Occlusal teeth surface accuracy of milled complete dentures: a comparison between different manufacturing techniques

Abstract

PURPOSE

This study aims to compare the occlusal trueness and precision of teeth manufactured using two modern digital milling processes.

MATERIALS AND METHODS

A total of 38 complete dentures (CDs) were fabricated and analyzed. CDs in Group 1 (monolithic) (n = 19) were produced using a monolithic bicolor resin disk, whereas in Group 2 (oversize) (n = 19) were fabricated using the oversize process, which involves two separate resin disks of different colors. Two investigation methods were developed to evaluate trueness and precision: cusp area analysis and cusp vertex analysis. The study included three levels of analysis: a comparison of the two measurement methods, an evaluation of the monolithic versus oversize processes, and an assessment of under- and overcontouring inaccuracies.

RESULTS

Statistical analysis using the Welch two-sample t-test, the non-parametric Wilcoxon signed-rank test, and the modified signed-likelihood ratio test (SLRT) revealed a statistically significant difference (P < 2.2 × 10⁻¹⁶) between the two measurement methods (vertex vs. area) for both the monolithic and oversize groups, with the vertex method demonstrating greater accuracy. The analysis of over- and undercontouring inaccuracies revealed that 55% of the surface for the monolithic process exhibited overcontouring, compared to 99% for the oversize process, indicating a strong tendency toward surface roughness in the latter.

CONCLUSION

The monolithic milling method exhibited significantly superior accuracy compared to the oversize process (P < .05). Additionally, the Reference Point System (RPS) metrological method proved more reliable than the best-fit method for comparing complex structures, offering more accurate estimates of both trueness and precision.

INTRODUCTION

Over recent decades, digital technologies available for the fabrication of complete dentures (CDs) have changed the conventional protocols for their design and manufacturing, leading to efficient clinical results and reduced processing time.¹,² Milling techniques to produce CDs are more precise than the traditional molding method, offering advantages such as the absence of polymerization shrinkage and reduced monomer release.³,⁴ In terms of mechanical properties, the milled resins used to manufacture CDs have higher microhardness than conventional polymerized polymethylmethacrylate (PMMA) resins.⁵,⁶,⁷,⁸,⁹,¹⁰,¹¹,¹²,¹³ Additionally, CAD-CAM resins differ significantly from conventional resins in physical properties such as surface roughness, which affects bacterial plaque adhesion, and hydrophobicity.⁶,¹⁴,¹⁵,¹⁶,¹⁷,¹⁸ From a chemical perspective, milled resins release minimal residual monomer, although studies indicate no statistically significant differences in residual monomer levels when compared to conventionally polymerized resins.⁶,¹⁶

Two primary methods exist for digitally fabricating CDs: additive manufacturing (3D printing) and milling. The most popular in the last decades has been the milling method that makes usage of proprietary resin discs, polymerized under high pressure and temperature.¹⁶ Wang et al.¹⁹ investigated the occlusion and the adaptation of the intaglio surface to the underlining mucosa as a function of the manufacturing process. This study showed the accurate adaptation to mucosa of digitally manufactured CDs. Baba et al.²⁰ reviewed various CD manufacturing techniques and confirmed that digitally produced dentures have enhanced physical properties.

This study compares the occlusal accuracy (trueness and precision) of teeth fabricated using two different subtractive techniques for producing CDs: the oversize process, which utilizes two pre-polymerized differently colored discs (one for teeth and one for gingiva), and the process for monolithic disc, which utilizes a unique pre-polymerized monolithic disc with a specific combination of pink and white resin. The single milling session of the monolithic disc is made possible by the shell geometry, which allows for the recreation of the gingival contours through a three-dimensional arch form design of pink and white resin that replicates the shape of a shell, and which is designed and patented based on natural dental anatomies. On the other hand, the oversize method has been proposed, maintaining the accuracy of the monolithic system through a two-stage approach. After milling the teeth in the white disc resin, and the base in the pink disc, the two parts are glued using a self-curing PMMA-based adhesive. The pink resin base remains anchored to its original disc during bonding, allowing it to be reinserted into the milling machine for refinement, thereby minimizing alignment inaccuracies.

