The searches and article selection were performed by three independent authors (SBH, EJK, and BK). Two authors (SBH and EJK) independently made a preliminary list of articles by screening the titles and abstracts for the meta-analysis and evaluated the full manuscript of studies with appropriate inclusion criteria to make a decision; Cohen's kappa was 0.88, indicating almost perfect inter-rater agreement. Discordance in article selection was resolved by debate and consultation with another author (BK). As a result, 17 articles were selected for the analyses.

The outcome measure was converted into a dichotomous value: 0 for loosened MIs, and 1 for unloosened MIs. Unloosed MIs were considered successful, while loosened MIs were considered unsuccessful. Five confounding factors were divided into two categories: patient factors and MI factors. The patient factors were age (< 20 years vs. ≥ 20 years), sex (male vs. female), and the jaw of insertion (maxilla vs. mandible), while the MI factors were length (< 8 mm vs. ≥ 8 mm) and diameter (≤ 1.4 mm vs. > 1.4 mm). The following criteria were used to select the articles: 1) studies on the stability of screws and implants when both items were used for orthodontic anchorage; 2) human clinical studies; 3) prospective and retrospective studies; 4) studies on MIs inserted into the posterior buccal region; and 5) studies that reported the success rates of MIs or explicitly included information that would enable a computation of the success rate with regards to any of the five confounding factors.

Exclusion criteria included articles related to the following: (1) systematic review or meta-analysis, (2) patient satisfaction, (3) orthognathic surgery, (4) radiographic evaluation, (5) microbiology, (6) case report, (7) in vitro studies, (8) literature review, (9) only one article from a particular research group, (10) studies based on the five confounding factors that used different dichotomizations (e.g., the age division of < 18 years vs. ≥ 18 years), and (11) studies on MIs inserted into the retromolar pad, lingual side, or palatal side. Since one of the assumptions of the meta-analysis was independence between studies, studies from specific authors were included only once and studies from coauthors of the chosen studies were excluded.

The following information was extracted from the 17 included studies: the first author's name, publication year, study design, type of temporary anchorage device (TAD), diameter and length of the MIs, number of patients, mean age of the patients, number of TADs, number of successes, success rate, definition of success and failure (Table 1).12346789101516171819202122 The ORs of the success of MIs with respect to these factors were directly calculated using the number of stable MIs and the number of inserted MIs for each category. The ORs calculated from the raw data were double-checked by two authors (SBH, HJL). The success of MIs was defined as the absence of clinically detectable mobility when the orthodontic force applied by a clinician was sustained regardless of the predetermined period, as the predetermined period varied between studies (Table 1).12346789101516171819202122

The quality of the studies was assessed for nonrandomized studies and then quantified using the 9-star Newcastle-Ottawa Scale (NOS),23 which is recommended by the Cochrane collaboration. The study quality was evaluated as being low (0-3 points), medium (4-6 points), or high (7-9 points). The NOS includes eight criteria for evaluating the quality of the included studies (Table 2),12346789101516171819202122 and these criteria are based on the following three categories: subject selection, comparability between groups, and outcome measurement. Any dis agreements were resolved by consensus.

The outcome of the meta-analysis was expressed as a combined OR since the OR can be used as the effect size for the three major study designs: cross-sectional, prospective, and retrospective.2425 The OR in each study was defined as the ratio of the odds (success/failure) of the MIs in two categories: patient factors and MI factors. In the jaw, an OR greater than 1 indicates that successful MIs are more likely to be located in the maxilla (jaw = 1), whereas an OR less than 1 indicates that successful MIs are more likely to be located in the mandible (jaw = 0). Heterogeneity was tested by the Cochran Q and I² statistic; if the p-value of the Q test is greater than 0.10, then no significant heterogeneity exists. It was tentatively suggested that low heterogeneity may be associated with I² values of less than 30%, while substantial heterogeneity may be associated with I² values of over 50%; if significant heterogeneity was identified, then additional subgroup and sensitivity analyses were performed and a qualitative review of the study was conducted until the causes of the heterogeneity were clearly determined.

The studies included in the meta-analysis differed in terms of their study design, participant characteristics, and treatment goals. Variability among these studies in a systematic review may be taken as heterogeneity. To investigate the source of the heterogeneity, a subgroup analysis can be used to answer specific questions about particular groups of patients, the type of intervention, or the type of study. In this study, subgroups were created based on the study design (retrospective study vs. prospective study), study quality (medium vs. high), and sample size of miniscrews implanted (< 200 vs. ≥ 200). A meta-analysis was performed for each subgroup, and the results were reported separately.

A sensitivity analysis is a collective method for verifying the robustness of results. We performed sensitivity analyses to assess the impact of each study on the combined effect size. The meta-analysis was repeatedly performed as follows: a meta-analysis that included all studies except the first study was performed followed by a meta-analysis that included all studies except the second study and so on until each study was excluded once. If the statistical significance of the result was influenced by removing one of the studies, the removed study was reviewed again to confirm the source of the heterogeneity.

Funnel plots have been widely used to detect the potential publication bias of studies in a meta-analysis. However, because the visual interpretation of funnel plots largely depends on the subjective impression of the observer,26 Begg's rank correlation test27 and Egger's linear regression test28 were used as more objective tests to detect publication bias in the present meta-analysis. Significant results (p < 0.05) suggest publication bias. Bias-corrected estimates were calculated using the trim-and-fill method, which accounts for unpublished data by imputing missing studies to yield an unbiased estimate of the effect size.

Meta-analyses and sensitivity analyses were performed using RStudio (ver. 0.96.315; RStudio Inc., Boston, MA, USA) and Comprehensive Meta-Analysis software (ver. 2.0; Biostat, Englewood, NJ, USA).

