2.6. Synthesis Method

Studies were grouped according to the methods of surface roughening. The SBS groups were (1) abrasion-no silane-bonding-metal brackets, (2) abrasion- [29, 35] HF-silane-bonding-metal brackets, (3) abrasion-silane-bonding-metal brackets, (4) Cojet- (a method of silica coating) silane-bonding-metal brackets, (5) diamond bur-no silane-bonding-metal bracket, (6) diamond bur-silane-bonding-metal bracket, (7) Er:YAG laser- (1.6 W, 20 Hz) silane-bonding-metal bracket, (8) HF-no silane-bonding-metal bracket, (9) Nd:YAG laser- (1 W) silane-bonding-metal bracket, (10) phosphoric acid 37%-silane-bonding-ceramic bracket, (11) phosphoric acid 37%-silane-bonding-metal bracket.

All the SBS studies except two had reported mean and standard deviations for all their groups. A study had reported median and the interquartile range instead of mean and standard deviation [30]. A formula ((Q3 − Q1)/1.35) [42] was used to convert the range into standard deviation. Also, another formula ((median = Q1 = Q3)/3) [42] was used to convert the median/quartile information into the mean. A study had reported merely mean SBS values without standard deviations [43], which was excluded from meta-analyses pertaining to SBS. All ARI scores had been reported in a way that raw data could be obtained from the presented data.

All the assessed studies were summarized as tables and also as forest plots. Heterogeneity was assessed using various measures including the I² statistic. The source of heterogeneity was not statistically assessed, since the studies were all in vitro and usually no study variables except for the main independent and dependent variables existed in each study, and also because many metasamples were small. For sensitivity analysis, forest plots and sample sizes were visually inspected by two statisticians, and almost no cases of extremities were found. Therefore, no statistical sensitivity analyses deemed necessary. Publication bias was assessed using the Egger regression.