The search strategy included relevant English databases, including Ovid-Medline, Ovid Embase, and the Cochrane Central Register of Controlled Trials, and Korean databases, such as KoreaMed and KMbase. The search terms were derived from those employed in Ovid-Medline and search functions, including controlled vocabulary (MeSH and Emtree), text words, logical operators, and truncation, were appropriately applied considering the characteristics of each database. The search was limited to the literature published between January 2000 and June 2023 in English and Korean (Supplementary Table 2).

We included original studies published in scientific journals with the full text available. We included studies as follows: those involving patients with facial asymmetry aged 18 years or older, focusing on orthognathic surgery, reporting one or more predefined outcomes, and conforming to a predetermined study design. We excluded non-original articles (reviews, editorials, letters, comments, abstracts, duplicate research); studies addressing the treatment of pathological conditions such as condylar hyperplasia; studies concerning the treatment of congenital facial deformities such as hemifacial microsomia and cleft lip and palate; documents not published in peer-reviewed journals (gray literature, abstracts, etc.); and studies not published in Korean. Two reviewers (YJK and MYK) independently conducted the literature selection process based on the predetermined inclusion and exclusion criteria. Inter-rater agreement was assessed using Cohen’s kappa statistic, which yielded a substantial agreement coefficient of 0.801. In cases of disagreement, a final selection of literature was made through discussions involving the entire research team.

Pilot data extraction was performed in several studies to standardize the process and enhance consistency between reviewers. Subsequently, two independent reviewers extracted data from the selected studies into a standardized form. Disagreements between the two reviewers were resolved by rechecking the data and further discussions with the clinical advisory committee.

The extracted data included (a) study characteristics (authors, year of publication, study design, period, and the number and location of research centers), (b) study population (inclusion/exclusion criteria, number of participants included/dropped out, sex, age, and type of dentoskeletal malocclusion), (c) methods (surgical techniques, use of SFA, computer simulation, analytic methods [2D vs. 3D] to assess asymmetry), (d) values of outcome variables, and (e) follow-up duration.

The methodological quality rating was determined using the Newcastle-Ottawa Scale (NOS) to assess the quality of non-randomized studies in the meta-analyses. Studies received one or two stars within each NOS category. A study with a score of ≥ 7 was considered high quality, those scoring 4–6 were considered moderate quality, and studies scoring < 4 were categorized as low quality. A critical assessment of the studies was conducted by two reviewers, and any discrepancies were resolved by a third.

We used the Grading of Recommendations, Assessment, Development, and Evaluation¹⁰ to evaluate the quality of evidence of the studies in which the meta-analysis was performed. Each outcome was rated based on the quality of evidence as high, moderate, low, or very low within five domains: risk of bias, imprecision, inconsistency, indirectness, and other biases.

The meta-analysis was conducted using the Hartung–Knapp–Sidik–Jonkman method, which is a random-effects model. Heterogeneity among the studies was assessed using the I² statistic, indicating the proportion of the total variation attributed to heterogeneity rather than chance. Heterogeneity was categorized as low (25–50%), intermediate (50–75%), or high (75–100%).

In cases of high heterogeneity, the effect estimate was reported as the standardized mean difference (SMD) or mean difference. To explore potential sources of heterogeneity, subgroup and sensitivity analyses were performed on studies exhibiting moderate or high heterogeneity. These analyses utilized predefined variables to investigate the factors contributing to heterogeneity.

Subgroup analyses were conducted to improve overall skeletal symmetry, which included all outcomes of skeletal symmetry and menton deviation considering surgery-related factors, such as the surgical techniques (one- vs. two-jaw), the use of SFA, utilization of computer simulation, and the analytical methods employed to evaluate asymmetry (2D vs. 3D). Skeletal relapse was evaluated by examining the SMDs of the overall skeletal symmetry from immediately after surgery to the follow-up periods of < 3, 3–12, and > 12 months. Additionally, relapse of menton deviation was observed for follow-up periods of < 3 months and > 3 months.

Standardized mean differences for the upper and lower dental midlines, midline discrepancies between the upper and lower dentition, and lip canting were assessed. The prevalence of signs and symptoms of TMD before and after surgery were analyzed. Quality of life was examined. Publication bias was assessed using funnel plots and Egger’s test.¹¹ Statistical analyses were performed using R (version 3.6.3, http://cran.r-project.org/) and comprehensive meta-analysis (version 3.3.070). Statistical significance was defined as P < 0.05.

