Three-Dimensional Accuracy of Clear Aligner Attachment Reproduction Using a Standardized In-House Protocol: An In Vitro Study

Abstract

This in vitro study aimed to quantitatively evaluate the accuracy of reproducing attachments for clear aligner therapy (CAT) using a standardized in-house fabrication protocol and to analyze discrepancies across maxillary tooth types. A custom attachment was designed on a symmetrical master model, and 30 experimental models were fabricated by three-dimensional (3D) printing, template construction, and bonding. Following scanning and superimposition, dimensional, angular, and positional deviations were quantified and statistically analyzed (p < 0.05). Results showed minor mean discrepancies but a consistent pattern of under-reproduction, most evident in the mesial and distal wall angles, as well as in the gingival bevel angle and attachment height. A significant trend was observed in the occlusal bevel, demonstrating marked extrusion in the anterior region that decreased posteriorly. Positional errors were minimal mesiodistally but substantial in the lingual and occlusal directions, with magnitudes varying by tooth type. In conclusion, this study identified consistent, predictable inaccuracies in a simulated in-house attachment reproduction protocol. These findings indicate that similar deviations may occur clinically, potentially affecting the predictability of CAT.

1. Introduction

The popularity of clear aligners stems from their advantages in aesthetics and comfort [1]. However, their early clinical use was limited by reduced efficacy in executing complex tooth movements. To enhance biomechanical performance, several innovations have been introduced, with the strategic incorporation of attachments being the most significant [2].

Attachments are critical components that improve aligner biomechanics by enhancing retention and fit, enabling directed forces on target teeth [3]. Advances in three-dimensional (3D) digital technology have further refined this process, allowing clinicians to virtually define the size, shape, and position of each attachment to facilitate specific, complex tooth movements [4,5].

Studies highlight the significance of attachment design in treatment efficacy. Larger attachments with sharper edges better achieve programmed tooth movements [6], while optimized horizontal and rectangular designs improve mesiodistal control [7]. Placement is equally critical: attachments are most effective when positioned gingivally for extrusion, centrally for intrusion, and incisally for torque [8]. Therefore, treatment success depends on both precise design and accurate clinical placement.

Despite these advances, attachment reproduction remains prone to error. Potential inaccuracies may arise during 3D printing of setup models, thermoforming of templates, and manual bonding procedures [5,9]. Collectively, such discrepancies compromise the fidelity of attachment reproduction in both shape and position, potentially jeopardizing outcomes [10].

Although prior studies have examined the influence of geometry and material properties on treatment, few have directly quantified the accuracy of attachment reproduction [11,12]. This gap is increasingly relevant with the rise of in-house aligner fabrication, driven by office-based 3D printing. In these decentralized models, minimizing process errors becomes a direct clinical responsibility. Therefore, this in vitro study aimed to quantitatively evaluate the accuracy of reproducing virtually designed attachments using a standardized protocol.

2. Materials and Methods

2.1. Sample Selection

This retrospective study was approved by the Institutional Review Board (IRB) of Kyungpook National University Dental Hospital (IRB No. KNUDH-2025-07-03-00), with informed consent waived.

A digital maxillary model was selected from a patient who had completed non-extraction orthodontic treatment. Inclusion criteria were the absence of prosthetic restorations, normal tooth morphology, and good final occlusion.

To eliminate the confounding effects of bilateral asymmetry, a fully symmetrical digital master model was created. Scan data was reoriented using Meshmixer (Version 3.5.474, Autodesk Inc., San Francisco, CA, USA) so that the occlusal plane was parallel to the horizontal plane and the dental arch midline, aligned with the median palatine suture, coincided with the midsagittal plane. The right side of the reoriented model was digitally mirrored onto the left, producing a symmetric experimental base.

2.2. Attachment Design and Fabrication

The design and fabrication workflow is illustrated in Figure 1. A custom attachment with a square base with two distinct bevels was designed using Fusion (Version 2602.1.25 x86_64, Autodesk Inc., San Francisco, CA, USA). The base measured 2.5 × 2.5 mm. Attachment height, defined as the perpendicular distance from the base to the midpoint of the junction edge between the two bevels (H-mid) was set to 1.2 mm (Figure 2).