The present meta-analysis was performed on the success rates of MIs inserted in the buccal area only, since the following inconsistent sucess rates contributed to the heterogeneity. MIs inserted into the anterior region and palatal side that had serious effects on the outcome were excluded from the present meta-analysis because greater root proximity in the anterior region31 and significantly higher success rates on the palatal side32 have been reported. Since the success rates of miniscrews placed in the mandibular lingual side were lower than the rates for the buccal side in Lai and Chen's study,33 the success rate in the mandible was considered to be much lower than the rate in the maxilla. On the other hand, since the retromolar area showed the highest success rate (100%) in Lin et al.'s study34 and since this area was included as part of the mandible, the success rate in the mandible was reportedly higher than the rate in the maxilla.

Furthermore, prospective studies, compared to retrospective studies, showed higher success rates for MIs inserted in the maxilla. Since prospective studies usually have fewer potential sources of bias and fewer confounding factors than retrospective studies, prospective studies are deemed more reliable than retrospective studies. The higher success rates in the maxilla than in the mandible were reportedly due to the thicker mandibular cortical bone compared to the maxillary bone, overheating of the mandibular bone during drilling, and irritation during chewing.34 While the insertion region in the maxilla is keratinized gingiva, it is highly possible that the mandibular insertion region is free gingiva. Therefore, gingival inflammation is more likely to occur in the mandible compared to in the maxilla. Since inflammation leads to loosening, oral hygiene management can help prevent the miniscrews from loosening. Another recommended way to reduce the MI insertion failure in the mandibular buccal posterior region is with pilot drilling, which is different from pre-drilling. During pilot drilling, the MI is inserted into the notch with the fissure bur using a hand driver.35 On the other hand, pre-drilling is a technique using a bur with a smaller diameter than the miniscrew inserted, to make an opening that has a depth less than the thread length of the miniscrew. We found that MIs inserted in the maxilla had higher success rates than did MIs inserted in the mandible regardless of the study design.

Since females tend to get more orthodontic treatments than males for esthetic reasons, the proportion of females was significantly larger in most of the studies examined in our meta-analysis. Although no heterogeneity among the 13 studies was detected, subgroup analyses were performed using the criteria of study design, study quality, and sample size of MIs; each subgroup acquired homogeneity. Although males have a higher bone mineral density than females, the success rate of MIs was not significantly different between the two sexes. These findings were consistent with the results of previous studies.36

In the present meta-analysis, significantly higher success rates were identified in patients who were ≥ 20 years of age compared to in patients who were < 20 years of age. However, in studies that had a sample size of < 200 MIs, no significant differences were found between the success rates of MIs inserted in patients ≥ 20 years old and those who were < 20 years old. However, these studies had moderate heterogeneity (I² = 32.54%), indicating that a few of the studies had different success rates. Therefore these results should be interpreted with caution. Chen et al.16 reported that adolescents had a higher chance of loosening than adults, as thin cortical bone and low bone density were linked to the increased failure of MIs. This result in which increased loosening in younger patients was consistent with the findings of the present meta-analysis. Removable appliances or extraoral appliances such as facemasks may be suitable alternatives to MIs for adolescents. Females usually reach full physical development by 15-17 years of age, whereas males typically complete puberty by 18-19 years of age.37 If the data in the previously published articles had been categorized using these peaks of physical growth, the results might have shown remarkable differences between the success rates of MIs inserted in patients with a different age dichotomization.

The most common MI lengths in the buccal area are 6 mm and 8 mm. Based on the study by Crismani et al.,30 which recommended that lengths under 8 mm be avoided, the present meta-analysis divided MIs into lengths of < 8 mm or ≥ 8 mm. As for diameter, ≤ 1.4 mm vs. > 1.4 mm was selected because the common diameters in the buccal region are 1.4 mm and 1.6 mm, and because Crismani et al.30 recommended that miniscrew diameters under 1.2 mm should be avoided, while Kuroda et al.38 recommended that miniscrew diameters of 1.3 mm and over be used. Even when the criteria of diameter is changed to ≤ 1.2 mm vs. > 1.2 mm or ≤ 1.3 mm vs. > 1.3 mm, larger diameter MIs still show a higher success rate than smaller MIs. It should be noted that although significant differences in the success rates were found between the groups up to 1.4 mm, for the group of ≤ 1.5 mm vs. > 1.5 mm, no significant differences in the succes rates were observed. This result supports the results reported that smaller diameter MIs should be avoided by Crismani et al.30 and Kuroda et al.38 Since miniscrew retention depends on its mechanical interdigitation to the bone, the physical length and diameter of the screw plays an essential role in its placement. Compared to longer and larger screws, shorter and smaller screws have less mechanical interdigitation and can easily break during implantation. Thus, it is important to select an appropriate size when placing miniscrews in the interradicular space to provide orthodontic anchorage, and it is also essential to prevent the screw from being too close to the roots.38 In terms of clinical applications, the results of the present meta-analysis indicating the success rate of each type of screw (length/diameter) and the region where MIs are most successful (maxilla/mandible) can help reduce the failure rate of MIs.

Some of the limitations in the present meta-analysis are as follows: (1) The exclusion of LOE articles may contribute to a risk of publication bias; however, the publication bias could be reduced by using the trim-and-fill method. (2) Despite an exhaustive search of the literature, the number of identified studies satisfying the inclusion criteria was low. Among them, only 5 prospective studies were found, and the number of prospective study in a subgroup was only one (Table 3). With one prospective study, a meta-analysis could not be performed.

More prospective study, including randomized controlled trials, will be needed to make firm conclusions, and authors should choose appropriate studies since inconsistent success rates imposed by particular areas may contribute to the heterogeneity.