Accuracy of Preliminary Maxillary Canine and Anchorage Tooth Movement in Premolar Extraction Cases Using 12 In‐House Clear Aligners: A Randomised Control Trial Comparing Power Arm and Control

ABSTRACT

Objectives

To investigate maxillary canine movement accuracy and anchorage during space closure in first premolar extraction cases (maximum anchorage) using In‐House Clear Aligners (IHCAs).

Materials and Methods

A randomised controlled trial with a split‐mouth design recruited 16 adults in university setting. Each patient was randomly assigned by side for canine retraction using 12 IHCAs to both the experimental palatal power arm (Pa) and non‐Pa control (C). Accuracy was assessed using GOM Inspect by superimposing the virtual and actual digital models between pretreatment and 12th IHCA. Paired t‐test or Wilcoxon signed‐rank test was used to compare virtual‐power arm (VPa) versus actual‐power arm (APa) and virtual‐control (VC) versus actual‐control (AC). Root mean square error (RMSE) was calculated.

Results

Pa displayed a significant difference in preliminary canine distalisation (VPa 2.0 mm vs. APa 2.4 mm), while the control did not differ. Both Pa and control exhibited significantly greater actual distal crown tipping than virtual movement (VPa 4.4° vs. APa −6.3°/VC 4.4° vs. AC −4.3°). AC achieved more canine rotation than VC. RMSE was slightly greater in Pa than control for canine distalisation (Pa 0.6 vs. C 0.55 mm) and distal crown tipping (Pa 10.9° vs. C 8.99°). Conversely, Pa displayed better accuracy in canine rotation. For anchorage, Pa and control exhibited significantly greater actual mesialisation and mesial tipping than virtual. RMSE for anchorage mesialisation and mesial tipping were comparable between Pa and control.

Conclusions

Preliminary canine retraction using Pa may result in greater error in distal crown tipping but less rotation than control.

Trial Registration

ISRCTN 14020146 by the International Standard Randomised Controlled Trial Registry

1. Introduction

Clear aligner (CA) technology has undergone numerous innovations and developments. Contemporary CA treatment can be an option for managing complex malocclusions [1, 2, 3]. Numerous studies have reported that the accuracy and effectiveness of CA treatment can be influenced by several factors, including treatment planning [4], digital scanning [5], material elasticity [6], attachments and interproximal reduction [7], provider skill and experience [8], patient compliance [9], complexity of the case, etc. [10].

Treating extraction cases presents a significant challenge for CA due to inconsistent root alignment following space closure. Many clinicians have observed the ‘bowing effect’ [11, 12, 13]. A hypothesis suggested that the elastic properties of CAs may generate a clockwise moment in the anterior arch, causing lingual inclination and extrusion of incisors, leading to interference in the anterior area. In addition, there is an intrusive force component in the middle of the arch, causing mesial tipping and intrusion of the premolar and molar into the extraction space, potentially resulting in an anterior deep bite and a posterior open bite. Moreover, the lack of tooth support at the extraction site increases the susceptibility of the aligner to deflection. These effects contribute to the loss of anchorage [14, 15].

The reported outcomes using the Invisalign system in extraction cases can generally be classified into two main categories: (1) predictability [12, 13, 14, 16, 17] and (2) outcome assessment index ABO‐OGS [18, 19] or PAR index [10].

Dai et al. [14, 16], Ren et al. [17] and Feng et al. [13] conducted clinical studies to assess the accuracy and predictability of Invisalign treatment in premolar extraction cases. They found consistent evidence suggesting that the anticipated movement of canines and molars was not entirely achieved. Specifically, canines exhibited less distalisation than predicted, along with increased distal tipping, lingual tipping, mesial rotation and extrusion. In contrast, the molars displayed mesial movement, mesial tipping and intrusion. Several strategies have been proposed for counteracting these adverse effects. For example, Womack [11] and Gaffuri et al. [19] suggested adding a power arm (Pa) to the canines in conjunction with applying 3/16″ elastics to the first molar buttons to control canine root angulation.

1.1. Rationale

Recently, the In‐House Clear Aligner (IHCA) has gained popularity among orthodontists because it allows orthodontists to have complete control over the workflow of aligners, such as the software application, manufacturing protocols, material choices, etc. [20] However, substantial studies addressing the effectiveness of IHCA, particularly in extraction cases, are lacking.

To date, only one randomised, controlled trial (RCT) by Jaber et al. [10] has assessed the IHCA in first premolar extraction cases, using Little's Irregularity Index and the PAR index. However, no study has specifically examined the accuracy or predictability of using the IHCA in controlling tooth movement into premolar extraction sites. Therefore, this study aimed to investigate the accuracy of tooth movement using the IHCA.