Attachments were digitally positioned in symmetrical, mirror-image locations on the bilateral maxillary lateral incisors, first premolars, and first molars of the master model. Placement was standardized both horizontally and vertically: for lateral incisors and first premolars, attachments were centered at the mesiodistal midpoint of the labial/buccal surfaces, and for first molars, on the mesiobuccal cusp. Vertically, all attachments were placed within the middle third of the respective tooth surfaces (Figure 3a,b). The base plane of each attachment was oriented parallel to the tangent of the underlying surface. This finalized virtual model, with symmetrically placed attachments, was designated the “Digital Control 3D Model”.

From these digital designs, physical models were fabricated. To ensure adequate sample size per the Central Limit Theorem, two sets of 30 models each were 3D-printed with liquid resin at 50 µm layer thickness. (Detailed specifications for all equipment and materials are provided in

Table A1. Detailed specifications for the 3D printing workflow.

Component	Parameter	Specification
3D printer	Manufacturer	Anycubic a
	Model	Photon Mono 4K
	Printing technology	LCD-based masked stereolithography
	Light source	UV LED Array (385–410 nm)
	XY axis resolution	17 × 17 µm
Liquid resin	Manufacturer	RESIONE b
	Product Name	D01S (Dental Model Resin)
Printing Parameters	Layer thickness	50 µm
	Exposure time	2.8 s per layer
Post-processing unit	Manufacturer	Anycubic a
	Model	Anycubic Wash & Cure 2.0
	Power	25 W
	Input voltage	100–240 V 50/60 Hz
	Light source	UV LED 12 pcs (450 nm)
	Washing time	3 min
	Post-curing time	5 min

^a Shenzhen Anycubic Technology Co., Ltd., Shenzhen, China. ^b Dongguan Godsaid Technology Co., Dongguan, China.

). The first set (n = 30) consisted of models without attachments, printed from the symmetrical digital master model to serve as bonding bases; the second set (n = 30) comprised models with attachments, printed from the digital control models for the attachment template fabrication.

2.3. Fabrication of the Attachment Template

A total of 30 attachment templates were fabricated using the models with attachments. Templates were thermoformed from 0.5 mm Polyethylene Terephthalate Glycol-modified (PET-G) sheets with a universal pressure thermoforming machine. (Detailed specifications for all equipment and materials are provided in

Table A2. Detailed specifications for the template fabrication workflow.

Component	Parameter	Specification
Thermoforming machine	Manufacturer	Scheu-Dental a
	Model	Biostar®
Thermoforming parameters	Heating temperature	Not applicable b
	Heating time	25 s
	Forming pressure	87 psi
	Cooling time	30 s
Thermoforming Sheet	Manufacturer	JOEL Tech c
	Product name	GOODY®
	Material	PET-G d
	Thickness	0.5 mm

^a Scheu-Dental GmbH, Iserlohn, Germany. ^b Time-controlled heating cycle. ^c JOEL Tech Inc., Seongnam, South Korea. ^d Polyethylene Terephthalate Glycol-modified.

). Each template was trimmed along the gingival margin (Figure 3c).

2.4. Attachment Bonding

The bonding protocol followed Invisalign^® guidelines (Align Technology, Tempe, AZ, USA), a widely accepted clinical standard ensuring relevance [13]. Each reservoir in the 30 templates was slightly overfilled with high-viscosity composite flowable resin (CharmFil Flow HV, Dentkist, Gunpo, Republic of Korea). The resin-filled template was then seated onto a base model without attachments, and firm occluso-gingival pressure was applied. Additional pressure was exerted with a metal spatula around each attachment to ensure adaptation and facilitate resin escape. The composite was light-cured for 20 s through the template, after which the template was carefully removed. Accessible excess composite or flash was meticulously removed with a dental explorer, completing fabrication of the experimental models (Figure 3d).

2.5. Three-Dimensional (3D) Measuring Analysis

The 30 physical experimental models with manually bonded attachments were scanned using an intraoral scanner (iTero lumina™, Align Technology) to generate digital experimental models (STL files). Both the digital control models and experimental models were imported into Geomagic Control X software (Version 2025.0.1 build no. 3, Hexagon AB, Stockholm, Sweden) for analysis.