Measurements were conducted using the Reference Point System (RPS), as previously described by authors.²¹,²² Two sets of measurements were collected: one based on the cusp area measurements, the other on the cusp vertex position. The null hypothesis of this in vitro study was that there is no difference of the mean deviation (Δ)-error values between the two different manufacturing methods. The secondary null hypothesis was that there is no difference between the area-based and vertex-based measurement techniques.

MATERIALS AND METHODS

In this study, 38 CDs were manufactured and analyzed. Group 1 (monolithic) consisted of 19 CDs produced using a monolithic bicolor PMMA resin disc (Ivotion^®, Ivoclar, Schaan, Lichtenstein), while group 2 (oversize) comprised 19 CDs using the Oversize Technique (Ivotion, Ivoclar), which involves two distinct PMMA resin disks, one for teeth (Ivotion^® Dent, Ivoclar, Schaan, Lichtenstein) and one for the denture base (Ivotion^® Base, Ivoclar, Schaan, Lichtenstein) (Table 1). A metrological approach based on the RPS was employed to compare the two groups, using two measuring methods to investigate cusp geometry: surface area analysis and vertex-based analysis.

Table 1. Chemical characteristics of the materials used in the study (Data by manufacturer).

Product name	Manufacturer	Composition
Ivotion Base	Ivoclar, Schaan, Liechtenstein	Polymethyl methacrylate (PMMA) > 90%; co-polymer, pigments
Ivotion Dent	Ivoclar, Schaan, Liechtenstein	Double crosslinked polymethyl methacrylate (PMMA)
Ivotion (Monolithic)	Ivoclar, Schaan, Liechtenstein	Highly cross linked PMMA tooth material with premium denture base material

An edentulous maxillary plaster model, created from a polyvinyl siloxane (PVS) impression of an edentulous patient archived in the dental clinic, was used as the starting point. The use of a clinical model, rather than a standardized sample, enhanced the clinical relevance of the study by simulating a real patient scenario. The plaster cast was scanned with a benchtop scanner (3Shape Generation Red E2 scanner, 3Shape, Copenhagen, Denmark), and the resulting 3D digital model was modified via CAD software to include five geometrically defined cones (Fig. 1).

Fig. 1. The computer aided design of the master model with the landmark cones (v₁ − v₅).

The vertices of the cones, called V_1–5, allowed the creation of an “X, Y, Z” reference system. The least square plane through the cone vertexes defines the Z axis. The origin O_XYZ of this coordinate system lies on that plane and coincides with the point identified by the intersection of the projections of the segments V_1-V4 and V_2-V₅ on the plane, while the direction of the Y axis points towards the frontal cone (V₃). The vertices, oriented towards the occlusal plane, allowed the use of the RPS system for alignment, providing five points easily identifiable through geometric construction of the vertex of the cones. This digital model was used to fabricate a titanium master model via laser melting (M280; EOS, Krailling, Germany), which eliminated risks of distortion or imperfection during subsequent procedures. To reduce optical scattering during scanning, the titanium model was coated with a dental opacifier (Spy Spray, Apex Dental srl, Milan, Italy). It was then scanned using a 3Shape Generation Red E2 scanner, compliant with ISO 12836, ensuring an accuracy of 10 µm. Using a design software (3Shape Dental System, 3Shape), a digital maxillary CD incorporating 14 teeth from the software’s tooth library was designed. The previously modeled cones were integrated into the intaglio surface of the prosthetic base.

The finalized STL file was sent to a milling machine (PrograMill PM7, Ivoclar, Schaan, Lichtenstein), which provides a manufacturing accuracy of ± 10 µm, per manufacturer specifications. Nineteen CDs were milled starting from a single monolithic disc, combining base and teeth in a unified structure and completed in one milling session using the proprietary shell geometry. The remaining 19 CDs were produced by using the Oversize Technique (Ivotion, Ivoclar, Schaan, Lichtenstein), in which base and teeth are fabricated separately from pink and white PMMA discs, respectively. During gluing, the base remains anchored to its resin disk, to allow its repositioning in the disk holder and maintain its initial position in the milling machine and to complete the milling process according to the CAD project. It refines only the overcontours in the occlusal area, eliminating any inaccuracy introduced after gluing, and to avoid bias, no adjustments were made to the occlusal surfaces in either group. All prostheses were analyzed at the Design Tools and Methods in Industrial Engineering Laboratory of the University of Padua. Scanning was performed using an optical 3D scanner (AuRum3d scanner, Open Technologies, Brescia, Italy) and the scan software (Optical RevEng Dental, Open Tech 3D s.r.l., Brescia, Italy), enabling 360° acquisition of the CD under controlled lighting and temperature conditions. Each denture was then coated with a thin layer of red opacifying spray (Spy Spray, Apex Dental srl, Milan, Italy), to enhance scan accuracy by reducing possible light scattering.

