Effect of Die Ditching on the Marginal Fit of CAD/CAM-Fabricated Co-Cr Crowns: An In Vitro Comparative Study

Abstract

Background

Manual die ditching can compromise CAD/CAM crown marginal accuracy beyond clinical thresholds. This in vitro study evaluated how conventional die ditching versus non-ditching of removable stone dies influences the marginal fit of digitally fabricated cobalt-chromium crowns.

Material/Methods

A mandibular first molar typodont prepared for a complete crown was duplicated into 20 stone dies, randomized into non-ditched (Gp CN) or ditched (Gp CD) groups (n=10 each), with the typodont as a reference (Gp T). Gp CD underwent standardized ditching, while Gp CN received minimal trimming. Following scanning, digital design, and milling, marginal fit was assessed via 3D STL superimposition at 8 landmarks. Landmark-specific root mean square (RMS) deviations were averaged for mean marginal RMS (primary outcome), while landmark-level values served as exploratory data. Non-parametric tests (Kruskal-Wallis with Dunn’s post hoc, Mann-Whitney U, Friedman) were used as primary analyses, with Welch t tests and Cohen’s d calculated only as sensitivity measures (α = 0.05).

Results

In terms of the primary outcome, crowns on non-ditched dies showed lower mean marginal discrepancies (0.086 ± 0.070 mm) than those on ditched dies (0.147 ± 0.200 mm). A secondary outcome analysis using 3D STL superimposition showed that Gp CD had higher root mean square (RMS) deviations, suggesting that ditching reduces the stability of the finish-line registration during scanning. Gp CN values stayed well within the 120-μm clinical acceptability threshold, but Gp CD values were observed to cross this threshold.

Conclusions

Within this CAD/CAM workflow, conventional die ditching was associated with larger marginal discrepancies and fewer measurements within the 120-μm clinical threshold, with localized effects most evident buccally.

Introduction

Decayed or fractured teeth can be restored through direct and indirect restorations (IR), with the latter being fabricated exclusively in dental laboratories via casting of dental alloys or computer-assisted design/computer-assisted machining (CAD/CAM) (additive or subtractive). Global dental crowns and bridges are projected to reach 5 to 7 billion by 2034 [1], which historically originated in the 18th century through the fine work of Pierre Fauchard [2], while evolving through 19th and 20th centuries with addition of porcelain jacket crowns [3], acrylic resins, and aluminous porcelain [4]. Precision die systems for fabrication of restorations were introduced in early 1900 [4], which continue to remain in place despite improvements through CAD/CAM milling and three-dimensional printing [5]. While traditional fabrication depended on the skills of dentists and technicians (primarily the lost wax technique), this process is increasingly replaced by CAD/CAM due to reduced reliance on subjective variables, making it more accurate and clinically efficient [6,7]. In the lost wax technique, die precision aids in creating well-contoured cervical margins and reproducing surface anatomy [8], which reduces clinical failures associated with the periodontium. An individual die isolates a prepared tooth, enabling precise contouring and enhancing marginal adaptation [9,10]. Since restorative material shrinkage results in marginal discrepancy (over, under, open, short margins) die ditching minimizes or eliminates such subject-induced errors [11,12]. However, improper ditching, such as excessive apical depth or deep angulation, does change the planned wax pattern, resulting in thicker margins, abnormal emergence profiles, and clinical or subclinical plaque retention [13,14].

Marginal fit is a critical determinant of the long-term success of full-metal and metal-ceramic crowns. Discrepancies exceeding the clinical threshold of 120 μm promote biofilm accumulation, secondary caries, and cement dissolution [15,16], and simultaneously predisposing restorations to several biological and technical complications [17]. Past studies on full-metal and metal-ceramic restorations achieved acceptable gaps between 9 and 112 μm [18]. McLean and von Fraunhofer’s classic long-term clinical study proposed that marginal gaps up to approximately 120 μm should be regarded as clinically acceptable for full-coverage restorations, which remain the benchmark for present contemporary evaluations [19]. Recent in vivo data comparing metal-ceramic crowns (95.6 μm) with CAD/CAM zirconia (77.4 μm) underscores that while marginal adaptation are both are within clinical limits, small variations in technique can still influence gap magnitudes [20]. Moreover, over-contoured cervical surfaces are more detrimental to gingival health than slightly under-contoured surfaces [11,14,21].

To achieve this clinical precision, fixed prosthodontic workflows rely on modifiable working casts that allow individual dies to be removed and placed back (eg, the PINDEX system), thus allowing detailed access to individual preparation feature while maintaining accurate spatial relationships within the arch [10,11,13,14]. Other removable-die systems, such as straight dowel-pin casts and Di-Lok trays, similarly aim to combine the stability of a solid arch form with the flexibility of individually accessible dies for waxing and finishing [11,14]. These die systems are routinely employed for fabricating full-metal and metal-ceramic restorations because they permit meticulous evaluation of finish lines, proximal contacts, and occlusal morphology without compromising the integrity of the overall cast [11]. After the working cast is produced, the individual preparation is sectioned and refined to create a definitive die that accurately reproduces the prepared tooth structure and finish line [14]. Excessive ditching can distort emergence profiles, resulting in over-contoured crowns, which may negatively impact gingival health [11,12,14].