Superimposition was conducted in two steps to ensure accuracy. First, rough pre-alignment of each experimental model to its control was performed using the software’s “initial alignment” function. This step was followed by fine superimposition on a tooth-by-tooth basis using a best-fit algorithm, with the entire surface of the clinical crown defined as the reference area, excluding the attachment surface [14].

Several parameters were evaluated to evaluate the dimensional and positional accuracy of attachment reproduction. Linear measurements were recorded with a precision of 0.001 mm, and angular measurements with 0.01 degrees.

Overall attachment accuracy was assessed using the “3D Compare” function, which calculates shape discrepancies between reference (control) and measured (experimental) surfaces, and visualizes them as a color map (Figure 4). For this study, the analysis focused on four attachment surfaces: the gingival bevel, occlusal bevel, mesial wall, and distal wall. The mean surface deviation for each was calculated and recorded.

Additional parameters were measured to evaluate the attachment’s profile deviations further. Measurement planes on experimental models were constructed by selecting relevant surfaces with a lasso tool, followed by application of a Least-Squares Method (best-fit algorithm). The bevel angle was defined as the angle between the attachment base plane of the control model and the corresponding bevel surface (gingival or occlusal) of the experimental model. The nominal angles on the control model were 38.66° for the gingival bevel and 50.19° for the occlusal bevel (Figure 5a). Wall angle was defined as the angle between the base plane of the control model and the mesial or distal wall of the experimental model, with a value of 90° for both (Figure 5b). Gingival and occlusal bevel deviations were calculated as the mean distance between the nominal bevel plane of the control model and the corresponding surface of the experimental model, with positive values indicating extrusion and negative values indicating intrusion. Attachment height was defined as the distance from the attachment base of the control model to the H-mid of experimental model (Figure 6).

The 3D positional deviation of each attachment was assessed. This deviation was determined by measuring the displacement of the H-mid on the experimental model relative to its planned position on the control model. For this analysis, the planned H-mid position on the digital control model was set as the origin (0,0,0) of a local coordinate system. Deviation of the H-mid on the experimental model was then calculated along three axes: x-axis (mesiodistal, positive = mesial shift), y-axis (buccolingual, positive = labial/buccal shift, and z-axis (occluso-gingival, positive = gingival shift) (Figure 7).

2.6. Statistical Analysis

All analyses were performed using JASP software (Version 0.95; JASP Team [2025], University of Amsterdam, The Netherlands), with statistical significance set at p < 0.05. Measurement data were grouped by tooth type: lateral incisor (G2), first premolar (G4), and first molar (G6). As the experimental models were derived from a perfectly symmetrical master model, right- and left-side measurements were treated as independent observations. Therefore, from the 30 fabricated models, a statistical sample size of 60 was obtained for each group (n = 60).

Data normality was tested using the Shapiro–Wilk test to determine the appropriate statistical method. Intra-operator reliability was assessed by re-measuring 15 randomly selected models after a two-week interval and calculating the intraclass correlation coefficient (ICC).

Measurement discrepancies among the three tooth groups (G2, G4, and G6) were compared using one-way analysis of variance (ANOVA) or, when appropriate, the Kruskal–Wallis H-test. Post hoc tests were performed for pairwise comparisons when significant differences were detected.

Separately, one-sample t-tests (or Wilcoxon signed-rank tests) were used to assess the accuracy of specific profile parameters relative to their design. These analyses were limited to the gingival bevel angle, occlusal bevel angle, and attachment height to determine whether their means differed significantly from their nominal values defined in the Computer-Aided Design (CAD) model.

The sufficiency of the sample size used in this study was verified through a power analysis using G*Power software (Version 3.1.9.7; Heinrich-Heine-Universität Düsseldorf, Düsseldorf, Germany). For the primary comparison among the three groups (ANOVA), a sample size of 53 per group was required to detect a medium effect size (f = 0.25) with 80% power at an alpha level of 0.05. For the one-sample t-tests, a minimum of 34 samples was required to detect a medium effect size (d = 0.5). Therefore, the sample size of 60 per group used in this study was considered more than sufficient for all planned statistical analyses.

3. Results

Intra-operator reliability for both linear and angular measurements was excellent, with ICC values ranging from 0.998 to 0.999.

Results for 3D deviations of the four attachment surfaces are summarized in

Table 1. Comparison of the three-dimensional (3D) surface deviation (mm) of attachment surfaces among tooth groups.