Each denture was scanned three times from different orientations — palatal, occlusal, and vertical position — allowing alignment of the internal and external geometries. The resulting complete 3D models were exported as a STL format for analysis.

Then, the reference model to compare the scans was created, and the CAD file used for manufacturing the prothesis was aligned to the same reference system that was previously materialized on the titanium master (Fig. 1). In a modeling software (Rhino 7, Rhinoceros^®, TLM Inc., Barcelona, Spain), a defeaturing operation removed all but the relevant cone and cusp geometries. Based on the definition of cusp, the meshes corresponding to the occlusal grooves and lateral walls were eliminated, allowing each cusp to be made as a separate three-dimensional entity, named according to the dental element and cusp position. This process yielded 24 regions of interest, corresponding to the cusps of four molars and four bicuspids (Fig. 2).

Fig. 2. (A), (B) Reference model before (A) and after (B) the modification performed with Rhino 7 software. Note the isolated cusps in element 27, as well as the reference cones without a vertex. Part of the dental structure has been eliminated to facilitate the subsequent steps.

These reference geometries were imported into the inspection and metrological evaluation software (GOM Software 2022, Zeiss, Oberkochen, Germany) and compared with the 38 scanned dentures. The scan was first prealigned to the reference model by means of a best-fit operation (Fig. 3A). The geometry of the six cones on the palatal side was sampled and the vertex of each cone was mathematically computed. The final alignment was obtained through a RPS operation based on the six cone vertexes (Fig. 3B) in compliance with the reference geometry approach previously presented by the authors.²¹,²²

Fig. 3. (A), (B) Differences in the best-fit alignment (left image) and RPS (right image) of a monolithic prosthesis. Note how the RPS alignment provides a more realistic picture of the actual situation, despite the deviation being greater. The deviation was 0.0278 mm for the RPS alignment and 0.0006 mm for the best-fit alignment.

By the measuring method 1 (vertexes of the cusp), the cusp vertex was defined as the point with the highest Z-coordinate on the cusp surface (Fig. 4). Corresponding vertex coordinates were identified on both the reference and scanned models. Spatial deviations (Δ_X, Δ_Y, Δ_Z, Δ_XYZ) were calculated and exported in .csv format.

Fig. 4. Allocation of the vertex (origin of green arrow) located on a cusp (red area). The disk is represented by the green line in this projection. In the thumbnail at the top left, the disk (green) is visible from another perspective.

By the measuring method 2 (area of the cusp), a cloud of ~300 equidistant points was mapped onto the surface of each cusp in the reference model (Fig. 5). Corresponding points in the scanned data were identified along the surface normal to the reference geometry. Deviations (Δ_X, Δ_Y, Δ_Z, Δ_XYZ) were computed for each point across each cusp of all 38 specimens, resulting in a dataset of approximately 290,000 deviations.

Fig. 5. (A) The cloud of points identified on the surface of each cusp of a molar; (B) The map of the cusp areas.

RESULTS

The milling accuracy was evaluated using both measuring methods in order to compare their measuring efficacy (Tables 2 and 3).

Table 2. Mean deviation for each measure for the monolithic system, obtained through the area and the vertex measurement methods (mm).

Measurement Method	Δx	Δy	Δz	Δxyz
Area	0.025	0.019	0.051	0.063
Vertex	0.088	0.088	0.068	0.164

Table 3. Mean deviation for each measure for the oversize system, obtained through the area and the vertex measurement methods (mm).