Die ditching is generally not preferred as a routine step for CAD/CAM-fabricated crowns; when it is used, it should be minimal and controlled rather than aggressive. CAD/CAM workflows depend on accurate 3D capture of the true tooth or stone-die geometry; excessive trimming around the finish line can distort the emergence profile and change the reference surface the scanner records [22,23]. Technical guidelines for open CAD/CAM systems (eg, Renishaw) emphasize preparing accurate stone models, sectioning them as appropriate, and then scanning the preparations and defining margins in software, without recommending heavy mechanical reshaping of the die beyond standard trimming [24,25]. Studies on scanning die stones with 3Shape laboratory scanners show that the marginal fit of CAD/CAM crowns is influenced mainly by stone properties and scanner performance, not by surface modifications like powders or aggressive alterations, and recommend selecting die stone based on physical properties rather than additional surface treatments [26]. Over-ditched dies can lead the technician to wax or design crowns with over-contoured cervical areas to “fill” the artificially deep sulcus, which is known to worsen plaque retention and gingival health. Many laboratory teaching sources for hybrid analog-digital workflows still describe light ditching or margin outlining on stone dies simply to make the finish line easier to visualize before scanning, followed by digital margin placement in the CAD software [24]. Despite widespread adoption of CAD/CAM workflows, the management of stone dies remains largely based on traditional analog practice [5,14,25,26]. Classical teaching advocates mechanical ditching to enhance finish-line visibility and facilitate contouring [11,12,14], yet optical-scanning research and manufacturer guidance emphasize the need to use geometrically stable dies and warn that deep grooves and loss of surrounding anatomy can reduce scanner trueness, complicate alignment, and force software interpolation of missing data [24,26–29]. Existing investigations have focused primarily on scanner type, stone material, and finish-line configuration [5,22,23,25–29], with virtually no quantitative evidence on how conventional ditching versus non-ditching of removable dies affects the coordinate accuracy of the digitized model or the marginal fit of CAD/CAM crowns relative to the original tooth. Accordingly, there is a clear need for a controlled in vitro study that isolates the effect of die ditching within an otherwise standardized CAD/CAM workflow. The present investigation was thus aimed and designed to compare ditched and non-ditched dies derived from the same preparation, using three-dimensional superimposition of STL files to evaluate coordinate deviations and marginal discrepancies at predefined landmarks and to interpret these values against established thresholds for clinical acceptability. By clarifying whether conventional ditching improves, worsens, or does not influence the marginal fit of digitally fabricated crowns, this study aims to provide evidence-based guidance for analog-digital hybrid protocols and to resolve the current inconsistency between traditional laboratory routines and modern digital recommendations.

Material and Methods

Ethics

This in vitro study received institutional ethics approval (College of Dentistry, Jazan University; Approval no: CODJU-2415I).

Study Design

This was an experimental in vitro comparative study with randomized allocation of dies to ditched (Gp D) and non-ditched (Gp N) groups, using the prepared typodont tooth (Gp T) as a fixed reference to assess the effect of die ditching on the marginal fit of digitally fabricated crowns. The individual fabricated digital crown designs (n=10 per group) served as the primary independent statistical unit for all comparative analyses.

Sample Size Estimation

Sample size was calculated for a 3-group comparison using G*Power 3.1.9.7 (Universität Kiel, Germany), based on a medium effect size of 0.35 (derived from Cohen’s benchmarks for F-tests), α = 0.05, and power of 0.80. The primary outcome defined for this estimation was the mean marginal root mean square (RMS) deviation. Although the prepared typodont (Gp T) serves as the geometric reference, it was included as a formal comparison group (n=10 digital crown designs) to establish the baseline variance inherent in the digital scanning and design workflow. Including the typodont-derived crowns (Gp CT) as a separate comparison group in the power analysis ensured that variance attributable to the scanner and design process alone, independent of die processing, was incorporated into the effect size estimation. This resulted in a total sample of N=30 digital crown designs (n=10 per group), which was deemed sufficient to detect significant differences in marginal fit.

Specimen preparation

Typodont Selection and Tooth Preparation Standardization

Table 1 presents the list of materials used and their respective manufacturers and specifications as declared by the manufacturer and current literature. A standard complete veneer crown preparation was performed on a mandibular first molar chosen from a detachable PVC typodont model (Foshan Tanton Technology Co., Ltd) (Figure 1A). The typodont has realistic occlusal anatomy, ideal occlusal contacts, detachable individual teeth, a soft gingival mask, and a strong PVC material. The Shofu Crown & Bridge Preparation Kit (Shofu, Inc., Japan) was used to accomplish tooth preparation using a standard air-driven high-speed handpiece with coolant (distilled water). Recommended diamond points and depth-cutting points were used during the tooth preparation. The tooth preparation followed academic textbook guidelines and recommendations for a full veneer preparation, which had preparation features of 1.2 mm axial reduction, 1.5 mm (nonfunctional cusp), 2.0 mm (functional cusp) occlusal reduction, a supragingival 0.5–1.0 mm chamfer finish line and total occlusal convergence of 6° to 8° [30,31] (Table 1). All point and line angles were rounded and smooth with no visible undercuts in the preparation. The technique was performed by a single operator who was unaware of the study outcome and die processing procedures. The prepared tooth was evaluated by a team of 2 experts who provided their consensus after evaluating the tooth preparation based on a rubric. Prior to the definitive preparation, the operator completed pilot preparations on typodont molars using the same kit and rubric to standardize technique, and the 2 evaluators reached consensus on preparation quality to minimize operator-dependent variability.

Table 1.

List of instruments and materials used.