	G2 (n = 60) Mean ± SD		G4 (n = 60) Mean ± SD		G6 (n = 60) Mean ± SD		p-Value
Gingival bevel	−0.151 ± 0.045		−0.138 ± 0.035		−0.131 ± 0.045		0.067	†
Occlusal bevel	0.089 ± 0.036	A	0.034 ± 0.030	B	−0.049 ± 0.026	C	<0.001	†
Mesial wall	0.009 ± 0.028	A	−0.021 ± 0.029	B	−0.019 ± 0.032	B	<0.001	†
Distal wall	0.012 ± 0.025	A	−0.001 ± 0.025	B	−0.009 ± 0.032	B	<0.001	*

G2, lateral incisor group; G4, first premolar group; G6, first molar group; SD, standard deviation. Values in the same row with different superscript letters (A, B, C) are significantly different according to the post hoc tests. Values without superscripts are not statistically significant. * ANOVA: Standard post hoc tests (such as Tukey’s HSD or Games-Howell). † Kruskal–Wallis: Dunn’s post hoc.

and illustrated in Figure 8. For the gingival bevel, a consistent pattern of surface intrusion (negative values) was observed across all tooth groups, with mean deviations ranging from −0.131 mm to −0.151 mm; however, differences among groups were insignificant (p = 0.067). Conversely, the occlusal bevel showed a significant trend from surface extrusion (positive values) in the anterior region to intrusion (negative values) posteriorly. The lateral incisor showed a positive deviation (G2: +0.089 mm), whereas the first molar showed a negative deviation (G6: −0.049 mm), with significant differences among all groups (p < 0.001). Similar trends were noted for mesial and distal walls: the lateral incisor group (G2) showed slight extrusion (positive values), while premolar (G4) and molar (G6) groups showed slight intrusion (negative values). For both walls, deviations in the G2 group were substantially different from those in G4 and G6 (p < 0.001).

The measurements and comparative results for the attachment profile—including bevel angles, wall angles, bevel deviations, and attachment height—are summarized in

Table 2. Comparison of measured bevel angles (°) and attachment height (mm) with their nominal values by tooth group.

	Gingival Bevel Angle			Occlusal Bevel Angle			Attachment Height
	Mean ± SD	p-Value		Mean ± SD	p-Value		Mean± SD	p-Value
Nominal value	38.66			50.19			1.200
G2 (n = 60)	38.09 ± 1.49	0.004	*	50.19 ± 1.36	0.993	*	1.087 ± 0.041	<0.001	†
G4 (n = 60)	37.61 ± 1.13	<0.001	*	50.05 ± 1.37	0.308	†	1.055 ± 0.043	<0.001	†
G6 (n = 60)	37.32 ± 1.53	<0.001	†	49.89 ± 1.80	0.271	†	1.021 ± 0.046	<0.001	†

G2, lateral incisor group; G4, first premolar group; G6, first molar group; SD, standard deviation. * One-sample t-test; † Wilcoxon signed-rank test.

,

Table 3. Comparison of attachment bevel and wall angles (°) among tooth groups.

	G2 (n = 60)		G4 (n = 60)		G6 (n = 60)
	Mean ± SD		Mean ± SD		Mean ± SD		p-Value
Gingival bevel	38.09 ± 1.49	A	37.61 ± 1.13	AB	37.32 ± 1.53	B	0.018	†
Occlusal bevel	50.19 ± 1.36		50.05 ± 1.37		49.89 ± 1.80		0.442	†
Mesial wall	69.84 ± 5.25	A	79.15 ± 5.25	B	78.50 ± 5.42	B	<0.001	†
Distal wall	71.22 ± 5.19	A	77.57 ± 3.81	B	75.17 ± 4.28	C	<0.001	†

G2, lateral incisor group; G4, first premolar group; G6, first molar group; SD, standard deviation. Values in the same row with different superscript letters (A, B, C) are significantly different according to the post hoc tests. Values without superscripts are not statistically significant. † Kruskal–Wallis: Dunn’s post hoc test.

and

Table 4. Comparison of bevel deviations (mm) and attachment height (mm) among tooth groups.