1.2. Objectives

The purpose of this RCT was to compare virtual (V) versus actual (A) tooth movement achieved during space closure with a conventional thermoforming IHCA in adults requiring maxillary first premolar extraction (maximum anchorage). The trial focused on the first phase of maxillary canine retraction (12 IHCAs). In addition, it aimed to compare the predictability of canine and anchorage movements between the palatal power arm (Pa) and non‐power arm control (C) groups.

2. Materials and Methods

2.1. Trial Design

This trial was approved by the Institutional Review Board (IRB), Faculty of Dentistry and Faculty of Pharmacy (COA.No.MU‐DT/PY‐IRB 2020/022.1304) and is registered in the International Standard Randomized Controlled Trial clinical registry under reference number ISRCTN 14020146. We used the Consolidated Standards of Reporting Trials (CONSORT) 2010 guidelines.

A single‐centre, RCT with a 1:1 split‐mouth ratio was designed (Figure 1A). There were no changes in the design of the methods after trial commencement.

FIGURE 1.

(A) Research design. (B) Location of triangular prism‐shaped attachments. (C) In‐house clear aligner (IHCA)Re and the power arm.

2.2. Sample Size Calculation

As there has been no research on canine tooth movement measurements in extraction cases using IHCA, we had to infer canine distalisation from the mean and standard deviation (SD), as reported by Ren et al. [17].

Considering our randomisation/split‐mouth design, a paired category was used. Sample size calculation was performed using the Statulator website [21]. To achieve a power of 80% and a two‐sided significance level of 5%, a mean difference of 1.3 mm between pairs, and SD of 1.7 mm, the study required a minimum sample of 17 patient pairs. Twenty‐one volunteers were recruited in the event of dropouts.

2.3. Randomisation

In each patient, canine retraction using 12 IHCAs was randomly assigned to the experimental Pa side and the other side as the non‐power arm control.

Sequence generation randomisation was conducted using an online random generator to allocate the side of Pa of the ipsilateral maxillary canines and molars (left or right). The sequence randomisation list was prepared before the intervention, and allocation concealment and assignment were performed by one resident (PS) who had no clinical involvement in the trial.

Because the treatment used a conspicuous Pa, it was impossible for the participants and researchers to be blinded to the method used at all stages.

2.4. Inclusion/Exclusion Criteria

Adult orthodontic patients receiving IHCA treatment between 2019 and 2022 at the Orthodontic Clinic, Faculty of Dentistry were enrolled in this clinical study. Eligibility was based on the following inclusion criteria: age 18 years or older, Angle's Class I or II, with maxillary anterior teeth proclination/protrusion, mild crowding in maxillary arch, requiring bilateral extraction of the maxillary first premolar teeth, requiring maximum anchorage of the maxillary arch. Exclusion criteria: medically compromised patients, previous orthodontic treatment, craniofacial anomalies, severe rotation or asymmetrical position of the maxillary canines, missing maxillary permanent teeth except the third molars, periodontal diseases, systemic pathologies, dental anomalies such as hypercementosis, and non‐compliance.

2.5. Intraoral Scanning

Intraoral scans (iTero, Align Technology Inc., San Jose, CA, USA) were performed at pretreatment and canine retraction at the 12th IHCA.

2.6. In‐House Laboratory Workflow

2.6.1. Virtual Tooth Movement and Dental Compensation

The scanned STL file at pretreatment was imported into Ortho‐Analyser software (3Shape, Copenhagen, Denmark). The first premolars planned for extraction were identified as missing teeth in the virtual model.

Using this software, both maxillary canines were distalised 3 mm into the extraction space. The virtual plan in our study was modified from the protocol suggested by Lombardo et al. [22]. Linear movements of each tooth in combined direction was 0.5 mm/model, mesio‐distal tip or rotation not exceeding 2°/model and torque not exceeding 1°/model.

Dental compensation (overcorrection) for tooth movement was performed for each patient. This included adding angulation and extrusion to the canine and anchorage teeth (except second molars) to mitigate the bowing effect. This involved 8° of distal root tipping for the maxillary canine and 5° of mesial root tipping for the anchorage teeth. Additionally, extrusion was applied to maxillary second premolar and first molar, with approximately 0.7 and 0.5 mm, respectively.

2.6.2. Attachments

Triangular prism‐shaped attachments were placed on maxillary lateral incisors, canines, second premolars and first molars, presumably because this shape is more forgiving, with fewer undercuts than traditional rectangular attachments [23] (Figure 1B). A block‐out area for palatal Pa placement on the maxillary canine and first molar was prepared in the digital model. This area appeared as a long, oval‐shaped object; however, it did not serve as a palatal attachment.