Material	Manufacturer	Specifications and features
Typodont models	Foshan Tanton Technology Co. Foshan, China	Key features: Realistic anatomy, correct occlusion. Teeth removable/replaceable. Soft gingiva, Durable PVC resistant to high-speed handpiece cutting. Product: Typodont Detachable 28 PVC Teeth Model (multi-purpose articulated anatomy model). Type: Dental teaching/training typodont with removable teeth
Crown & Bridge Preparation Kit	Shofu, Inc., Japan	Type: FG diamond bur kit for crown and bridge, veneer and onlay preparation Kit composition: 14 FG diamond points in 3 grit sizes (Regular, Fine, Superfine). Key features: Designed for tooth preparation for crowns, bridges. Depth-cutting diamond points, 2 tissue-protective end-cutting burs, autoclavable aluminum bur block for storage and sterilization
Castone Dental Stone (Type III gypsum)	Dentsply Sirona (formerly Dentsply Prosthetics), USA	Type: ADA Type III dental stone (laboratory stone for models). Catalog sizes: 25 lb, 50 lb, 100 lb bags (cream color variants). Catalog number: 499133 for 44 lb (cream). Setting time 8–10 min, setting expansion 0.08%, compressive strength [Wet: 3000 psi, dry: 8000 psi] Key features: High-strength, low-expansion, excellent flow and detail reproduction, Minimal setting expansion
Aquasil Ultra+ Soft Putty and Extra-Low Viscosity (XLV)	Dentsply Sirona, USA	Type: Vinylpolysiloxane (VPS) / Addition silicone, Type 0 (putty consistency) Package: 2 × 450 ml (Base + Catalyst) with dosing spoons; Mixing time 30 seconds; working time 1: 30 min, dimensional stability < 0.05%; Total setting time 4: 30–5: 00 min; Permanent deformation ≤ 0.5%; working time 1: 10 min, set time 5: 00 min, Linear dimensional change < 0.05%, Key features: Smart Wetting and high tear strength
PINDEX Model and Die Fabrication System	Coltène/Whaledent AG, Altstätten, Switzerland	Type: Precision pin drilling system Drilling unit – PINDEX Laser (manual drilling, laser positioning); Pins - Brass pins: Short (10 mm, grey sleeve), Medium (16 mm, brown sleeve), Long (20 mm, white sleeve), Dual-Pin/Tri-Plus (anti-rotation). Sleeves – Plastic (adjustable fit) or metal; color-coded by pin length. Drills Carbide PX-119 (3 pcs/set), 1.8 mm diameter; parallel channels
3Shape E Series Lab Scanner	3Shape A/S, Copenhagen, Denmark	Type: Blue LED multiline structured light scanner Model E4. Key features: 4 × 5 MP cameras, 4 μm accuracy per ISO 12836, 9 sec full arch scan speed, color texture capture, blue LED multiline structured light technology
Medit Design software	Medit Corp., Seoul, South Korea	Version 2.1.4; Measurement Metrics (calculates deviations using the root mean square 2 superimposed scans D mesh analysis and comparison application with superimposition, measurement (distance/area/angle), alignment (manual/automatic), editing tools (bridge creation, mesh resolution control, mirroring, offset/thicken/Boolean operation Prodns)

Abbreviations: PVC, polyvinyl chloride; ADA, American dental association; lb, pounds; min, minute; %, percent; ml, milliliter; mm, millimeter; XLV, Extra-Low Viscosity; LED, light emitting diode; MPa, megapascal; sec, seconds.

Figure 1.

(A) Prepared first molar tooth for single crown (B) Final impression (C) Complete STL image of ditched model (D) Complete STL image of un-ditched model (E) Comparison of 2 crown fabricate with ditched model and control models (F) Superimpose of ditched and control models (G) Superimpose of 2 crown fabricate with diched model and control models. Figure created using MS PowerPoint, version 20H2 (OS build 19042,1466), Windows 11 Pro, Microsoft corporation).

Working Model Fabrication

Two distinct procedures were impression procedure and PINDEX die system processing. A single full-arch impression of the prepared typodont was made using Aquasil Ultra Soft Putty (regular set) in a stock tray combined with Aquasil Ultra Extra-Low Viscosity (XLV) wash (Dentsply Sirona, USA) (putty reline impression technique) [30,32], mixed 1: 1 by volume (putty hand-mixed for 30 seconds; XLV auto-mixed in Pentamix cartridge), seated with firm finger pressure for at least 3 minutes, and allowed to polymerize for 4.5 to 5 min to ensure dimensional change below 0.05% and permanent deformation ≤ 0.5% [31,32] (Table 1, Figure 1B). Immediately after setting, the impression was rinsed, air-dried, and inspected under magnification to verify complete reproduction of the finish line and absence of tears or voids; defective impressions were discarded and remade using the same protocol to maintain a single, high-quality master impression. Each approved impression was then immersed in 0.5% to 1% sodium hypochlorite solution for 10 minutes as recommended for elastomeric impressions, ensuring complete coverage of all surfaces. Following immersion disinfection, the impression was again rinsed with water, gently air-dried to avoid pooling solution in critical areas, visually inspected for dimensional distortion or surface defects, and then poured with Castone Type III dental stone to fabricate the working cast. The working cast was poured within 30 minutes using a Castone Type III dental stone (Dentsply Sirona, USA), mixed with distilled water at the manufacturer’s powder-to-water ratio under vacuum, vibrated slowly into the impression to minimize air entrapment, and allowed to set for at least 45 minutes (initial set 8–10 min, low setting expansion 0.08%) before separation (Table 1). The master cast base was trimmed perpendicular to the long axis of the teeth and then processed with the PINDEX Model and Die Fabrication System (Coltène/Whaledent, Altstätten, Switzerland) by drilling 1.8-mm parallel channels from the base using the PINDEX Laser unit, inserting brass pins (short/medium/long, color-coded sleeves) centrally in each abutment, and securing them with cyanoacrylate to permit precise repositioning. After pin placement, the cast was sectioned from the occlusal aspect using the PX-200 handsaw (0.18–0.25 mm blades) to create removable dies (Figure 1C), ensuring that the saw kerf did not encroach upon the finish line; each die was then carefully separated, lightly trimmed to remove excess stone, and re-seated in the parent cast to confirm passive, repeatable seating without rocking, forming the standardized working model set for subsequent non-ditched and ditched die processing [14,30,31].