A comprehensive database search yielded 2,928 articles, including 2,694 in English and 234 in Korean. After removing duplicate studies, 2,175 articles remained for further evaluation based on the selection criteria. Through a rigorous review process, 168 articles were selected based on their titles and abstracts, and a final set of 49 articles were included after a thorough examination of the full texts (Figure 1).

Supplementary Table 3 summarizes the NOS results. Most of the studies were assessed as moderate quality,^12-48 while a few were assessed as high quality.^14,49-62 The studies were classified as high quality when the comparability of cohorts was appropriately addressed based on the study design or analysis.

The outcome parameters of the studies included in the meta-analysis were examined. Factors such as risk of bias, indirectness, inconsistency, and imprecision were not serious issues. The certainty of evidence regarding menton deviation is very low, owing to publication bias. However, publication bias did not significantly affect the results of menton deviation analysis. Certainty was low for other outcomes (Supplementary Table 4).

The meta-analysis included 1,796 participants. The publication timeline revealed 1 article from 2003 and 8 articles from 2019, and a significant proportion (51%) of the studies were published within the last 5 years (2019–2023). The studies originated from various countries: 19 from South Korea, 12 from Taiwan, 9 from Japan, 7 from China, 1 from Hong Kong, and 3 from Europe. The mean number of participants per study was 41 (range: 8–228), with a mean age of 24.6 years. In addition, 33 studies focused on skeletal Class III malocclusion. The follow-up periods varied across studies, ranging from a minimum of 15 days to a maximum of 3 years, with an average follow-up time of 9.7 months.

In the analysis of skeletal symmetry based on the follow-up period, studies with a follow-up period of < 3 months exhibited the highest SMDs (–1.36, P < 0.01), indicating a significant change due to surgery, while studies with a follow-up period of 3–12 months also showed a significant improvement in skeletal symmetry (SMD = –1.34, P < 0.01). However, in studies with a follow-up period > 12 months, the SMD was less than that in those with a shorter follow-up period (SMD = –0.60, P > 0.05) (Figure 5). The analysis of menton deviation based on the follow-up period showed similar results, with a greater SMD for studies with a follow-up period of < 3 months^12,14; studies with a follow-up period > 3 months showed fewer SMDs (Supplementary Figure 9).

Orthognathic surgery significantly reduced maxillary and lip canting. However, lip canting correction was found to be less predictable than maxillary canting.⁶³ Kim et al.²⁴ demonstrated a relationship between lip canting and canting correction of the anterior maxillary transverse occlusal planes. Lip asymmetry is influenced by muscular factors, and surgery may not cause significant changes in lip-line canting.⁶⁴ Lee et al.⁵⁰ reported an average lip cant of 3.12 for patients undergoing one-jaw surgery, which reduced to 1.27 after surgery.

Facial asymmetry can only be analyzed in the frontal view; therefore, a frontal cephalogram in addition to a lateral cephalogram is required for diagnosis and surgical planning. However, in cases of facial asymmetry, 3D distortion of the mandible is often observed due to various yaw and roll patterns, which limits a comprehensive understanding of the mandibular morphology using only 2D imaging modalities.³⁰ Owing to technological advances in medical imaging, 3D imaging modalities, such as computed tomography (CT), intraoral scanners, and facial scanners, are readily available to clinicians. Integrating these datasets allows computer simulation surgery, which is reported to outperform traditional methods in enhancing facial midline symmetry.^17,54

Patients with facial asymmetry often show condylar deformation with degenerative changes, which can be accurately analyzed using CT.⁶⁵ Study groups using 2D and 3D imaging modalities for surgery and those who planned surgery with or without computer simulation all showed significant improvements in asymmetry. However, it was not possible to directly compare the outcome variables of the studies that used 2D and 3D imaging modalities due to the nature of the meta-analysis. Additionally, the outcome variables used in these studies were mainly menton deviations and dental midline discrepancies, which do not fully represent facial asymmetry.