2.6.3. Model Printing

The digital setup was divided into six models and one attachment template model. The digital models were then exported to Meshmixer software (Autodesk Inc., San Francisco, CA, USA) to prepare a hollow base with a shell thickness of 3 mm.

Digital models were printed in 20° oblique orientation using Formlabs V4 photopolymer resin for dental model (Somerville, Massachusetts, USA) on Formlabs 3+ with printing resolution of 100 μm, then underwent post‐curing process according to manufacturer's instruction.

2.6.4. IHCA Thermoforming

Thermoplastic sheets of 0.5 and 0.75 mm (PET‐G: 3A MEDES, Gyeonggi‐do, Republic of Korea) were thermoformed on each printed 3D model with a pressure moulding device (Biostar; Scheu Dental, Iserlohn, Germany). Subsequently, the IHCA was trimmed and finished at 2 mm above the cervical margin.

2.7. Clinical Procedures

On the experimental side, stainless‐steel Pa with heights of 12 and 8 mm were bonded to the cervical 1/3 area on the palatal side of the canine and ipsilateral first molar, respectively. The level of the canine relative to the molar creates the direction of force in the disto‐occlusal vector. The attachments were bonded with the template tray using the hybrid composite material Filtek Z350 XT (3 M ESPE; St. Paul, MN, USA). Cutouts were made on the IHCA to bond the buttons to the maxillary canines and mandibular first molars for Class II elastics.

Participants were then referred for maxillary first premolar extraction on the same day. The IHCA was delivered within 1 week of extraction.

A super‐elastic power chain (TOMY Inc., Tokyo, Japan) of 80–100 g was engaged between the maxillary canine and molar in the Pa group. The power chain was changed every six weeks (Figure 1C).

Patients were instructed to wear the IHCA for at least 22 h/day and change to a new IHCA every week. Two‐ounce 3/16" Class II elastics were used every day after a week of aligner delivery. Follow‐ups were scheduled every 6 weeks to check for off‐track teeth. If any problems were found, management was reinforced using chewies, elastic traction, back‐tracking or rescanning to create a new IHCA, depending on the severity of the deviation. Data from intraoral scanning were obtained at pretreatment and the endpoint when patients had received the 12th IHCA.

2.8. Deviation Analysis

Deviation analysis was performed to evaluate the accuracy of IHCA using the GOM Inspect suite software (Carl Zeiss GOM Metrology, Braunschweig, Germany). Superimposition between virtual and actual digital models and measurement procedures were modified from previous research as described below (Figure 2) [24, 25].

FIGURE 2.

Superimposition method using GOM inspect suite. The master files were imported to GOM inspect suite software. Pretreatment model (Blue) and 12th IHCA (Grey). (A) Construction of a local coordinate system by fitting dental planes on the master model. (B) Selecting and defining surface points for each tooth on the master model. (C) Superimposition of the actual model with the master file, firstly using the 3‐points alignment method. (D) Followed by the local best‐fit function at the palatal area. (E) Superimposition of the canine firstly, by using the Geometric element method. (F) Second superimposition of the canine by local best‐fit. Thereafter, software did the automatic transfer of the points from the master file and link to the actual file. (G) Linear measurements of surface points deviation. (H) Rotation angle between mesio‐distal lines. (I) Tip and torque angles between the axis lines.

2.8.1. Model Superimposition

Pretreatment and the 12 IHCA digital models (both virtual and actual) of each patient were superimposed in STL format by a single operator using GOM software (Figure 2A–F). The coordinate system and surface points were constructed using a pretreatment virtual model (Figure 2A,B). Superimposition between virtual and actual model was performed in three steps. First, the 3‐point alignment method was used (Figure 2C), followed by local best‐fit superimposition performed using the surface area around the palatal rugae and palatal vault [26] (Figure 2D). Then each tooth anatomy (canines, second premolars, first molars) was superimposed. Because canines have moved significantly from their original locations, additional steps were required, beginning with the Geometric Element method (Figure 2E) and followed by canine local best‐fit superimposition (Figure 2F). Thereafter, we performed the ‘copy and link’ function. The GOM software automatically selected (artificial intelligence) surface points on the virtual tooth, which were linked to the actual tooth to ensure repeatability.