Die Processing and Randomization to Die Ditching

After fabrication of the PINDEX working cast, 20 removable dies representing the prepared mandibular first molar were identified and sequentially numbered from 1 to 20, then randomly allocated into 2 equal experimental groups (non-ditched die: Gp N, ditched die: Gp D) using a computer-generated randomization list, ensuring n=10 dies allocated per condition before subsequent subdivision for crown fabrication and coordinate analysis. Die ditching followed classical prosthodontic guidelines [14,30,31], in which the cervical sulcus around the finish line was mechanically deepened to enhance margin visibility and facilitate contouring (Figure 1C). In this study, 1 trained dental technician, blinded to measurement outcomes, performed controlled ditching on all Gp D dies using standardized recommended hand instrument and magnification, while Gp N dies received only minimal trimming without sulcus deepening, consistent with manufacturer and optical-scanning recommendations that emphasize geometrically stable dies for CAD/CAM workflows (Figure 1D). The dental technician was trained on a series of trial dies until consistent ditch depth and angulation were achieved under magnification, and only then proceeded with the study specimens under the same protocol. The magnitude and configuration of ditching in Gp D were based on traditional protocols [14], in which a shallow V-shaped groove extending approximately 0.5 to 1.0 mm apical to the finish line is created to expose the margin, whereas Gp N dies intentionally preserved surrounding cervical anatomy in line with contemporary evidence discouraging aggressive ditching because it may distort emergence profile and compromise scanner trueness. All processed dies (Gp N and Gp D) were re-seated into the parent cast to confirm passive, repeatable seating, labeled only with anonymized codes, and then forwarded to the digital operator so that subsequent scanning, crown design, and marginal fit measurement were carried out under blinded conditions with respect to ditching status.

Sample Grouping

The study used a hierarchical grouping structure that differentiated between reference geometry, die ditching, and corresponding digital crown designs. The primary groups were the prepared typodont tooth (Gp T), the non-ditched stone dies (Gp N), and the ditched stone dies (Gp D), all derived from the same standardized mandibular first molar preparation. For evaluation of restoration behavior, each model group had an associated digital crown subgroup, yielding 3 crown categories: Gp CT (crowns designed directly from the typodont scan), Gp CN (crowns designed on non-ditched dies), and Gp CD (crowns designed on ditched dies). This structure allowed the margin geometries of ditched and non-ditched dies to be compared with the typodont-prepared tooth geometries at 8 predefined landmarks, facilitating cross-group analyses (Gp T vs Gp N vs Gp D) to isolate the specific influence of die ditching on coordinate accuracy and digital marginal fit. Each die subgroup and each digital crown subgroup comprised 10 unique specimens (n=10). Each digital crown within these groups (n=10) contributed 8 landmark measurements, which were treated as nested observations within the independent specimen to account for the repeated-measures structure of the data.

Digital Workflow

3Shape E4 Scanning and Medit Design Crown Fabrication

All physical models (prepared typodont and stone dies) were digitized with a 3Shape E-series E4 lab scanner (3Shape A/S, Copenhagen, Denmark) after daily calibration using the manufacturer’s reference block. Separate scans were obtained only for the prepared typodont reference (Gp T) and for each non-ditched (Gp N) and ditched (Gp D) die, which were mounted on the scanner plate to capture the full crown preparation and adjacent reference anatomy (Figure 1E, 1F). The resulting scan data were exported as STL files and imported into Medit Design software v2.1.4 (Medit Corp., Seoul, South Korea), where the typodont STL was defined as the reference model and the die STLs were aligned using automatic best-fit registration with manual refinement as needed (Figure 1G). Using these aligned datasets, 3 crown designs were generated digitally in Medit Design: Gp CT crowns based on the typodont scan, Gp CN crowns based on non-ditched die scans, and Gp CD crowns based on ditched die scans. Each full-coverage crown was designed with uniform thickness (approximately 1.0 mm axial, 1.5–2.0 mm occlusal), standardized occlusal morphology, and a consistent emergence profile. The crown designs were then exported as separate STL files for subsequent milling and 3D superimposition analysis.

Crown Milling and Sintering (CAD/CAM Full-Metal Veneer Crown Production Protocol)

Full-metal veneer crowns were produced as monolithic cobalt-chromium (Co-Cr) restorations using a CAD/CAM milling-sintering workflow in which the approved crown designs were nested in pre-sintered soft Co-Cr blanks and milled on a 5-axis milling unit, following the alloy manufacturer’s recommended bur sequence, spindle speed, and dry-milling parameters to minimize tool wear and preserve marginal integrity [33]. After milling, each oversized green-stage Co-Cr crown was separated from the blank, sprues were trimmed without contacting the finish line region, and the restorations were placed on sintering trays and sintered at approximately 1300 to 1350 °C under protective gas to full density (≈ 10% volumetric shrinkage), a protocol shown to yield homogeneous, distortion-free Co-Cr frameworks with marginal discrepancies comparable to conventionally cast or laser-sintered metal-ceramic crowns [34]. Following sintering, crowns were allowed to cool slowly within the furnace, then inspected under magnification for defects, lightly finished and polished only on non-critical external surfaces, ultrasonically cleaned in distilled water, and stored dry until fit evaluation, ensuring that any differences in marginal accuracy reflected die ditching condition rather than variability in the CAD/CAM metal manufacturing process [33].

