patterns, DMLS Direct metal laser-sintering, HM Hybrid manufacturing, RLS Repeated laser sintering

Table 5
Summary of included studies (measurement information) SLM Selective laser melting, SLS Selective laser sintering, CLW Conventional lost-wax technique, CAD/RP Computer-aided design/ Rapid prototyping, LWT Lostwax technique, CAD/RPS Selective laser melting from stone model, LWTR Lost-wax technique from resin model, Rpi Indirect rapid prototyping, RPd Direct rapid prototyping, ΜIi Indirect milling, ΜId Direct milling, CON Conventional lost-wax technique, MEP Metal 3D printing, RPC Resin printing and subsequent casting, DMLS Direct metal laser-sintering, C-M Conventional casting of milled sacrificial patterns, C-P Conventional casting of printed sacrificial patterns, HM Hybrid manufacturing, RLS Repeated laser sintering, NR Not reported, Co-Cr Cobalt-chromium

Table 6
Summary of included studies (manufacturing information) Conventional casting of printed sacrificial patterns, DMLS Direct metal laser-sintering CAD Computer-aided design, SLM Selective laser melting, RP Rapid prototyping, NR Not reported, CAD-CAM Computer-aided design-Computer-aided manufacturing, RPd Direct rapid prototyping, MId Direct milling, RPD Removable partial denture, C-M Conventional casting of milled sacrificial patterns, C-P

Table 7
Evaluation of risk of bias by modified MINORS scale 0: not reported, 1: reported but inadequate, 2: reported and adequate; The global ideal score being 24 for in vitro studies and 26 for clinical studies.Risk of bias of individual study judged by total score: < 16 high, 16-18 medium, > 18 low Methodological

Table 8
Results of subgroup analysis

Table 9
Results of two studies included in qualitative description