	G2 (n = 60)		G4 (n = 60)		G6 (n = 60)
	Mean ± SD		Mean ± SD		Mean ± SD		p-Value
Gingival bevel deviation	−0.157 ± 0.037	A	−0.145 ± 0.030	B	−0.113 ± 0.033	C	<0.001	*
Occlusal bevel deviation	0.128 ± 0.036	A	0.069 ± 0.030	B	−0.010 ± 0.021	C	<0.001	†
Attachment height	1.087 ± 0.041	A	1.055 ± 0.043	B	1.021 ± 0.046	C	<0.001	†

G2, lateral incisor group; G4, first premolar group; G6, first molar group; SD, standard deviation. Values in the same row with different superscript letters (A, B, C) are significantly different according to the post hoc tests. * ANOVA: standard post hoc tests (such as Tukey’s HSD or Games–Howell). † Kruskal–Wallis: Dunn’s post hoc test.

and illustrated in Figure 9 and Figure 10. Analysis of bevel angles showed that the gingival bevel angle was considerably smaller than its nominal value (38.66°) in all three groups (Table 2, p ≤ 0.004). Inter-group comparison (Table 3) further revealed that the G2 angle was significantly greater than that of G6 (p = 0.018). Conversely, the occlusal bevel angle did not differ substantially from its nominal value in any group, nor were intergroup differences observed (Table 2 and Table 3). For wall angles (Table 3), both mesial and distal walls were more than 10° smaller than the 90° nominal value. Intergroup analysis showed that the mesial wall angle in G2 was considerably smaller than in G4 and G6 (p < 0.001). For the distal wall, significant differences were found among all groups, with the angle increasing in the order G2 < G6 < G4 (p < 0.001).

Bevel deviation analysis (Table 4) revealed consistent intrusion (negative values) of the gingival bevel, most pronounced in G2 (−0.157 mm) and decreasing posteriorly (G2 > G4 > G6 in magnitude; p < 0.001). Conversely, occlusal bevel deviation showed a trend from extrusion in the anterior (G2: +0.128 mm) to near-nominal positioning in the posterior (G6: −0.010 mm), with all groups differing significantly (p < 0.001).

Attachment height was significantly less than the 1.2 mm nominal value in all three groups (Table 2, p < 0.001). Intergroup comparison (Table 4) demonstrated a progressive decrease from anterior to posterior (G2 > G4 > G6, p < 0.001).

Results of the 3D positional deviation analysis are presented in

Table 5. Comparison of attachment positional deviation (mm) among tooth groups.

	G2 (n = 60)		G4 (n = 60)		G6 (n = 60)
	Mean ± SD		Mean ± SD		Mean ± SD		p-Value
X-axis (mesiodistal)	0.018 ± 0.073		−0.013 ± 0.074		0.019 ± 0.109		0.053	†
Y-axis (buccolingual)	−0.113 ± 0.041	A	−0.138 ± 0.062	B	−0.161 ± 0.090	C	<0.001	†
Z-axis (occluso-gingival)	−0.181 ± 0.041	A	−0.111 ± 0.043	B	−0.028 ± 0.046	C	<0.001	†

G2, lateral incisor group; G4, first premolar group; G6, first molar group; SD, standard deviation. Values in the same row with different superscript letters (A, B, C) are significantly different according to the post hoc tests. Values without superscripts are not statistically significant. † Kruskal–Wallis: Dunn’s post hoc test.

and Figure 11. In the mesiodistal direction (X-axis), mean displacement was minimal (< 0.02 mm across groups) with no significant differences (p = 0.053). Conversely, considerable variation was observed in the buccolingual direction (Y-axis) (p < 0.001). All groups exhibited lingual displacement (negative values), with magnitude increasing posteriorly (G2 < G4 < G6). Significant differences were noted in the occluso-gingival direction (Z-axis) (p < 0.001). All groups showed occlusal displacement (negative values), most pronounced in G2 and progressively decreasing posteriorly, with G6 approaching the nominal position.

4. Discussion

It is well established that clear aligners (CAs) provide less precise control over tooth movement compared to conventional fixed appliances [7,15]. Consequently, optimizing attachment design and prescription is crucial to mitigate this limitation. Accurate reproduction of digitally designed attachments on the tooth surface is a fundamental prerequisite for achieving the intended biomechanics [16].