3D Superimposition Analysis (Marginal Fit Assessment at Predefined Landmarks)

The primary outcome measure was the mean root mean square (RMS) marginal deviation per digital crown design. Individual landmark measurements were recorded as secondary, exploratory data points to map position-specific errors, such as those at buccal and lingual locations. Statistical testing at the level of individual landmarks was therefore considered secondary and exploratory, with interpretation anchored primarily on the per-crown mean RMS deviations. Marginal fit was evaluated by importing all STL datasets (prepared typodont, non-ditched dies, ditched dies, and the 3 crown subgroups Gp CT, Gp CN, Gp CD) into Medit Design v2.1.4, designating the prepared typodont model as the reference, and superimposing each corresponding crown restoration (Gp CT, Gp CN, or Gp CD) onto its reference geometry (Gp T, Gp N, or Gp D) using automatic best-fit alignment followed by manual verification to ensure accurate global registration (Figure 2). Eight predefined marginal landmarks were identified on the reference model (mesio-lingual, mesial, mesio-buccal, buccal, disto-buccal, distal, disto-lingual, and lingual). For each digital crown design, the 8 landmark-specific RMS values were averaged to obtain a single mean marginal RMS deviation per crown, which served as the primary quantitative outcome for all between-group comparisons. These specific individual landmarks were then transferred to each superimposed pair. Medit Design’s Root Mean Square (RMS) distance tool was used to calculate the absolute 3D deviation (mm) between the crown margin and the surface, generating color-coded maps and numerical values for subsequent statistical analysis. For each landmark and subgroup, the distribution of marginal discrepancies (median, IQR, minimum, maximum, mean rank score) was computed. These RMS-based deviations were then additionally interpreted against the 120-μm clinical acceptability threshold to classify gaps as clinically acceptable or unacceptable [19].

Figure 2.

Flowchart showing study design, hypothesis, study sequence and division of groups based on independent variables. Figure created using MS PowerPoint, version 20H2 (OS build 19042,1466), Windows 11 Pro, Microsoft corporation).

Statistical Analysis

Data distribution for marginal discrepancy values at each of the 8 landmarks was reviewed via normality tests. Because variables were non-normally distributed with heterogeneous variances, non-parametric statistics were adopted as the primary analytical approach, and only these non-parametric P values were used for inferential decision-making, with Welch t tests and Cohen’s d reserved for descriptive sensitivity analyses. The independent statistical unit was defined as the individual digital crown design (n=10 per group), and the primary outcome was the mean marginal RMS deviation. To account for the repeated-measures structure of the 8 landmarks nested within each digital crown, a Friedman test was utilized to evaluate positional variance. For the primary 3-group comparison among Gp CT, Gp CN, and Gp CD, a Kruskal-Wallis one-way ANOVA on ranks was applied. Pairwise differences between groups through Dunn’s post hoc test were analyzed only when the primary test indicated significance. Direct comparisons of marginal discrepancies between non-ditched and ditched groups were further evaluated using Mann-Whitney U tests, while parametric Welch t tests and Cohen’s d were calculated solely as sensitivity analyses to assess effect magnitude. Clinical interpretability was established by benchmarking all median marginal gap values against the 120-μm clinical acceptability threshold; deviations ≤ 120 μm were classified as clinically acceptable. Differences in the probability levels were considered significant if the P values were equal to or less than 0.05 (P ≤ 0.05).

Results

Overall Marginal Fit (Primary Endpoint)

The principal analysis of the pooled data across all 8 different landmarks revealed that the overall mean marginal discrepancy was significantly higher for crowns measured on ditched dies (0.147 ± 0.200 mm) compared to those on non-ditched dies (0.086 ± 0.070 mm; P = 0.0127) (Table 2). The standardized effect size (Cohen’s d) for this difference was −0.407, with a medium effect size. Furthermore, the RMS error, representing total three-dimensional misfit, was 0.111 mm for non-ditched casts and 0.247 mm for ditched casts (Table 2).

Table 2.

Comprehensive marginal fit analysis – parametric and non-parametric comparisons with effect sizes and clinical acceptability across 8 positions (non-ditched vs ditched dental casts).

Position	M ± SD (Gp N)	M ± SD (Gp D)	T test (P value)	Cohen's d	MW U (P value)	% ≤ 120μm (N/D)	Friedman Pos
Mesio-lingual	0.050 ± 0.035	0.086 ± 0.065	−1.542 (0.1456)	−0.690	35.0 (0.2712)	100/80	χ2 = 9.00*
Mesial	0.073 ± 0.091	0.192 ± 0.283	−1.265 (0.2323)	−0.566	32.5 (0.1964)	80/60	χ2 = 9.00*
Mesio-buccal	0.083 ± 0.047	0.189 ± 0.288	−1.144 (0.2808)	−0.514	55.0 (0.7328)	90/80	χ2 = 9.00*
Buccal	0.089 ± 0.057	0.175 ± 0.235	−1.127 (0.2858)	−0.503	39.0 (0.4258)	70/50	χ2 = 9.00*
Disto-buccal	0.088 ± 0.054	0.095 ± 0.194	−0.111 (0.9134)	−0.049	73.5 (0.0805)	70/90	χ2 = 9.00*
Distal	0.090 ± 0.053	0.124 ± 0.118	−0.824 (0.4253)	−0.372	47.0 (0.8498)	70/50	χ2 = 9.00*
Disto-lingual	0.141 ± 0.129	0.161 ± 0.091	−0.400 (0.6944)	−0.179	28.0 (0.1016)	70/50	χ2 = 9.00*
Lingual	0.077 ± 0.033	0.152 ± 0.236	−0.983 (0.3502)	−0.445	55.0 (0.7312)	100/70	χ2 = 9.00*
Overall	0.086 ± 0.070	0.147 ± 0.200	−2.540 (0.0127)*	−0.407	2929 (0.3552)	81/66