2.9. Data Collection

2.9.1. Linear Measurements

The measurements were performed by an outcome assessor. The cusps of the canines, buccal cusps of the second premolars and mesio‐buccal cusps of the first molars were recorded as landmarks to assess tooth displacement (Figure 2G). The virtual and actual distance changes in tooth position in the mesio‐distal, intrusive‐extrusive and buccolingual directions in both the Pa and control groups were recorded.

2.9.2. Angular Measurements

Angular measurements of the canine, second premolar and first molar were performed, which comprised the mesiodistal (rotation) and buccolingual (tip and torque) angulation of the tooth axis (Figure 2H,I).

2.10. Statistical Analysis

Statistical analyses were performed using SPSS (version 22.0; IBM, Armonk, NY, USA). The Shapiro–Wilk test was used to test the normal distributions. Paired t tests or Wilcoxon signed‐rank tests were used to test for significant differences in tooth movement.

2.11. Root Mean Square Error (RMSE)

RMSE represented accuracy in our study according to the formula,

RMSE = sqrt [(Σ(Ai—Vi)²)/n].

The sum of the squared differences between the virtual (APa‐VPa) and actual values (AC‐VC) was divided by the number of observations, and the square root of the result was determined to yield the RMSE.

2.12. Error Analysis

Repetition of surface point selection, superimposition and all measurements between the virtual and actual digital models were repeated for six pairs of digital models. Thereafter, Dahlberg's error formula was calculated to evaluate precision.

3. Results

3.1. Participant Flow and Recruitment

The CONSORT flow diagram is shown in Figure 3. Initially, 21 patients were recruited and randomised for treatment. However, data from five patients were not collected for deviation analysis (Figure 3).

FIGURE 3.

Consolidated Standards of Reporting Trials (CONSORT) flow diagram of patient's recruitment, treatment and analysis. The participant number analysed (w/denominators) was 16/21(76%) for each group. Abbreviations: AC = actual control, APa = Actual power arm, VC = virtual control, VPa = virtual power arm.

3.2. Patient Characteristics

Patients' demographic data were as follows: mean age 23.40 ± 4.60 years; skeletal type I (3 cases) and II (13 cases); Angle classification I (5 cases) and II div I (11 cases); mean crowding in the maxillary arch, −3.05 ± 2.11 mm; mean overjet, 6.17 ± 1.86 mm; and mean treatment time, 17.50 ± 5.19 weeks.

3.3. Outcomes

Tables 1, 2 display measurements of the extent of the three linear and three angular changes in the maxillary canine, second premolar and first molar.

TABLE 1.

Maxillary canine movement from pretreatment to 12th IHCA; Comparison of virtual and actual tooth movements in all directions. Results of paired T‐test or Wilcoxon signed‐rank test (n = 16).

Movement	Power arm mean ± SD [95% CI]		p	Control mean ± SD [95% CI]		p	RMSE
	VPa	APa	VPa_APa	VC	AC	VC_AC	PA	C
Distalisation (mm) = + Mesialisation (mm) = −	2.01 ± 0.26 [1.87–2.15]	2.38 ± 0.57 [2.08–2.68]	0.01	2.08 ± 0.46# [1.83–2.32]	2.32 ± 0.66 [1.97–2.67]	0.08#	0.61	0.55
Extrusion (mm) = + Intrusion (mm) = −	0.71 ± 0.47 [0.46–0.96]	0.42 ± 0.61 [0.10–0.75]	0.02	0.70 ± 0.45 [0.46–0.94]	0.41 ± 0.43# [0.17–0.64]	0.003#	0.52	0.39
Buccal (mm) = + Lingual (mm) = −	1.08 ± 0.41 [0.86–1.30]	1.13 ± 0.58 [0.82–1.44]	0.49	1.14 ± 0.49 [0.87–1.39]	1.10 ± 0.56 [0.80–1.40]	0.73	0.27	0.36
Tip (°) MCT = + DCT = −	4.37 ± 1.86# [3.38–5.36]	−6.30 ± 3.10# [−7.95 to −4.65]	0.000#	4.42 ± 1.34# [3.71–5.13]	−4.31 ± 2.25 [−5.51 to −3.10]	0.000#	10.90	8.99
Torque (°) BCT = + LCT = −	−3.28 ± 2.71 [−4.73 to −1.84]	0.43 ± 3.20 [−1.28–2.13]	0.001	−3.07 ± 2.94 [−4.63 to −1.50]	−0.25 ± 2.11 [−1.37–0.88]	0.004	5.02	4.29
Rotation (°) M‐in = + D‐in = −	−0.14 ± 4.12 [−2.34–2.05]	−2.02 ± 5.45# [−4.93–0.88]	0.39#	−0.94 ± 5.71 [−3.98–2.10]	−8.30 ± 4.84 [−10.88 to −5.73]	0.000	6.33	9.12

Note: Favourable movement = + (Distalisation, Extrusion, Buccal, Mesial Crown Tip (MCT), Buccal Crown Torque (BCT), Mesial‐in rotation). Non‐favourable movement = − (Mesialisation, Intrusion, Lingual, Distal Crown Tip (DCT), Lingual Crown Torque (LCT), Distal‐in rotation).

