Literature results
Database searches yielded 4122 potentially relevant entries. After removing duplicated publications by Endnote X9, 3063 articles remained. Among the remaining articles that qualified for the title and abstract review, 2530 were excluded because of their apparent irrelevance. After reading the full text, 81 pieces of literature were finally included, with a total of 27227 eyes (Fig. 1).
Basic Characteristics And Methodological Quality Assessment Of Included Studies
The basic information of the included literature is shown in the Supplement (eTable1 in the Supplement), and all were RCTs.
The Jadad score of each study is more than or equal to 3 points, with 11 studies having 7 points, 15 studies having 6 points, 20 studies having 5 points, 13 studies having 4 points, and 22 studies having 3 points. All the studies with 3 points were unable to achieve double-blind due to the different physical properties of the interventions. Altogether, the overall quality of the included articles is acceptable.
Cochrane Risk Bias Assessment is shown in the Supplement (eFigure1 in the Supplement).
Network Meta-analysis Results
Evidence network
Based on direct comparison data, relationships were formed between all interventions. Each vertex of the relationship diagram represents different interventions, the vertex size represents the sample size included in each intervention, the line between the vertices indicates the direct comparison between the two interventions, and the width of the line is proportional to the number of studies of each pair of interventions. There is direct or indirect evidence between different interventions, and the basic conditions for performing network meta-analysis are met (Fig. 2).
Consistency Test
Inconsistency model test results for axial length and refraction showed P > 0.05, indicating that the global inconsistency was insignificant. Thus, a consistency model can be used to fit.
The node splitting method was used to test for local inconsistency for the two outcomes of axial length and refraction separately (eTable2 in the Supplement). The results of axial length showed no inconsistency between any two interventions (P > 0.05). The refraction results showed local inconsistency between “control-1%Atropine + Racanisodamine” and “1%Atropine-1%Atropine + Racanisodamine (P < 0.05) and no local inconsistency between any two remaining interventions (P > 0.05).
Heterogeneity Test
Any two interventions of axial length and refraction were analyzed by fixed effect model and random effect model, respectively, and the results showed that the two models had consistent results, suggesting that the heterogeneity of the included studies was small (eTable3 in the Supplement).
Axial
Seventy-two studies reported axial length changes, totaling 35 interventions. Network meta-analysis results showed that except for progressive addition spectacle lenses, axial length growth after treatment by other interventions was less than control. Among them, four interventions such as orthokeratology (-0.36[-0.53, -0.20], P < 0.05), 1%Atropine (-0.39[-0.65, -0.14, P < 0.05), 1%Atropine + bifocal spectacle lenses (-0.60[-1.1, -0.13], P < 0.05) and 0.01%Atropine + orthokeratology (-0.47[-0.79, -0.14], P < 0.05) were statistically significant from control (Fig. 3A). Network meta-analysis results of axial length changes are presented in the Supplement (eTable4 in the Supplement).
For slowing axial progression, the cumulative probability ranking results for each intervention showed that the top ten were: 1%Atropine + bifocal spectacle lenses (88.6%) > 0.01%Atropine + orthokeratology (80.9%) > 1%Atropine (75.9%) > 0.01%Atropine + single vision spectacle lenses (74.0%) > orthokeratology (72.9%) > 1%Atropine + progressive multi-focal spectacle lenses (68.2%) > 0.5%Atropine (67.8%) > 0.1%Atropine (65.8%) > 0.05%Atropine (64.7%) > 1%Atropine + Racanisodamine (64.0%), suggesting 1%Atropine + bifocal spectacle lenses may be the most effective measure to delay axial growth (Fig. 4A).
Diopter
Sixty-eight studies reported refractive changes, totaling 36 interventions. The results of network meta-analysis showed that except for blue-violet light filtering spectacle lenses, undercorrected single vision spectacle lenses, compound tropicamide eye drops, and 1%Atropine + Racanisodamine, the progress of refraction after treating by other interventions was less than control; among them thirteen interventions such as progressive multi-focal spectacle lenses (0.42[0.00, 0.85], P < 0.05), orthokeratology (0.55[0.31, 0.80], P < 0.05)、progressive multi-focal soft contact lenses (0.78[0.20, 1.4], P < 0.05), 1%Atropine (0.43[0.11, 0.75], P < 0.05), 0.5%Atropine (0.69[0.24,1.1], P < 0.05), 0.1%Atropine (0.43[0.11, 0.75], P < 0.05), 0.05%Atropine (0.56[0.24, 0.89], P < 0.05), 0.01%Atropine (0.34[0.13, 0.54], P < 0.05), 1%Atropine + bifocal spectacle lenses (1.1[0.55, 1.6],P < 0.05), 1%Atropine + progressive multi-focal spectacle lenses (0.66[0.17,1.1], P < 0.05), 0.01%Atropine + single vision spectacle lenses (0.70[0.18, 1.2], P < 0.05), 0.01%Atropine + orthokeratology (0.81[0.38, 1.2], P < 0.05), bifocal spectacle lenses + Massage (0.85[0.16, 1.5], P < 0.05) were statistically significant from control (Fig. 3B). Network meta-analysis results of refraction changes are presented in the Supplement (eTable4 in the Supplement).
For slowing refraction progression, the cumulative probability ranking results for each intervention showed that the top ten were: 1%Atropine + bifocal spectacle lenses (94.8%) > 0.01%Atropine + orthokeratology (85.7%) > bifocal spectacle lenses + Massage (84.3%) > progressive multi-focal soft contact lenses (82.0%) > 0.5%Atropine (78.7%) > 0.01%Atropine + single vision spectacle lenses (77.9%) > 1%Atropine + progressive multi-focal spectacle lenses (75.9%) > 0.05%Atropine (70.9%) > orthokeratology (69.6%) > single vision spectacle lenses + Red light (69.2%), suggesting that 1%Atropine + bifocal spectacle lenses may be the most effective measure to slow refraction progress (Fig. 4B).
Cluster Analysis For Ranking Of Two Outcomes
Cluster analysis of the cumulative probabilities of 34 common interventions showed that all interventions fell into 4 categories, with increasing effectiveness in delaying myopia progression from top right to bottom left; there were a total of 12 interventions in the top right category, of which 5 (41.67%) were combination interventions and 4 (33.33%) were Atropine. This result suggested that combination interventions may most effectively delay myopia progression (Fig. 5).
Publication Bias
From the funnel plot of axial length and refraction, it can be found that each point is scattered and asymmetrical incompletely, suggesting that there may be some publication bias. Some scattered points are distributed at the bottom of the funnel plot of each study indicator, indicating a small sample effect (eFigure2 in the Supplement).