Abbreviations: N, non-ditched; D, ditched; t, Welch t-statistic (df ≈ 18); MW U, Mann-Whitney U (two-tailed); Cohen’s d, effect size; %, percentage; P, probability value; M, mean; SD, standard deviation.

Statistical tests: Cohen’s d = standardized effect size measuring the magnitude of difference between 2 group means in standard deviation units. Cohen’s d scale = E 0.2 [Small; Disto-buccal (−0.049)]; 0.5 [Medium; Mesial (−0.566) ^**]; 0.8 [Large; None]; standard threshold 0.50 (Medium-Clinically noticeable]. Cohen’s d: negative = higher in ditched; |d| > 0.5 = medium effect”^*. MW U (P value):

Mann-Whitney U test; lower U favors non-ditched.

% ≤ 120 μm (N/D) = Percentage of measurements ≤ 120 micrometers (μm) at each position for Non-ditched (N) vs Ditched (D) groups. Clinical Interpretation: 120 μm = clinically acceptable marginal gap threshold (McLean & von Fraunhofer, 1971); ≤ 120 μm: Acceptable (prevents plaque, caries, periodontitis); > 120 μm: Unacceptable (risks complications).

Friedman Pos = χ² statistic from Friedman test - a non-parametric repeated-measures ANOVA testing if deviations differ significantly across the 8 marginal positions within each group. Friedman Pos: χ² from Friedman test across 8 positions within group (P = 0.44 both groups); Clinical interpretation: Ditching affects overall fit level (t test sig.) but not position pattern consistency (Friedman non-sig.).

Statistical significance:

^*P < 0.05;

negative d/t favors non-ditched; χ² = 9.00 (P = 0.44 both groups, non-sig.). RMS: Non-ditched 0.111 mm; Ditched 0.247 mm.

Clinical Acceptability and Distribution

Using a 120-μm threshold for clinical acceptability, 81% of all measurements in the non-ditched group were within the acceptable range, whereas this proportion decreased to 66% in the ditched group. At the level of individual positions, the non-ditched group demonstrated 70% to 100% acceptability, while the ditched group ranged from 50% to 90%. The Friedman test indicated that the pattern of gaps across the 8 positions did not differ significantly within either group (χ² = 9.00; P = 0.44), indicating that the magnitude of discrepancies increased overall rather than altering the site-specific pattern (P ≤ 0.05) (Table 2).

Site-Specific Comparison (Secondary Endpoints)

While the overall discrepancy was significant, individual position-wise comparisons between non-ditched (Gp N) and ditched (Gp D) casts did not reach statistical significance (all t test P > 0.14, all Mann-Whitney P > 0.08) (Table 2). However, the Cohen’s d values for several sites fell within the small-to-medium range (−0.37 to −0.69), particularly at the mesio-lingual and mesial positions (Table 2). When comparing the crown restoration margin directly to the reference typodont (Gp CT) and versus the ditched die (Gp CD), the buccal landmark was the only site to show a statistically significant difference (P ≤ 0.05) (Table 3). Post hoc Dunn tests with Bonferroni correction (corrected α = 0.0167) confirmed this significant difference at the buccal site (P = 0.0152), while all other crown-to-model comparisons (CT vs CN or CT vs CD) remained non-significant (Table 4).

Table 3.

Comparative differences in median values of predetermined surface landmarks between the crown restoration margin and typodont model/ditched/non-ditched die.

Landmark	Subgroups	N	Median	IQR	Minimum	Maximum	MRS	H Statistic	P value
Mesio-lingual	Gp CT	10	0.020	0.004	0.015	0.09	20.75	5.48	0.06452
	Gp CN	10	0.011	0.01	0.01	0.08	12.55
	Gp CD	10	0.016	0.01	0.009	0.09	13.2
Mesial	Gp CT	10	0.120	0.005	0.11	0.19	17.7	1.9357	0.3799
	Gp CN	10	0.036	0.105	0.004	0.3	12.45
	Gp CD	10	0.159	0.251	0.003	0.352	16.35
Mesio-buccal	Gp CT	10	0.013	0.009	0.01	0.07	15.0	0.0571	0.9718
	Gp CN	10	0.012	0.04	0.009	0.1	15.6
	Gp CD	10	0.030	0.04	0.003	0.12	15.9
Buccal	Gp CT	10	0.003	0.0005	0.001	0.006	10.1	6.2161	0.0446*
	Gp CN	10	0.012	0.026	0.001	0.04	16.8
	Gp CD	10	0.025	0.07	0.002	0.25	19.6
Disto-buccal	Gp CT	10	0.050	0.011	0.01	0.055	17.95	4.1605	0.1249
	Gp CN	10	0.050	0.076	0.009	0.151	17.65
	Gp CD	10	0.024	0.041	0.002	0.07	10.9
Distal	Gp CT	10	0.030	0.005	0.01	0.09	17.1	0.6926	0.7073
	Gp CN	10	0.030	0.04	0.003	0.08	15.55
	Gp CD	10	0.022	0.045	0.003	0.068	13.85
Disto-lingual	Gp CT	10	0.05	0.012	0.01	0.057	15.15	4.0635	0.1311
	Gp CN	10	0.09	0.1	0.009	0.5	19.6
	Gp CD	10	0.017	0.035	0.01	0.11	11.75
Lingual	Gp CT	10	0.022	0.0028	0.015	0.029	16.2	0.0961	0.9531
	Gp CN	10	0.022	0.023	0.002	0.1	15.15
	Gp CD	10	0.018	0.023	0.009	0.28	15.15