Abbreviations: #, non‐parametric; AC, control side of actual model; APa, power arm side of actual model; VC, control side of virtual model; VPa, power arm side of virtual model.

TABLE 2.

Maxillary second premolar and first molar movement from pretreatment to 12th IHCA; Comparison of virtual and actual tooth movements in all directions. Results of paired T‐test or Wilcoxon signed‐rank test (n = 16).

Movement	Power arm mean ± SD [95% CI]		p	Control mean ± SD [95% CI]		p	RMSE
	VPa	APa	VPa_APa	VC	AC	VC_AC	PA	C
Second Premolar
Distalisation (mm) = + Mesialisation (mm) = −	0.23 ± 0.29 [0.07–0.38]	−0.39 ± 0.28 [−0.54 to −0.24]	< 0.001	0.21 ± 0.25# [0.08–0.34]	−0.44 ± 0.29# [−0.60 to −0.28]	< 0.001#	0.69	0.72
Extrusion (mm) = + Intrusion (mm) = −	0.48 ± 0.34 [0.30–0.65]	−0.16 ± 0.34 [−0.34–0.02]	< 0.001	0.55 ± 0.25 [0.42–0.69]	−0.03 ± 0.30 [−0.19–0.13]	< 0.001	0.73	0.67
Buccal (mm) = + Lingual (mm) = −	0.17 ± 0.19 [0.07–0.27]	0.04 ± 0.33 [−0.13–0.21]	0.022	0.15 ± 0.27 [0.01–0.29]	−0.05 ± 0.25 [−0.18–0.08]	0.005	0.24	0.30
Tip (°) DCT = + MCT = −	2.75 ± 1.57 [1.91–3.59]	−0.40 ± 1.90 [−1.42–0.61]	< 0.001	2.87 ± 1.25 [2.21–3.54]	−0.59 ± 1.69 [−1.49–0.31]	< 0.001	3.77	3.78
Torque (°) BCT = + LCT = −	1.08 ± 1.07# [0.50–1.65]	0.08 ± 1.56 [−0.75–0.91]	0.007#	0.64 ± 1.05# [0.09–1.20]	0.22 ± 1.56 [−0.39–0.84]	0.179#	1.58	1.28
Rotation (°) D‐in = + M‐in = −	0.91 ± 2.09# [−0.21–2.03]	0.06 ± 1.67# [−0.83–0.95]	0.032#	0.50 ± 1.80# [−0.46–1.46]	−0.72 ± 1.00 [−1.25 to −0.18]	0.006#	1.50	1.78
First Molar
Distalisation (mm) = + Mesialisation (mm) = −	0.25 ± 0.28# [0.10–0.39]	−0.30 ± 0.23 [−0.42 to −0.18]	< 0.001#	0.27 ± 0.30# [0.11–0.43]	−0.41 ± 0.27 [−0.56 to −0.26]	< 0.001#	0.63	0.77
Extrusion (mm) = + Intrusion (mm) = −	0.61 ± 0.37 [0.41–0.81]	−0.15 ± 0.33 [−0.32–0.02]	< 0.001	0.65 ± 0.26 [0.51–0.79]	−0.04 ± 0.26 [−0.18–0.09]	< 0.001	0.85	0.76
Buccal (mm) = + Lingual (mm) = −	0.27 ± 0.21 [0.16–0.38]	0.15 ± 0.28 [−0.01–0.29]	0.039	0.30 ± 0.22 [0.18–0.42]	0.10 ± 0.25 [−0.03–0.23]	0.014	0.25	0.33
Tip (°) DCT = + MCT = −	2.66 ± 1.34# [1.94–3.37]	−0.12 ± 2.18 [−1.28–1.04]	< 0.001#	2.70 ± 1.25 [2.03–3.36]	−0.25 ± 1.50 [−1.05–0.55]	< 0.001	3.28	3.18
Torque (°) BCT = + LCT = −	0.73 ± 1.08 [0.15–1.30]	0.44 ± 1.21 [−0.21–1.08]	0.455	0.65 ± 1.14 [0.04–1.26]	0.45 ± 1.43 [−0.31–1.21]	0.726	1.49	2.20
Rotation (°) D‐in = + M‐in = −	0.76 ± 1.29# [0.08–1.45]	0.72 ± 1.49 [−0.08–1.51]	0.938#	0.98 ± 1.41# [0.22–1.73]	−0.21 ± 1.31 [−0.91–0.49]	0.004#	1.40	1.69

Note: Favourable movement = + (Distalisation, Extrusion, Buccal, Distal Crown Tip (DCT), Buccal Crown Torque (BCT), Distal‐in rotation). Non‐favourable movement = − (Mesialisation, Intrusion, Lingual, Mesial Crown Tip (MCT), Lingual Crown Torque (LCT), Mesial‐in rotation).