Measurement Method	Δx	Δy	Δz	Δxyz
Area	0.066	0.053	0.138	0.172
Vertex	0.065	0.091	0.173	0.224

To compare the vertex and area measuring process, the trueness of the monolithic process was evaluated both with respect to the mean position error of each vertex (Δ_{xyz Vertex-measurements}), and with respect to the mean position error of the different area points identified on the cusp itself (Δ_{xyz Area-measurements}). A mean 0.063 mm Δ_{xyz Area-measurements} error between the CAD designed and the CAM manufactured CDs was identified for the area measuring process. A mean 0.164 mm Δ_{xyz Vertex-measurements} error was documented for the vertex measuring process. The oversize method reported a higher mean error for both the area and vertex measuring processes, equal to 0.172 mm and 0.224 mm, respectively. To check whether a statistically significant difference existed between the mean values of each measurement method (Vertex and Area), the Welch two sample t-test was used. The P-value of 2.2 × 10⁻¹⁶ confirmed the statistically significant difference between the area and the vertex measurements, for both monolithic and oversize group. The t-value showed that the difference is more evident in the monolithic method (−21.894 absolute Δ_xyz-value, effect size measured by Cohen’s d = −1.45) than for the oversize (−14.836 absolute Δ_xyz-value, Cohen′s d = −0.98), with a lower mean Δ_xyz-error absolute value for the monolithic method regardless of the evaluation method. These results suggest that the two methods of measuring accuracy are different in assessing the accuracy of the milling process.

To evaluate the relative variability and consequently the precision, the coefficients of variation (CV), given by the ratio between the standard deviation and the mean value, were compared using the Modified signed log-likelihood ratio test (SLRT). For the monolithic, the CV of the area measurements (0.826) was higher than the CV of the vertex measurements (0.508), showing that a variability of the data in the area measurements with respect to the mean value is present and is greater than the vertex measurements. For Group 1 (monolithic), the P-value (2.2 × 10⁻¹⁶) confirmed that this difference was statistically significant, confirming a worst precision of the area measurement method. For the oversize, the CV of the area measurements (0.254) was slightly lower than that of the vertex measurements (0.271), confirming a slightly lower precision of the vertex measurement method. However, the P-value (.189) demonstrated that this difference is not statistically significant in the oversize group. The relative variability analysis showed that the accuracy measured through the surface analysis is more variable for the monolithic method than for the oversize method.

After comparing the monolithic and oversize process, the statistical analysis was focused on the differences between the two manufacturing methods (monolithic and oversize), to determine the best trueness of manufacturing of the occlusal surface of CD teeth. For this purpose, individual-specific differences for each method were computed; the non-parametric Wilcoxon signed-rank test showed that the monolithic (Δ-error_{Area measurements} value 0.063 mm; Δ-error _{Vertex measurements} value 0.164 mm) process was more accurate than the oversize (Δ-error_{Area measurements} value 0.172 mm; Δ-error _{Vertex measurements} value 0.224 mm), showing better trueness regardless of the evaluation method (by area or by vertices) used (Fig. 6). The effect size for the vertexes method is 0.812, while for the area method is 0.877. Both the P-value (< .05) and the modified SLRT for the difference of CVs confirmed a statistically significant difference between monolithic and oversize process, attesting the best accuracy of the monolithic manufacturing process.

Fig. 6. (A) The box plots representing the mean Δ_XYZ distance for the vertex measurements. (B) The box plots representing the mean Δ_XYZ distance for the area measurements.

The CV of the Δ_{xyz Area-measurements} for the monolithic group is 0.575, while that for the oversize process is 0.208; the CV of the Δ_{xyz Vertex-measurements} for the monolithic group is 0.262, while that for the oversize process is 0.126. Consequently, the monolithic process showed lower precision than the oversize process for both the area measurement methods.

Finally analyzing the under- or overcontouring quality of the inaccuracies, the extent of the inaccuracy was evaluated to understand whether it was due to an excess or a lack of material created during the milling process. The analysis along the Z axis showed 55% of overcontouring for the monolithic, i.e. an almost equal distribution between the excess and defect areas (Fig. 7). For the oversize, the percentage of positive values exceeded 99%, showing a clear tendency towards over-contouring of this manufacturing process (Fig. 8).

Fig. 7. Distribution of the Δ-error at the surface level of a monolithic sample with the scale limit set to 0.300 mm. The elements of quadrant 3 and 4 (left side) are oversized compared to the initial design, while those of the quadrants 1 and 2 are undersized. This tendency was found in all the monolithic samples.