Abbreviations: Gp, group; N, number of specimens; IQR, Interquartile range; H, difference between 2 or more groups of an independent variable on a continuous dependent variable; P, probability value; MRS, mean rank score.

Interpretation of Groups: C = crown restoration, T = typodont teeth dimensions, N = non-ditched die, D = ditched die.

Test employed – One-way ANOVA on ranks (Kruskal-Wallis H test); Statistical Interpretation: level of the degree of significance was determined on the value of P < 0.05;

^*significant.

Table 4.

Post hoc (Dunn test) results of median values showing multiple group (between subgroups) comparisons between crown restoration, typodont model, ditched and non-ditched die.

	Comparison subgroups (CT, CN, CD)			Comparison subgroups (T, N, D)
	Subgroups	MRD	P value	Subgroups	MRD	P value
Mesio-lingual	Gp CT-Gp CN	8.2	0.0352	Gp T− Gp N	−6.25	0.1073
	Gp CT-Gp CD	7.55	0.0525	Gp T− Gp D	−10.7	0.0058*
	Gp CN-Gp CD	−0.65	0.8674	Gp N− Gp D	−4.45	0.2515
Mesial	Gp CT-Gp CN	5.25	0.1804	Gp T− Gp N	−11.3	0.0038*
	Gp CT-Gp CD	1.35	0.7305	Gp T− Gp D	−14.8	0.0001*
	Gp CN-Gp CD	−3.9	0.3197	Gp N− Gp D	−3.5	0.3702
Mesio-buccal	Gp CT-Gp CN	−0.6	0.8756	Gp T− Gp N	−6.95	0.0733
	Gp CT-Gp CD	−0.9	0.8144	Gp T− Gp D	−3.55	0.3603
	Gp CN-Gp CD	−0.3	0.9376	Gp N− Gp D	3.4	0.3809
Buccal	Gp CT-Gp CN	−6.7	0.0870	Gp T− Gp N	13.3	0.0006*
	Gp CT-Gp CD	−9.5	0.0152*	Gp T− Gp D	10.1	0.0095*
	Gp CN-Gp CD	−2.8	0.4746	Gp N− Gp D	−3.2	0.4115
Disto-buccal	Gp CT-Gp CN	0.3	0.9388	Gp T− Gp N	−9.4	0.0153*
	Gp CT-Gp CD	7.05	0.0713	Gp T− Gp D	−1.7	0.6611
	Gp CN-Gp CD	6.75	0.0842	Gp N− Gp D	7.7	0.0470
Distal	Gp CT-Gp CN	1.55	0.6915	Gp T− Gp N	10.35	0.0076*
	Gp CT-Gp CD	3.25	0.4055	Gp T− Gp D	6.75	0.0821
	Gp CN-Gp CD	1.7	0.6635	Gp N− Gp D	−3.6	0.3538
Disto-lingual	Gp CT-Gp CN	−4.45	0.2546	Gp T− Gp N	3.95	0.306
	Gp CT-Gp CD	3.4	0.384	Gp T− Gp D	−3.95	0.306
	Gp CN-Gp CD	7.85	0.0444	Gp N− Gp D	−7.9	0.0406
Lingual	Gp CT-Gp CN	1.05	0.7883	Gp T− Gp N	6.3	0.0947
	Gp CT-Gp CD	1.05	0.7883	Gp T− Gp D	6.45	0.0871
	Gp CN-Gp CD	0	1	Gp N− Gp D	0.15	0.9683

Abbreviations: Gp, group; P, probability value; MRD, mean rank differences.

Interpretation of Groups: C = crown restoration, T = typodont teeth dimensions, N = non-ditched die, D = ditched die.

Test employed – Dunn Test; P < 0.05;

^*significant. Bonferroni correction - Corrected α = α / m = 0.05 / 3 = 0.01667.

Model Accuracy and Deviation

Direct comparison between the reference typodont (T) and the produced stone dies (N and D) showed significant deviations at several sites. Significant differences were noted at the mesial (P = 0.0038), buccal (P = 0.0006), and distal (P = 0.0076) positions when comparing the typodont to non-ditched dies. In the ditched group, the mesio-lingual (P = 0.0058) and mesial (P = 0.0001) positions also showed significant deviations from the typodont dimensions (Table 4).

Discussion

This invitro study evaluated the influence of die ditching on the crown coordinate deviations and associated marginal discrepancies during the fabrication of a full veneer cobalt-chrome crown restoration prepared through digital workflow. The findings of this study indicates that within this tested CAD/CAM workflow, die ditching was associated with a numerical increase in the marginal gaps of Co-Cr crowns when compared to the crowns that did not incorporate any ditching in its dies. While an overall trend of diminished fit was observed, statistical significance among the crown subgroups was localized to the buccal landmark (P = 0.0152), where medians increased from 0.003 mm (typodont) to 0.025 mm (ditched die). This highlights a site-specific vulnerability to analog distortions. However, the non-significant Friedman test (P = 0.44) suggests that ditching tended to elevate baseline discrepancies uniformly across the measured positions rather than creating isolated new hotspots of error. These results are broadly comparable to marginal gaps reported for most CAD/CAM and 3D-printed crowns in the literature, though direct comparisons are limited by varying measurement protocols and materials.