Abbreviations: #, non‐parametric; AC, control side of actual model; APa, power arm side of actual model; VC, control side of virtual model; VPa, power arm side of virtual model.

Bar plots (mean±SD) of distalisation‐mesialisation, tipping and rotation of virtual power arm (VPa), actual power arm (APa), virtual control (VC), actual control (AC), classified by canine, second premolar and first molar are shown in Figure 4. The primary outcomes of interest are shown below.

FIGURE 4.

Bar plots (mean/SD) of distalisation‐mesialisation, tipping and rotation of actual control (AC), actual power arm (APa), virtual control (VC), virtual power arm (VPa), classified by canine, second premolar and first molar. Paired t‐test or Wilcoxon signed‐rank test was used to compare VPa vs. APa and VC vs. AC (p < 0.05). Sign direction: Canine movement (+) = Distalisation, Mesial Crown Tip, Mesial‐in rotation. Canine movement (−) = Mesialisation, Distal Crown Tip (DCT), Distal‐in rotation. Second premolar and first molar (+) = Distalisation, Distal Crown Tip, Distal‐in rotation. Second premolar and first molar (−) = Mesialisation, Mesial Crown Tip, Mesial‐in rotation. (*) = statistically significance.

3.3.1. Canine Distalisation

In the Pa group, a statistically significant difference was observed between the mean virtual (2.01 ± 0.26 mm) and actual (2.38 ± 0.57 mm) canine distalisation, with the actual demonstrating more distalisation than virtual (Table 1). However, in the control group, no significant difference was observed between the mean virtual (2.08 ± 0.46 mm) and actual (2.32 ± 0.66 mm) movements.

The RMSE values of the Pa (0.61 mm) and the control (0.55 mm) groups differed by only 0.06 mm.

3.3.2. Canine Tipping

In the Pa group, a statistically significant difference was observed between the mean virtual (4.37° ± 1.86°) and actual (−6.30° ± 3.10°) distal crown tipping, with the actual demonstrating more distal crown tipping than the virtual movement (Table 1). Similarly, in the control group, a significant difference was observed between the mean virtual (4.42° ± 1.34°) and actual (−4.31° ± 2.25°) movements, with the actual demonstrating more distal crown tipping than virtual.

The RMSE values of the Pa group (10.90°) deviated more than those of the control group (8.99°), with a difference of 1.91°.

3.3.3. Canine Rotation

In the Pa group, no significant difference was observed between the mean virtual (−0.14° ± 4.12°) and actual (−2.02° ± 5.45°) movements (Table 1). Conversely, in the control group, a significant difference was observed between the mean virtual (−0.94° ± 5.71°) and actual (−8.30° ± 4.84°) movements, with the actual demonstrating more distal‐in rotation than the virtual.

The RMSE values of the Pa group (6.33°) deviated less than those of the control group (9.12°), with a difference of 2.79°.

3.3.4. Second Premolar Mesialisation

In the Pa group, a statistically significant difference was observed between the mean virtual (0.23 ± 0.29 mm) and actual (−0.39 ± 0.28 mm) movements, with the actual demonstrating more mesialisation than virtual (Table 2). Similarly, in the control group, a significant difference was observed in the mean virtual (0.21 ± 0.25 mm) and actual (−0.44 ± 0.29 mm) movements.

The RMSE values of the Pa (0.69 mm) and control (0.72 mm) groups deviated by only 0.03 mm.

3.3.5. Second Premolar Mesial Crown Tipping

In the Pa group, a statistically significant difference was observed between the mean virtual (2.75° ± 1.57°) and actual (−0.40° ± 1.90°) movements, with the actual demonstrating more mesial crown tipping than the virtual. Similarly, in the control group, a significant difference was observed between the mean virtual (2.87° ± 1.25°) and actual (−0.59° ± 1.69°) movements (Table 2).

The RMSE of the Pa group (3.77°) had a deviation similar to that of the control group (3.78°).