Fig. 8. Δ-error distribution at the surface level of an oversize sample with the scale limit set to 0.300 mm. Unlike the monolithic, a clear tendency towards oversizing is noted (as clearly shown by the deviation histogram), with a less marked difference between left and right quadrants.

DISCUSSION

The primary null hypothesis of this study that no difference in accuracy exists between the oversize and the monolithic processes for manufacturing CDs was rejected. Also, the secondary null hypothesis that no difference exists between the two metrological methods was was rejected, too.

The use of CAM process enabled the achievement of accuracy values that are significantly higher than those associated with conventional techniques.²¹,²³,²⁴ A further advantage of the digital workflow is the ability to utilize high-performance materials, with mechanical properties superior to traditional alternatives.²⁵ This study compared the monolithic and oversize processes, both proposed by the Ivoclar, for the fabrication of removable complete dentures. Both the monolithic and oversize systems employ industrial polymerization for disc production, ensuring high material stability and excellent impact resistance due to the inclusion of cross-linking agents. However, the oversize system yields superior impact resistance, requiring at least 1200 J/m² for fracture propagation, compared to 800 J/m² for the monolithic system. Milling has been considered the gold standard in denture fabrication due to its superior accuracy and precision compared to other techniques.⁷,²³,²⁶ All specimens in this study were milled using the same device under identical settings, as both milling time and speed are known to influence results.²⁷ Numerous studies have demonstrated that milling ensures excellent accuracy on the intaglio surface,²¹,²⁶,²⁸,²⁹ with Δ-error values consistently below the mucosal tolerance threshold of 300 mm, as accepted in the literature.³⁰ On the occlusal surface, although many authors report acceptable levels of accuracy,³¹,³²,³³,³⁴ some have documented deviations from the original CAD design.³⁵,³⁶,³⁷ This discrepancy is often attributed to the gluing process used to fix the denture teeth into the base,³⁵ which inherently introduces dimensional variation. To mitigate this, the oversize technique was developed. In this approach, definitive milling is performed after the teeth are bonded to the base, thereby eliminating positional errors induced by the bonding step. This method should offer a level of accuracy comparable to monolithic milling, which is completed in a single uninterrupted cycle and thus depends solely on the milling machine’s precision. The data of area analysis showed that the monolithic process is more accurate, with a mean Δ-error of 63 µm compared to the 172 mm of the oversize process. Although the accuracy of the monolithic was near that of the study of Herpel,³¹ which had obtained a Δ-error value of 65 µm, the oversize resulted in a lower accuracy than those in other studies.⁸ One comparative investigation³⁸ examining seven digital fabrication methods found that milling systems produced denture bases with inaccuracies as low as ± 21 µm, while 3D printing systems showed mean errors ranging from 58 to -64 µm. Graf et al.³⁹ also reported highly favorable outcomes for milling, with mean errors of -6 µm on the x-axis, 22 µm on the y-axis, and -20 µm on the z-axis.

In this study, the positive/negative Δ-error distribution is also more uniform for the monolithic process, with a very similar distribution between the under- and over-contouring areas (respectively, negative and positive Δ-error value), unlike the tendency towards over-contouring shown by the Oversize. This study also found that the Monolithic process exhibited a more balanced distribution of deviations, with under- and over-contouring occurring at similar frequencies (55% and 45%, respectively). In contrast, the Oversize process showed a pronounced tendency toward over-contouring (99%). This suggests that the bonding step may introduce vertical discrepancies, increasing the height of the occlusal surfaces. While both over- and under-contouring may contribute to plaque retention, further investigation is needed to establish clinical correlations. Clinically, these findings support the use of monolithic milling for achieving superior occlusal accuracy.