Mechanisms of Distortion

The localized distortions at the mesial, buccal, and distal landmarks between the typodont and the stone dies primarily stem from the physical properties of the materials and the preparation geometry. The use of Type III gypsum, with a setting expansion of approximately 0.08%, and the elastic recovery of the VPS impression, can disproportionately affect convex and proximal sites. This is corroborated by previous stone accuracy work showing that dies can be 0.1% to 0.9% larger or smaller proximally versus at other oral surfaces [35,36]. In this study, the buccal landmark’s vulnerability was highlighted by medians escalating from 0.003 mm (typodont) to 0.012 mm (non-ditched die) and finally to 0.025 mm (ditched die). This progressive deviation at some landmarks advocates that the convexity of the mandibular molar benefits the initial typodont scan but complicates subsequent stone replication and the manual die ditching process, as it has the potential to create an uneven ditch, which potentially leads to the margin errors during digitization.

Comparison with Methodologically Comparable Studies

Our findings regarding buccal bias are consistent with other site-specific analyses in the literature. Renne et al [37] reported a buccal mean gap of 79.32 μm versus a lingual mean of 50.39 μm for chairside CAD/CAM crowns, mirroring the directional trend found here. Similarly, an implant crown study on mandibular first molars reported the largest gaps at buccal sites (9.897 ± 5.755 μm), which were significantly greater than lingual and distal sites (P < 0.05) [40]. Kricheldorf et al [38] also found significant differences localized to the buccal (P = 0.012) and lingual (P = 0.016) sites when comparing CAD/CAM systems. While these studies confirm buccal marginal weakness, direct comparisons are limited by varying measurement protocols. For instance, the use of 3D STL superimposition (Medit RMS) in this study allows for precise tracking of analog-to-digital increments that may be absent in studies using post-cementation replica images or digital microscopy [18,37]. Such difference of methodology (eg, gap width vs RMS deviation) can yield a non-comparable value/s, highlighting the fact that the reported fit is highly dependent on the method employed [10,13,15].

Clinical and Laboratory Implications of Ditching

Traditional prosthodontic guidelines advocate for cervical ditching to improve finish-line visibility [14,30,31]. However, in this modern digital application, the standardized ditching protocol was associated with a notable reduction in margins meeting the 120 μm success threshold (dropping from 81% to 66%). This emphasis on seemingly small technical deviations along the treatment chain is in line with our previous clinical work on fixed prosthodontics [39], where radiographic evaluation of post–core restorations highlighted that minor departures from ideal preparation and laboratory procedures can measurably affect the technical quality and long-term prognosis of fixed prosthodontic reconstructions. This aligns with Sabouhi et al [12], who found that technician ditching produced larger vertical marginal gaps (436 ± 318 μm) than standardized methods in cast restorations. Although our CAD-milled Co-Cr crowns achieved a superior overall fit (86–147 μm means) due to the controlled digital workflow, the unique degradation effect of ditching remained evident. Our non-ditched dies achieved a mean marginal gap of 86 μm, aligning closely with results for cast Co-Cr (76 ± 61 μm) and with the marginal adaptation of CoCrMo crowns fabricated using computer-aided technologies reported in previous studies [33,36,40]. However, once ditching was performed, the gaps increased to 147 μm, approaching the upper limits of clinical acceptability.

Limitations and Future Directions

It must be noted that this study utilized an in vitro design, 1 specific mandibular molar preparation, and a single scanner-material chain. The results may vary with different tooth geometries (such as anterior preparations) or alternative scanner technologies. Extending these investigations to integrate detailed morphometric assessments of tooth and root anatomy, as previously performed in radiographic analyses of root-to-crown ratios [41], could help identify whether specific anatomical configurations are more susceptible to analog-digital distortions during die processing and scanning. Future research should explore whether “minimal trimming” provides a more balanced outcome than the standardized ditching protocol tested here, and should include longitudinal clinical studies to determine if these numerical differences translate into meaningful differences in periodontal health and restoration survival.

Conclusions

From a biological perspective, various studies indicate that both marginal fit and cervical contour influences the short and long-term health of periodontal tissues, with over-contoured cervical areas and deep artificial gingival sulci providing ideal places for plaque retention and gingival inflammation. Previous clinical work has indicated that over-contoured mandibular molar crowns can lead to higher plaque and gingival index scores compared to under-contoured restorations, underscoring the potential importance of accurate emergence profile reproduction. The findings of this in vitro study suggest that die ditching, within the tested workflow, was associated with an increase in marginal gap size and the creation of a deeper artificial sulcus. These factors could theoretically act in combination, increasing the risks of both geometric overbulk and microleakage, which might influence long-term periodontal and restorative outcomes. However, given the limited scope of this study, further research is required to explore graded levels of ditching depth and angulation, as well as alternative scanner technologies and die materials. Extending this analysis to full-ceramic and hybrid restorations would help determine whether the observed influence of ditching is consistent across different material systems. Furthermore, longitudinal clinical or fatigue studies are necessary to correlate 3D misfit data with clinical parameters such as plaque indices, secondary caries, and restoration survival. Such research would clarify whether the numerical differences observed in this study between non-ditched and ditched workflows translate into clinically meaningful differences in patient outcomes.