The increasing use of intraoral scanners and 3D printing has facilitated in-house aligner fabrication. To reflect this clinical reality, the present study employed commonly accessible tools: an intraoral scanner, a pressure thermoforming machine, and a conventional high-viscosity restorative resin. Furthermore, the measurement protocol was designed to minimize operator error and enhance reproducibility by defining points based on geometric centers rather than subjective landmarks. The parameters assessed—bevel angles, attachment height, and positional accuracy—were chosen for their direct biomechanical relevance in determining the magnitude and direction of orthodontic forces.

The surface reproducibility of the attachments was evaluated using the “3D Compare” function [14]. Mean deviation was adopted as the primary metric instead of the Root Mean Square (RMS) value to preserve information about the directionality of discrepancies. Recognizing that localized surface defects or protrusions could skew this global measure, the analysis was supplemented with linear deviation measurements between the nominal and fabricated planes of the functionally critical gingival and occlusal bevels.

Several characteristic trends in attachment fidelity were observed. A consistent vertical discrepancy pattern emerged, with attachments generally displaced occlusally. This tendency was most pronounced in the anterior teeth (G2) and gradually diminished posteriorly, approaching nominal dimensions. This positional shift was accompanied by gingival bevel intrusion and occlusal bevel extrusion, particularly in the anterior region. Although previous studies have evaluated vertical displacements, direct comparisons are challenging due to differing experimental designs and the absence of directional analysis in those reports [17,18]. These vertical deviations are clinically significant, as an occlusal shift combined with a tilted profile may generate unintended intrusive forces, potentially hindering planned extrusive movements. Furthermore, such discrepancies could compromise aligner fit, exacerbating the issue over time. Notably, this tendency was most prominent in the maxillary lateral incisor (G2), a tooth frequently reported to exhibit poor tracking and intrusion in CA therapy [19,20], highlighting the need for further investigation.

Attachment height was consistently reproduced as smaller than the nominal dimension across all tooth groups, with greater reduction towards the posterior. Although direct comparisons are limited, similar reductions in attachment height have been documented in previous studies [21,22].

Regarding angular fidelity, the occlusal bevel angle demonstrated high accuracy, showing no significant deviation from the nominal value. Conversely, the gingival bevel angle was consistently smaller than planned, with the most considerable discrepancy observed in the first molar (G6), averaging approximately 1.3°. This discrepancy was significantly greater than in anterior teeth, and may relate to the concurrent reduction in posterior attachment height. Mesial and distal wall angles exhibited the most pronounced deviations, being 10–20° less than the intended 90° angle. This effect may result from resin flash formation at the attachment margins during bonding, which hinders accurate reproduction of sharp angles [23]. This challenge was more pronounced in the anterior region. Therefore, special consideration is warranted when designing rectangular attachments for mesiodistal tooth movements or angulation control. The under-formation of both bevel and wall angles, although variable, is a consistent trend confirmed in prior reports [10,24].

Although a deviation of approximately 0.1 mm in attachment profile and position may be considered negligible in conventional orthodontics, its clinical impact in CA therapy is significant. Tooth movement is widely recommended to be limited to 0.25 mm per aligner, with angular changes between 0.5° and 2.0°, reflecting the physiological limits of the periodontal ligament space [7,25]. Reproduction errors of this magnitude can compromise aligner fit, alter the planned force system, and ultimately reduce treatment predictability.

Several factors may contribute, with the thermoforming process likely being a primary source. Thermoforming causes non-uniform thinning of the plastic sheet as it stretches to conform to the model’s geometry [26,27], producing variable thickness and differential cooling rates across the appliance. Combined with inherent thermal shrinkage, this phenomenon may induce complex dimensional changes [28]. Furthermore, temperature differences between the inner (model-contact) and outer (air-exposed) surfaces may generate internal stresses, potentially causing warpage [29,30].

Shrinkage direction may also depend on tooth position within the arch. Posterior segments exhibit primarily unidirectional shrinkage (buccopalatal), whereas anterior teeth, particularly lateral incisors near arch corners (G2), may experience forces tangential to the arch curvature. This geometrically complex pattern may underlie the specific deviations observed in G2 and warrants further investigation.