3.3.6. First Molar Mesialisation

In the Pa group, a statistically significant difference was observed between the mean virtual (0.25 ± 0.28 mm) and actual (−0.30 ± 0.23 mm) movements, with the actual demonstrating more mesialisation than the virtual. Similarly, in the control group, a significant difference was observed in the mean between the mean virtual (0.27 ± 0.30 mm) and actual (−0.41 ± 0.27 mm) movements (Table 2).

The RMSE of the Pa group (0.63 mm) exhibited less deviation than that of the control group (0.77 mm), with a difference of 0.14 mm.

3.3.7. First Molar Mesial Crown Tipping

In the Pa group, a statistically significant difference was observed between the mean virtual (2.66° ± 1.34°) and actual (−0.12° ± 2.18°) movements, with the actual demonstrating more mesial crown tipping than the virtual (Table 2). Similarly, in the control group, a significant difference was observed between the mean virtual (2.70° ± 1.25°) and actual (−0.25° ± 1.50°) movements.

The RMSE of the Pa (3.28°) and control (3.18°) groups revealed 0.1 mm difference in deviation.

3.4. Error Measurements

Dahlberg's formula revealed that all linear measurements did not exceed 0.1 mm and all angular measurements did not exceed 0.5° for any of the investigated variables.

3.5. Power Analysis

According to our data, detecting a mean of the differences of 0.37 mm between pairs (APa‐VPa), with an SD of 0.48 mm, a level of significance of 5% (two‐sided) and a final sample size of 16 patients, the resulting power calculated was 82%. (https://homepage.univie.ac.at).

3.6. Harm

At this stage, no harm was observed in any participant during the trial.

4. Discussion

The purpose of the present study was to deepen our understanding of unwanted tooth movement and the bowing effect of the IHCA in premolar extraction cases [11, 12, 13]. The research design of this RCT aimed to determine whether using palatal Pa, in addition to attachments and dental compensation, would increase the predictability of canine retraction and anchorage control. Data were collected at the 12th IHCA (made from six printed models).

In the first premolar extraction case, the canine was the first tooth to move distally during space closure. This movement can potentially result in a considerable bowing effect and loss of anchorage. Interestingly, for canines, the Pa group displayed greater actual canine distalisation than virtual movement (2.0 mm vs. 2.4 mm). The Pa and control groups exhibited significantly greater distal crown tipping than predicted (Pa 4.4° vs. −6.3°/control 4.4° vs. –4.3°). Furthermore, the control group achieved more canine rotation than virtual (−0.9° vs. −8.3°). The RMSE was slightly greater in Pa than in the control group for canine distalisation (Pa 0.6 vs. control 0.55 mm) and distal crown tipping (Pa 10.9° vs. control 8.99°). Conversely, the Pa group displayed lower error in distal‐in rotation (6.3° vs. 9.1°).

For anchorage, to counter the mesial tipping effect during space closure, all maxillary second premolars and first molars of both the Pa and control groups were virtually set up, the roots to tip mesially with the extrusion of the teeth. However, the crowns in both groups actually tipped mesially with the intrusion.

Moreover, the ratio between loss‐of‐anchorage and maxillary canine distalisation in the second premolar was 1:6.1 (16.4%) for the Pa group and 1:5.27 (18.97%) for the control group. In the first molar, the ratio was 1:7.93 (12.61%) for the Pa group and 1:5.66 (17.66%) for the control group.

These results revealed significant differences between virtual and actual tooth movement, which is consistent with previous studies using Invisalign. They reported that tooth movement was not fully achieved as predicted [12, 13, 14, 16, 17, 27]. In particular, the canines were found to show more distal tipping, lingual tipping, distal rotation and intrusion than predicted [16, 27].

In our study, the actual crown angulation of the maxillary canines in both the Pa and control groups showed more distal tipping than virtual movement. This finding agrees with Baldwin et al. [12] and Feng et al. [13]. Interestingly, Feng et al. [13] suggested the use of compensatory anti‐tipping of canines by setting the roots 23° distally towards the extraction site in complete treatment. In our study, the design was only 8° for approximately 40% of the extraction site.

The actual retraction of the maxillary canines in the present study showed greater distalisation than virtual in Pa group. These findings contrast with those of Dai et al. [16] and Ren et al. [17] who reported insufficient retraction of the maxillary canines using the Invisalign system. A possible explanation for the observed over‐distalisation than predicted in our study could be the insufficient wrapping of plastic, the auxiliary force from the palatal power chain and the use of buccal Class II elastics. Insufficient plastic wrapping caused distal tipping of the maxillary canines into the wider part or block‐out area of the aligner. The maxillary canines demonstrated intrusion with distal‐in rotation, which was consistent with the finite element analysis of Zhu et al. [15]. In addition, we also observed that the off‐track cases appeared to be poor cooperation in wearing Class II elastics.