The metrological comparison system involved the use of a RPS as an alternative to the commonly adopted superposition measuring method: the superimposition protocol is based on the best-fit concept, trying to obtain the lowest possible deviation (Δ-error) value. For example, by performing the same protocol of this study using the superimposition method, accuracy values even lower than 10 µm might be obtained (mean Δ-error = 6 µm). The RPS method, however, provided a more accurate representation of the true deviations, as shown in prior literature.²¹,²²

This operation ensures that the alignment is based on minimizing the distances between the reference cone vertices and the actual (measured) cone vertices. As a result, the alignment is determined by the best fit of the five cone vertices. In contrast to the “standard” best-fit method, which considers all available points, this approach prevents defective zones from interfering with the alignment. As shown in Fig. 3A, the “standard” best-fit minimizes the error across the entire geometry. Consequently, the average error at the cusp tends toward zero, making it impractical to identify sources of error. In contrast, the RPS alignment is independent of the actual cusp geometry, allowing geometric differences to be immediately identified and quantified. As illustrated in Fig. 3B, one side clearly shows a lack of material, while the other exhibits excess material. These discrepancies allow for the immediate identification of systematic trends in the manufacturing process, particularly a tendency toward a negative rotation of the actual geometry around the y-axis. Understanding this behavior can enable adjustments in the manufacturing setup to optimize the final geometry. However, the reasoning behind this systematic tendency is not straightforward and may depend on specific manufacturing conditions. A more in-depth discussion of the benefits of this type of alignment has been previously provided in the literature.²² For these reasons, RPS alignment is considered superior to best-fit alignment for detecting trueness and precision.

Regarding the two measurement approaches (area or vertex Δ-error calculation), the possibility of using only the vertex of the cusp for the accuracy evaluation reported values that were on average higher than those obtained through the area evaluation method. These two quantitative and repeatable evaluation methods were specifically developed to overcome limitations of conventional surface metrology, which typically produces color maps (Figs. 7 and 8) indicating deviation trends but lacks precision in numerical reporting. In this case, the reference model and actual data (measured data) are visually overlapped and a color is attributed to each point of the reference model based on the distance of the actual data from reference model. While this representation is useful to detect trends and interpret results, it does not provide a direct numerical quantity. Manual deviation sampling using “flag” tools introduces subjectivity and lacks reproducibility. In contrast, the proposed methods are automated yielding unbiased, statistically analyzable data.

This difference is due to the method used to identify the vertex, which is defined as the point with the highest Z coordinate among all the points that form the surface of the cusp, without considering the values of adjacent points. In an almost flat and “horizontal” (normal to the Z direction) surface, such as that of the cusp of a prosthetic molar or premolar, a multitude of points could have similar values of the Z coordinate, potentially determining an incorrect estimate of the accuracy in the case of two points that are distant from each other along the X and Y axes, but with almost identical Z values. Indeed, the Δ-error is calculated as the average between the XYZ distances. This limitation is consistent with the finding of a greater difference between the two investigation methods in the Monolithic, where the greater overall accuracy showed the limits of the vertex measurement. Additionally, vertex identification is labor-intensive and operator-dependent, making it susceptible to human error. In contrast, the area-based method, although it requires greater computing power and involves a large amount of data, is largely automated and reproducible because it is automatically managed by the software, making it preferable for larger sample sizes. Moreover, the cusp vertex has limited clinical value, as it is rarely directly involved in the occlusion of an edentulous patient. Many studies demonstrated that the occlusal discriminatory capacity in edentulous patients can reach values of 80 µm in subjects who have been rehabilitated for a long time.⁴⁰,⁴¹,⁴²,⁴³ In the current study, Monolithic dentures showed Δ-error values (63 µm) well below this threshold, while the Oversize process slightly exceeded it (172 µm). The RPS method allowed for clear identification of systematic over-contouring in the Oversize group—likely due to adhesive layer thickness—whereas traditional best-fit methods would likely have obscured such trends by averaging out opposing deviations.²²

Limitations of this study include the use of a single milling system and resin type. Future investigations with broader sample sizes and different milling technologies are needed to validate the reproducibility of these findings and to explore correlations between occlusal accuracy and functional outcomes.

CONCLUSION

The Monolithic manufacturing process demonstrated significantly greater accuracy than the Oversize process, with Δ-values well below the clinically accepted threshold for edentulous patients.

The RPS metrological method was confirmed to be superior to traditional best-fit (superimposition) techniques for evaluating of complex geometries, such as the convex-concave surfaces of CD teeth offering a more realistic estimation of true accuracy.

The area-based measurement method was shown to yield more reliable and clinically relevant results than the vertex-based approach, particularly when high-accuracy is required.