Skeletal symmetry outcomes related to transverse dimensions such as frontal ramal inclination and distance from the gonion to the midsagittal reference plane are strongly associated with facial contours. In patients with a translation asymmetry type, transverse movement of the distal segment results in correction of frontal ramal inclination with stability and improves soft tissue facial contour.^15,20 A study comparing 2D and 3D planning in cleft patients with severe discrepancies showed that 3D planning yielded superior outcomes in terms of facial contour asymmetry, but we excluded the study due to cleft palate patients.⁶⁶ Additionally, Udomlarptham et al.¹² concluded that 3D planning was particularly advantageous to achieve bilateral mandibular contour symmetry.

The effects of surgical intervention on TMD symptoms vary across studies. Consistent with the results of a previous systematic review focusing on various patients with skeletal malocclusion who underwent orthognathic surgery,⁶⁷ our analysis of patients with facial asymmetry revealed an overall improvement in the prevalence of pain.

However, differences in TMJ symptoms have been observed depending on the surgical technique used. Patients who underwent intraoral vertical ramus osteotomy (IVRO), intraoral vertical-sagittal ramus osteotomy (IVSRO), or sagittal split ramus osteotomy (SSRO) without fixation experienced a substantial reduction in TMJ symptoms.^45-47,61,62 The favorable effect of surgeries with no fixation of the bone segments on the TMJ has been corroborated in previous studies. Fujimura et al.⁶¹ reported 92% and 83% improvements in clicking in patients who underwent IVSRO and IVRO, respectively. Chen et al.⁶⁸ also demonstrated a 71.4% improvement in pre-existing TMJ clicking in patients who underwent IVRO for mandibular setback. They showed that the rate of improvement of TMJ symptoms was not associated with the amount of setback. The alleviation of TMD symptoms after surgery may be attributed to the absence of fixation between the proximal and distal segments, allowing the proximal segment to attain a stable position.⁶⁹

In contrast, some patients who underwent conventional SSRO with fixation exhibited TMJ symptoms.^46,47,62 Hu et al.⁷⁰ found that 8% of initially asymptomatic patients developed TMD symptoms 6 months postoperatively, emphasizing the challenges in replicating the original condylar position during SSRO. This difficulty poses a risk of excessive pressure on the articular disc or an unfavorable condylar position, potentially resulting in joint noise, pain, or exacerbation of pre-existing TMD symptoms.⁶⁹ However, Toh and Leung reported that, despite the surgical technique used (IVRO, SSRO, or a combination of the two techniques), 12.5% of patients who were previously asymptomatic developed signs and symptoms after surgery.⁶² AlWarawreh et al.⁷¹ reported that most patients who underwent SSRO with rigid fixation showed no change in the signs and symptoms of TMD. Eshghpour et al.⁷² studied the biomechanical stress of three different modalities, two bicortical screws, three bicortical screws, and a miniplate, with SSRO using finite element analysis and concluded that the miniplate showed the least displacement of the bony segments.

Long-term stability is a crucial factor determining the success of orthognathic surgery. Short-term relapse can be attributed to factors such as condylar displacement during surgery, while long-term stability is influenced by condylar remodeling and skeletal growth.⁷³ We revealed that skeletal symmetry improvements remained stable even when relapse patterns were identified up to 3 years post-surgery. Most studies demonstrated stable menton and lower midline positions following surgery, suggesting no significant relapse. Most skeletal relapses occurred within the first 12 months postoperatively, with similar amounts of relapse observed between the immediate postoperative period and the 3-month to 12-month postoperative period. These relapse patterns align with previous studies reporting that most surgical relapses occur within 1 year after surgery.^39,74

Our study has several limitations. First, we included ramus height and body length as parameters to assess skeletal symmetry. However, postsurgical changes in these variables may be influenced by the surgical technique, specifically SSRO and IVRO. Despite the high prevalence of SSRO in the included studies, the surgical technique may have influenced the outcomes related to skeletal symmetry. Second, although the subgroup analysis was performed based on predefined variables, high statistical heterogeneity was observed in most groups. One possible explanation for this is the limited number of studies on these variables. Distinct types of facial asymmetries can be present, including rotation, translation, and yaw. However, the specific type of asymmetry and amount of surgical movement in the jaws were not consistently defined across the included studies, which may have contributed to the heterogeneity of the outcome parameters. Finally, as we focused on the outcome of surgery in patients with facial asymmetry, additional factors concerning skeletal stability, such as positional changes of the proximal segment and condylar head remodeling, were not studied. Further studies are warranted to focus on factors associated with long-term stability.