Risk of bias
The risk associated with randomization was considered low in most articles because of the detailed, rigorous randomization methods used in those studies. Only Tunçer et al. 12 defined the risk of bias in their article as moderate because randomization was not performed in a fully objective manner but by individual determination of patient eligibility based on met criteria.
The risk associated with disclosure of the group assignment was considered low in all RCTs due to the use of opaque, sealed envelopes or other equivalent randomization methods.
The use of specific treatment methods and the differences between them were known to both study participants and clinicians. Patients were aware of their participation in the study and signed a consent form before the treatment. For this reason, blinding of participants and personnel to treatment status was impossible, and the risk of bias in that criterion was defined as high.
Although blinding of participants was impossible, the overall risk of bias was reduced in some studies by blinding of assessors during the outcome analysis. In the articles by
Al-Sibaie et al.11, Al-Imam et al.14, Chen et al.2 and Tunçer et al.12, the assessor was not directly involved in the study and did not know its purpose; hence, the risk of bias was considered low. In the article by Sadeka et al.5, the assessor could easily define the purpose of the study (patients treated with labial or lingual appliances); thus, blinding was not possible, and the risk was considered high. In turn, the study by Davoody et al. 10 did not provide any information concerning the blinding of assessors. However, it was stressed that the assessors could easily identify patients in each group based on an analysis of cephalograms (presence or absence of microimplants). A similar risk may have occurred in Al-Sibaie et al.'s study11. However, the authors explicitly stressed the blinding of assessors in that study. As it was difficult to determine the actual effect on bias, the risk in Davoody et al.10 article was finally defined as unclear.
In all studies, a complete set of data was obtained from patients and analyzed. Therefore, the risk in that category was considered low.
In Davoody et al.10 and Chen et al.2 articles, the authors stressed the possibility of attrition bias. In the first study, the loss of participants was approximately 30%, and in the second study, it was approximately 6%. Hence, the risk of selective reporting in Davoody et al.10 article was defined as high, while in Chen et al.2 article, it was defined as moderate. In that category, the article by Tunçer et al.12 was also of moderate risk. The reason was selectivity during the eligibility of participants for the study.
Additional potential sources of bias were identified in several studies. In Sadeka et al. 5 article, the authors emphasized that a different vertical position of microimplants in the study groups, which was not taken into account during the outcome analysis, may have influenced the inclination of the upper incisors. In contrast, Chen et al.2 considered the impact of different anchoring methods used in individual patients to be a potential source of error in the results. Xu et al.3 stressed that patient treatment was conducted according to clinical standards, i.e., it was frequently tailored to the individual patient’s needs and modified with the course of therapy. Therefore, direct comparison of specific treatment methods was not possible, which may have affected the lower statistical significance of the results obtained. Tunçer et al.12 pointed to the fact that the collection of patients' molecular samples began too late from the start of the trial and that the subsequent samples were collected at too-large intervals. In that case, however, those measurements were not relevant to our analysis. Finally, given the difficulty in determining the true impact of those limitations on the obtained results, the risk in the studies by Sadek et al.5, Chen et al.2 and Xu et al.3 was considered unclear.
The articles by Deepak et al.4, Lee et al.6 and Koyama et al.13 received 4 points (maximum) for the patient selection criterion. The remaining articles (Ruan et al.9, Jeea et al.8) lost 1 point due to the lack of controls. In terms of confounding factors, also due to the lack of controls, the articles by Ruan et al.9 and Jeea et al.8 received no points, whereas the remaining articles were given the maximum score (2 points). Regarding the last criterion (assessment of study effects), all articles lost 1 point due to the lack of assessor blinding. Moreover, the study by Ruan et al.9 lost 1 point due to a significant reduction in the number of patients confirmed eligible for the outcome analysis compared to the initial group size. Ultimately, the abovementioned study received 1 point for that category.
Outcome analysis
There are many different methods for torque control during anterior retraction; however, most of them were not sufficiently analyzed; hence, their efficiency is not fully validated. According to the present systematic review, not all torque control methods differ in terms of efficiency in a statistically significant way.
Among the torque control methods presented in this systematic review, it is worth examining those with the greatest statistically validated efficacy.
The use of corticotomy during retraction significantly reduces the lingual tipping of the maxillary incisors14, and the technique for corticotomy may be significant for torque control, that is, incisions should be made on both the vestibular and palatal sides. In Tunçer et al.'s12 study, where incisions were made only on the vestibular side, no statistically significant difference in torque loss was obtained compared to that in controls.
The use of TISADs during retraction also significantly reduces the lingual tipping (torque loss) of the maxillary incisors. This is because a vector of force used for retraction moves closer to the center of resistance of the retracted teeth than, for example, during traditional retraction, where the force is applied to maxillary molar brackets. Furthermore, traditional retraction is frequently performed in two steps, which may further influence the greater retroclination of incisors. This is confirmed by Sibaie et al.11.
When retracting the anterior teeth, it is more advantageous to use vestibular mechanics than lingual mechanics5. In the latter, the force vector lies farther from the center of resistance of the retracted segment, which makes it more retruded.
The use of methods such as an intrusion archwire5,10 or the PASS system2 during incisor retraction should also be mentioned. In molar brackets, the PASS system has an additional slot with a 25° inclination. The insertion of an additional archwire into that slot and its attachment to incisors has a similar effect to the use of an intrusion archwire – it strengthens the anchorage in the molar area and causes intrusion of incisors as well as their protrusion, thus increasing torque control during retraction. In the studies, an intrusion archwire was used both in combination with microimplants (Sadeka et al.5) and without microimplants (Davoody et al.10), obtaining better torque control results when skeletal anchorage was included. Nevertheless, the sole application of the intrusion archwire during retraction (Davoody et al. 10) turned out to be more effective for torque control than isolated skeletal anchorage (Al-Sibaie et al.11). This highlights the role of vertical dimension control during retraction in maintaining the correct incisor torque.
The use of intra-maxillary elastics also results in less lingual tipping of incisors during retraction than, for example, the use of a power chain for this purpose7. This difference is probably a result of the much lower force acting on the anterior segment when using intra-maxillary elastics (approx. 100 g) instead of a power chain (approx. 250 g).
The influence of age on the degree of incisor inclination control is probably small, as suggested by Ruan et al.9. Although the retroclination of incisors in that study was greater in the adult group, there was no statistically significant difference.
The statistical analysis showed that the use of additional elements that control the torque resulted in less lingual tipping of incisors during extraction than the closure of dental gaps without using those methods (Table 4). Nevertheless, not all methods were equally effective. The greatest difference between controls and the group treated with assisted torque control was shown in the case of the traditional 2-step retraction with the use of an additional intrusion archwire (Table 4). When comparing the efficiency of torque control during en-masse retraction across studies and ranking it from highest to lowest, the following order of studies was obtained: Sadeka et al.5, Zao et al.7, Chen et al.2, Al.-Sibaie et al.11 and Al-Imam et al.14 (Table 1). However, it should be noted that the scientific credibility of those studies, ranked from largest to smallest, was characterized by a different order of the studies: Al-Imam et al.14, Al-Sibaie et al.11, Chen et al.2, Sadeka et al.5, Davoody et al.10 and Zhao et al.7 (Table 2, 3).
The influence on the aggregate outcome of the statistical analysis also varied. Ranking from the highest aggregate result, the order of the studies was as follows: Al.-Imam et al.14, Al-Sibaie et al.11, Sadeka et al.5, Chen et al.2, Zhao et al.7 and Davoody et al.10 (Table 4). In summary, the outcomes of studies proving the high efficiency of torque control are, alas, frequently associated with low statistical reliability.