Thermoforming-induced variations in thickness and gap width were well-reported in previous studies, showing superior fit in the anterior region and gingival margin compared to posterior and occlusal surfaces, respectively [26], with the most minor gaps at buccal and buccal–gingival areas, and the largest at the palatal, palatal-gingival, and incisal/occlusal surfaces in PET-G aligners [31]. Such volumetric changes can significantly affect template fit and, consequently, attachment reproducibility.

These thermoforming dynamics are likely further complicated by the distinct anatomical morphology of different teeth. Anterior teeth, such as the lateral incisors, are characterized by longer, narrower, and more tapered crowns, whereas posterior teeth are shorter and wider. During thermoforming, the heated sheet makes initial contact with the incisal edges of the anterior teeth, where it cools rapidly and loses some of its flowability. The material must then stretch significantly down the long clinical crowns. In contrast, over the posterior teeth, a larger area of the sheet drapes over shorter crowns with less severe stretching. It is plausible that these differences in material stretching and cooling rates, dictated by tooth morphology, are directly related to the characteristic error patterns observed in this study. Further research is warranted to clarify these complex interactions.

Another contributing factor is the polymerization shrinkage of the composite resin used for bonding [32]. Previous studies have quantified this effect, reporting that attachments can lose approximately 20% of their planned volume during light curing, independent of the light source [33]. The extent of shrinkage is material-dependent, as increased composite filler content has been shown to reduce polymerization shrinkage while enhancing final bond strength [34].

Based on the findings of this study, several clinical guidelines can be cautiously proposed. Our results indicated a significant under-reproduction of the mesial and distal wall angles by 10–20 degrees. This inaccuracy is likely attributable to the formation of a resin puff or flash at the tooth-attachment interface, a phenomenon that is geometrically more challenging to control with sharp 90° angles [23]. Therefore, clinicians should consider using only the necessary amount of composite and avoiding resins with excessive flowability [21]. From a design perspective, it may be beneficial to avoid creating sharp 90-degree angles between the attachment walls and the base. Instead, slight modifications to the attachment’s mesiodistal position could be considered to maintain the intended force vector. Furthermore, given the pronounced extrusion of the occlusal bevel in the anterior region, clinicians might consider a digital overcorrection by designing intentionally more intruded occlusal bevels, especially when extrusion is a key treatment goal. Finally, for cases where the biomechanics demand absolute dimensional accuracy, an alternative workflow involving the bonding of prefabricated attachments prior to the initial scan for aligner fabrication could be considered.

This study identified consistent tendencies and discrepancies between virtually planned attachments and their physical reproductions. These predictable patterns of inaccuracy can serve as a reference for clinical decisions regarding attachment design and placement, composite resin selection, and template fabrication protocols. However, several limitations should be noted. Potential errors from manual fabrication by a single operator and the 3D printing workflow could not be excluded. Furthermore, as an in vitro study, it could not account for the complexities of the intraoral environment, where the observed deviations could be accentuated. Clinical factors such as saliva contamination, limited visibility and access due to the lips and cheeks, minor patient movements, and intraoral temperature fluctuations were not replicated. These variables can compromise both the bonding process and the perfect seating of the template, likely leading to even greater inaccuracies in vivo. Therefore, follow-up studies with controlled protocols and larger sample sizes are warranted to validate these findings and explore methods to improve reproduction fidelity.

5. Conclusions

This in vitro study simulated the attachment reproduction process and quantitatively evaluated its fidelity across different maxillary tooth types. Key findings include the following:

• Three-dimensional Surface Deviation: A consistent gingival bevel intrusion was observed across all tooth groups, with no significant inter-group differences. Conversely, the occlusal bevel exhibited a considerable trend, shifting from extrusion in the anterior region to slight intrusion in the posterior region.
• Attachment Profile Deviation: The most significant discrepancies were found in the mesial and distal wall angles, which were consistently 10–20° smaller than the planned 90° angle. The gingival bevel angle was also consistently reduced, whereas the occlusal bevel angle demonstrated high fidelity to the design. Attachment height was substantially smaller than planned and progressively decreased from anterior to posterior teeth.
• Positional Deviation: Mesiodistal (X-axis) displacement of the attachment center was minimal and not significantly different among groups. However, consistent deviations were observed in the other two planes: lingual (Y-axis) displacement increased posteriorly (G6), while occlusal displacement (Z-axis) was greatest anteriorly (G2) and decreased towards the posterior (G6).