The anchorage loss direction in our study was consistent with that of other publications using the Invisalign system, in which the molars were found to move mesially with crown tipping towards the extraction site [10, 13, 14, 16, 17]. Interestingly, Feng et al. [13] recommended an anti‐tipping design of the second premolar and first molar at 9.5° and 8.7°, respectively. This possibly indicates that a greater anchorage preparation was required in our study.

Theoretically, the Pa provides an extension, allowing a force vector to move closer to the centre of resistance of the tooth, and presumably, less distal tipping of the canine should be achieved, as reported in several studies [28, 29]. Interestingly, the Pa group had significantly less control of canine root movement. We hypothesised that this phenomenon was due to the significant deflection of CA toward the extraction site. In addition, the height of the palatal Pa may have been inadequate to counteract this large moment. However, this palatal force positively helped decrease the distal‐in rotation of the canine in the Pa group compared with the control.

4.1. Strengths

The RMSE accuracy was used in our study instead of the percentage accuracy formula, which is an inappropriate tool for interpreting our results. In the extraction cases, some raw data in the virtual movement approached zero and some values in the actual movement were many times higher than those in the virtual movement, which overinflated the percentage values. In addition, there is a possibility that the datasets produced both positive and negative deviations between the virtual and actual movements, which could lead to one result neutralising the other.

4.2. Limitations

This study has some notable limitations. These findings cannot be generalised to the Invisalign system because of the use of different thermoplastic materials. Previous research found that the elastic properties of the aligner materials played a crucial role in achieving accurate tooth movement [6]. In our study, we utilised PET‐G materials with thicknesses of 0.5/ 0.75 mm, which were effective for conventional thermoforming of CA. However, a newer material, multilayer‐polyurethane used in Invisalign, provides enhanced flexibility, durability and potentially improved treatment accuracy [30]. Recently introduced direct 3D‐printing materials have not yet been applied to extraction cases. One significant benefit of using the traditional method was the reduction in 3D printing models and less waste of model resin.

Our study reported the first phase of 3 mm canine distalisation (the first set of 12 IHCAs). Because of the bowing effect during space closure in premolar extraction cases, we aimed to fabricate the IHCAs in several subsets instead of using the entire virtual setup, as in commercial CAs. However, we found that the achieved canine movement was only 2 mm, which may indicate an inaccuracy between the software's movement distance and actual tooth movement. The next phase of our canine distalisation research is ongoing to evaluate whether there are any differences in accuracy during the different phases of space closure.

Various factors may be involved in the accuracy measurements, such as the scanners, software for setup, 3D printers and superimposition. It is possible that there was a slight discordance between the superimposed structures. These limitations were described by Adel et al. [31].

Future research should involve prospective studies that compare the outcomes between various CA materials.

In summary, in the first phase of canine retraction using the IHCA, palatal Pa does not decrease distal crown tipping. Attachments and overcorrection should be considered to help achieve the predicted changes in extraction cases.

5. Conclusion

Based on the present findings, it can be concluded that

• Maxillary canines achieved slightly greater distalisation than predicted in the palatal Pa group. Both Pa and control groups achieved greater distal crown tipping than predicted.

• Maxillary canines in the control group achieved more distal‐in rotation than predicted.

• Maxillary second premolars and first molars in both the Pa and control groups achieved greater mesial crown tipping than predicted.

• According to the RMSE, a palatal Pa may result in greater error in distal crown tipping but less rotation of canine movement.

• Inaccuracy observed in the anchorage was comparable between both groups.

Author Contributions

Natnicha Vongtiang: investigation, software, data collection, data curation, formal analysis, data interpretation, writing – original draft. Nathaset Tongkitcharoen: investigation, software, data‐collection, writing ‐ original draft. Sawitt Eurutairat: investigation, software, data‐collection. Somchai Manopatanakul: conceptualisation, final approval manuscript. Peerapong Santiwong: review, final approval manuscript. Nita Viwattanatipa: conceptualisation, project administration, methodology, funding acquisition, clinical registry, IRB submission, investigation, supervision, data interpretation, validation, writing – review and editing, final approval manuscript.

Ethics Statement

Institutional Review Board (IRB), Faculty of Dentistry and Faculty of Pharmacy (DTPY‐IRB) No. COA.No.MU‐DT‐DT/PY‐IRB 2020/022.1304.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request.